Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
This thread still begs the issue of benchmarking a representative business application across multilple platforms. It may beg the issue, but the answer isn't (solely, or even much) a benchmark. Running a single payroll application (for example) doesn't cut it either. These systems are built to run hundreds or even thousands of applications with different profiles on one machine. (Are there benchmarks like that?) How do you benchmark a 777 compared to a motor scooter? They both move, they both carry people, but is any sort of simple comparison possible? A single number doesn't help much. The mainframe inevitably forces the question: How do you measure the total performance of the whole IT organization, in business terms? That paragraph I wrote in response to Dean gives some idea of the dimensionality of that question. As mentioned, many outsourcing companies do a decent job making such measurements. They have to because it dictates how they bid and how much profit they can make. IMHO, IT organizations which don't do a good job in this area -- which focus way too much on SPECint :-) -- are the ripest candidates for outsourcing as the company's CEO and CFO grow thoroughly disgusted with IT mismanagement. - - - - - Timothy Sipples IBM Consulting Enterprise Software Architect Specializing in Software Architectures Related to System z Based in Tokyo, Serving IBM Japan and IBM Asia-Pacific E-Mail: [EMAIL PROTECTED] -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
Timothy, Internal customers do not ask your question, How do you measure the total performance of the whole IT organization? They are really only interested in the total cost of acquistion or internal chargeback (excluding power and environmentals) and application performance. -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
William, 1 - You state that internal customers are only concerned with chargeback and application performance, yet if that was the case, why are the putting up with web apps with 10-12 second response time and the requirement that they move their hands from the keyboard to the mouse and back repeatedly, increasing user work while they used to have a CICS screen with sub-second response time. 2 - Companies that exclude power and environmental issues (including sq foot of floor space, labor to upgrade power panels etc) in the chargeback are doing themselves a huge disservice. If you have chargeback, don't just charge for part, charge for all. Wayne Driscoll Product Developer JME Software LLC NOTE: All opinions are strictly my own. -Original Message- From: IBM Mainframe Discussion List [mailto:[EMAIL PROTECTED] On Behalf Of William Richter Sent: Tuesday, July 24, 2007 7:31 AM To: IBM-MAIN@BAMA.UA.EDU Subject: Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article) Timothy, Internal customers do not ask your question, How do you measure the total performance of the whole IT organization? They are really only interested in the total cost of acquistion or internal chargeback (excluding power and environmentals) and application performance. -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
On Tue, 24 Jul 2007 07:30:36 -0500, William Richter [EMAIL PROTECTED] wrote: Timothy, Internal customers do not ask your question, How do you measure the total performance of the whole IT organization? They are really only interested in the total cost of acquistion or internal chargeback (excluding power and environmentals) and application performance. ...which actually makes Tim's point about I/T mismanagement. To allow decisions to be made that do not take into account the overall welfare of the organization is irresponsible. The department may not be capable of evaluating the impact on all of I/T, but someone above the department level should be responsible for reconciling decisions that fly in the face of the best interest of the entire organization or company. But you probably knew that already and were just stating fact. And I agree with your assessment. I don't agree with the philosophy. Bill Seubert System z I/T Architect IBM Corporation [EMAIL PROTECTED] -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
On Jul 24, 2007, at 11:25 PM, Bill Seubert wrote: On Tue, 24 Jul 2007 07:30:36 -0500, William Richter [EMAIL PROTECTED] wrote: Timothy, Internal customers do not ask your question, How do you measure the total performance of the whole IT organization? They are really only interested in the total cost of acquistion or internal chargeback (excluding power and environmentals) and application performance. ...which actually makes Tim's point about I/T mismanagement. To allow decisions to be made that do not take into account the overall welfare of the organization is irresponsible. The department may not be capable of evaluating the impact on all of I/T, but someone above the department level should be responsible for reconciling decisions that fly in the face of the best interest of the entire organization or company. But you probably knew that already and were just stating fact. And I agree with your assessment. I don't agree with the philosophy. Gee wiz, I thought that was why the CIO was there and makes all the big bucks. Ed -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
On Sat, 21 Jul 2007 12:00:28 +, Ted MacNEIL [EMAIL PROTECTED] wrote: Could that be because it's irrelevant? There's more to a transaction than the processor speed, regardless of the platform. And THAT, friends, is the jist of the entire discussion. And it's a point that more and more customers are realizing. As Timothy pointed out in a posting just a bit ago, the clock rate is slower. Whoop de freakin' do. There is far, far more to how much work a computer system can do (note: I avoid the word processor - it's all about the whole system) than the speed of the chip. I have this discussion with customers all the time, and it doesn't take a very long discussion for them to realize that processor speed and system performance and system capacity are distinctively different concepts that are only remotely related. Bill Seubert System z I/T Architect IBM Corporation [EMAIL PROTECTED] -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
This thread still begs the issue of benchmarking a representative business application across multilple platforms. What application architecture runs across all platforms (z/series, intel, pseries) and operating systems (z/OS, Linux, Unix)? I'd like to suggest SAS. A SAS application that is both cpu and IO intensive would make an interesting benchmark, testing both the strengths and weaknesses of the various platforms and operating systems. -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
-Original Message- From: IBM Mainframe Discussion List [mailto:[EMAIL PROTECTED] On Behalf Of William Richter Sent: Monday, July 23, 2007 9:58 AM To: IBM-MAIN@BAMA.UA.EDU Subject: Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article) This thread still begs the issue of benchmarking a representative business application across multilple platforms. What application architecture runs across all platforms (z/series, intel, pseries) and operating systems (z/OS, Linux, Unix)? I'd like to suggest SAS. A SAS application that is both cpu and IO intensive would make an interesting benchmark, testing both the strengths and weaknesses of the various platforms and operating systems. snip This has been my point. But the common applications use data bases or are accounting applications, or are payroll (a different type of accounting). So a pseudo accounting system report -- say statements that need to know what it was that was ordered, what has been paid, what hasn't, how much is owed, etc. would be a good benchmark tool. And if it were written in COBOL, given that there are COBOL compilers for these platforms, it would give a good throughput demonstration and benchmark. This would drive I/O just like a business application in the real world, could give a definition as to what a transaction is, etc. Regards, Steve Thompson -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
On 20 Jul 2007 23:35:18 -0700, in bit.listserv.ibm-main you wrote: - Original Message - From: Timothy Sipples [EMAIL PROTECTED] Newsgroups: bit.listserv.ibm-main To: IBM-MAIN@BAMA.UA.EDU Sent: Friday, July 20, 2007 10:43 PM Subject: Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article) Re: Supposed factor improvements over time in the integer performance of processors, there are some faulty numbers in this discussion, or at least misleading. It has to do with cores versus chips. Dean, with all due respect, no matter how much you try to fuzz it, they're not directly comparable. The X86 architecture going to two cores and now quad cores was the only way X86 engineers could simulate a Moore's Law improvement. But the dirty little secret is that two cores most definitely does not equal doubling the clock speed of a single core. I think your math is pretending otherwise, but that's not the real world of business computing. I didn't do any math. I reported SPECint numbers - for both single core and dual core. The numbers I provided were single processor, single core results up to June 2006. Another dirty little secret is that today's typical X86 software is lousy at taking advantage of multi-cores. And yet another dirty little secret is that almost all software vendors charge more for multi-core, so moving to the supposedly higher performance multi-core design might actually raise your cost of computing. (This is a very real problem now. Single core processors are still in demand, especially for light duty test servers, development servers, branch servers, and education/training servers, in order to minimize the cost of the software.) Why is this relevant to the discussion, except as a way to again move it away from the question of processor performance. I understand the desire to defend the faith at all costs - but this is just a simple little issue. If processor performance doesn't matter, then why is the fight to defend it so fierce? Either mainframe CPUs are slower, or they are not. Instead of all these 'dirty little secrets', and 'leading technology' arguments that have nothing at all to do with the issue, except to widen the discussion so it can be 'won', we either present the figures and deal with them, or just agree to disagree. I suspect that slower or faster may depend on workload. If there is a lot of decimal arithmetic (native on z, simulated in part or all on Intel, Power and RISC), the mainframe will be a lot faster for the arithmetic part. On the other hand, IBM came out with XPLINK and other tweaks because C/C++ performance was so bad on z series. I would like to see a test with optimized COBOL and web server on the various platforms. Also a COBOL and DB2 benchmark across platforms would be useful. As someone who likes COBOL and z, I am dismayed by the probable demise of both in part because of what I believe to be bad management of both products. Oh, there's another dirty little secret. Execution errors are becoming more frequent as clock speeds increase, temperatures rise, and densities shrink. Keeping those electronics flowing in the right places is getting tougher, and more often they're leaping out of their little cages resulting in two plus two not equalling four. This is most unacceptable in the financial transaction processing world, for example, which is why IBM mainframes protect against execution errors. It's yet another metric SPECint doesn't seem to report, the long-term processor error rate. This, of course, is a red herring. We've already had Tom Marchant claim that IBM is leading in process technology, and now we are hearing that these improvements are causing increased errors that are unacceptable for mainframes. This is typically called FUD. Since a large percentage of PCs are sold with non-parity, non-ECC memory, I doubt that anyone knows the true error rate of Intel processors. I wonder how many glitches attributed to Windows are actually hardware problems. It would be instructive to know how many times a processing error is actually detected by the z hardware. If you want to look at integer performance on a benchmark, stick to per core numbers if you're comparing cores. And you'll discover that processor engineers are struggling to increase core speeds, and Moore's Law has probably stalled already. Maybe that's why Intel is cutting back on RD? :-) More FUD. As I said - I compared single chip, single core performance. This multi-core problem is not new to IBM. The solutions (plural) require a total system design perspective, including software. Such as Linux and open source. Yep, only IBM has the answer. The FUD gets more intense. I think John Gilmore is right - this thread has probably run its course. I'm a staunch mainframe advocate, but I think it's OK to give credit where it is due. I haven't seen anything to rebut the notion that mainframe processors are slower than other architectures,
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
Clark F Morris wrote: On 20 Jul 2007 23:35:18 -0700, in bit.listserv.ibm-main you wrote: snip Since a large percentage of PCs are sold with non-parity, non-ECC memory, I doubt that anyone knows the true error rate of Intel processors. I wonder how many glitches attributed to Windows are actually hardware problems. It would be instructive to know how many times a processing error is actually detected by the z hardware. snip I think the MODE STATUS command can be used to see how many machine checks of various types there have been. They should also be logged to LOGREC, though I think the MODE command may be able to modify this some. But you have to know what is being logged, I seem to recall that on the 3090 recoverable memory errors weren't logged until there had been some number of them. Of course, you are probably talking about all z machines rather than one. There used to be a service (R+ I think) that tracked this kind of info. I didn't cover all machines, but anybody that subscribed to the service sent in LOGREC info and received a summary of all machines being monitored. I don't know how accurate it was, vendors might be tempted to modify the recording process in their machines to not show some kinds of errors. -- Richard -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
OK, Dean, here you go: a single IBM System z core employs a lower clock oscillator rate than an X86 core (new 2007 v. new 2007) And that fact is approximately 0.1% relevant to the achievement of particular business outcomes within a particular time at a particular fully measured risk-adjusted cost profile with a particular set of service qualities in a particular physical space with a particular environmental impact using particular applications, middleware, and operating systems. - - - - - Timothy Sipples IBM Consulting Enterprise Software Architect Specializing in Software Architectures Related to System z Based in Tokyo, Serving IBM Japan and IBM Asia-Pacific E-Mail: [EMAIL PROTECTED] -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
- Original Message - From: Timothy Sipples [EMAIL PROTECTED] Newsgroups: bit.listserv.ibm-main To: IBM-MAIN@BAMA.UA.EDU Sent: Friday, July 20, 2007 10:43 PM Subject: Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article) Re: Supposed factor improvements over time in the integer performance of processors, there are some faulty numbers in this discussion, or at least misleading. It has to do with cores versus chips. Dean, with all due respect, no matter how much you try to fuzz it, they're not directly comparable. The X86 architecture going to two cores and now quad cores was the only way X86 engineers could simulate a Moore's Law improvement. But the dirty little secret is that two cores most definitely does not equal doubling the clock speed of a single core. I think your math is pretending otherwise, but that's not the real world of business computing. I didn't do any math. I reported SPECint numbers - for both single core and dual core. The numbers I provided were single processor, single core results up to June 2006. Another dirty little secret is that today's typical X86 software is lousy at taking advantage of multi-cores. And yet another dirty little secret is that almost all software vendors charge more for multi-core, so moving to the supposedly higher performance multi-core design might actually raise your cost of computing. (This is a very real problem now. Single core processors are still in demand, especially for light duty test servers, development servers, branch servers, and education/training servers, in order to minimize the cost of the software.) Why is this relevant to the discussion, except as a way to again move it away from the question of processor performance. I understand the desire to defend the faith at all costs - but this is just a simple little issue. If processor performance doesn't matter, then why is the fight to defend it so fierce? Either mainframe CPUs are slower, or they are not. Instead of all these 'dirty little secrets', and 'leading technology' arguments that have nothing at all to do with the issue, except to widen the discussion so it can be 'won', we either present the figures and deal with them, or just agree to disagree. Oh, there's another dirty little secret. Execution errors are becoming more frequent as clock speeds increase, temperatures rise, and densities shrink. Keeping those electronics flowing in the right places is getting tougher, and more often they're leaping out of their little cages resulting in two plus two not equalling four. This is most unacceptable in the financial transaction processing world, for example, which is why IBM mainframes protect against execution errors. It's yet another metric SPECint doesn't seem to report, the long-term processor error rate. This, of course, is a red herring. We've already had Tom Marchant claim that IBM is leading in process technology, and now we are hearing that these improvements are causing increased errors that are unacceptable for mainframes. This is typically called FUD. If you want to look at integer performance on a benchmark, stick to per core numbers if you're comparing cores. And you'll discover that processor engineers are struggling to increase core speeds, and Moore's Law has probably stalled already. Maybe that's why Intel is cutting back on RD? :-) More FUD. As I said - I compared single chip, single core performance. This multi-core problem is not new to IBM. The solutions (plural) require a total system design perspective, including software. Such as Linux and open source. Yep, only IBM has the answer. The FUD gets more intense. I think John Gilmore is right - this thread has probably run its course. I'm a staunch mainframe advocate, but I think it's OK to give credit where it is due. I haven't seen anything to rebut the notion that mainframe processors are slower than other architectures, and it doesn't seem like we are going to get there from here. Re: Token-Ring and Ethernet, yes, really lousy analogy. The progression in networking technology mainly had to do with the emergence of network switching, effectively obsoleting both Ethernet and Token-Ring. It had nothing in particular to do with Ethernet getting faster, because Token-Ring did, too (4, 16, 100 Mbps). Actually, according to the presenters it had *everything* to do with it. The point was that Ethernet was *cheap* and ubiquitous. Engineers could ratchet up the speed of the Ethernet network to overcome the inefficiency cheaper than they could ratchet up Token Ring speeds, and customers didn't care about efficiency. The reason I brought it up was because of the efficiency argument - which is, again, not the real issue. The real issue is the economics, and that was the point of the analogy. And now we come full circle, because guess what's inside even the latest System z9 mainframe? Yes, PCI, albeit far enhanced from the original. You can buy
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
I haven't seen anything to rebut the notion that mainframe processors are slower than other architectures Could that be because it's irrelevant? There's more to a transaction than the processor speed, regardless of the platform. - Too busy driving to stop for gas! -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
On 17 Jul 2007 19:21:33 -0700, in bit.listserv.ibm-main you wrote: Dean Kent wrote: - Original Message - From: Howard Brazee [EMAIL PROTECTED] Newsgroups: bit.listserv.ibm-main To: IBM-MAIN@BAMA.UA.EDU Sent: Tuesday, July 17, 2007 7:29 AM Subject: Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article) IBM is not positioning the mainframe to compete with a computer chip. Apples and bird seed. Customers of computer chips are computer manufacturers. IBM is positioning the mainframe to compete with server farms. That is something very different. And the x86 speed of a PC is not what customers care about when looking at alternatives for these business needs. I agree with your first point, but not your second. There *is* a reason that SPEC (and other benchmarking organizations) exist. These customers want a common performance metric to identify the value they are getting for the money they spend. Yes, reliability, fault-tolerance, data integrity, etc. are all factors too - but the mainframe does not have a lock on these features, other platforms do as well, including those based on x86. Maybe I can point out the dilemma better this way: People here have asked how managers can justify migrating their mission-critical applications off the mainframe and onto a 'PC'. At the same time, these people will say that there is no common metric to compare the various platforms, that they are just different. So, a manager who must make a business/financial decision is given no tools with which to make that decision - so is it any wonder that those decisions seem, well, random? One would think that *if* the mainframe can compete head-to-head with these other server systems that those in the business would *want* a common metric. They would *strive* to identify something that managers could use to make better decisions. This should, you would think, include the vendor who would benefit most by such information. Continually claiming that there is, and cannot be any comparison seems counter productive. You are *asking* these managers to go with the latest fad because they have nothing else to use as a guideline. If it can't compete, then perhaps it makes sense to claim that no such metric can be identified. As long as the workloads are completely different, then it makes sense. They they overlap, however, you are asking for people to flip a coin to choose unless you give them another tool to use. As for the car analogy in performance, I would suggest the following: I can look at horsepower, top speed, acceleration, luggage capacity, towing capacity, gas mileage and various other factors that are available for *all* vehicles. This allows me to make an intelligent, informed decision about which particular vehicle is best for my needs, whether it be a sportscar, a family vehicle, a farm vehicle or a large commercial vehicle. Instead, Timothy Sipples suggests (and I paraphrase from his reply to me) if you don't know, talk to your IBM rep - he'll tell you what you need. Sure, he'll tell me I need a Sun system instead of an IBM system - right? Or perhaps I should go talk to Sun or HP or Dell to find out what best suits my needs. If you care about the platform, you should care about the problem... or so it seems to me. Regards, Dean Can the mainframe z900/z990/z9 compete head to head with Intel? IMHO, yes. Head to head. You are talking about the same number of processors and same amount of RAM/Central Storage. We have two z990, a 304 and 303. A total of 7 (seven) CPU's and 20GB of total central stoage for the z/OS images. Now in our enviroment we do NOT have a test mainframe, we do not have test LPAR's. All production/test/development/QA/user accecptance testing are done with the same LPARs. There is a system programmer sand box on the 303. We are planning to migrate 80% of our workload off the mainframe on to Intel. If the Intel processors were really faster/better than the z990 CPU, then we sould be to get a single Intel box with 6 CPUs to replace our two mainframes. Right? Remember this is head to head. Are we? NOPE. In the end to replace 80% of 7 z990 CPU's and 20GB of RAM the Intel side will have Right now the plan is to ONLY have 96 total cores (some systems will have single core processors and some will have dual core) CPU's and 456 GB of RAM. This is a est. and they beleive that they may need to increase this by as much as 50%. So the head to head comparsion is 7 CPU's and 20GB to 96 CPU's and 456 GB of RAM. Doesn't seem to head to head to me. That is a about a 13:1 ratio on CPU's and 23:1 on RAM. If Intel was faster then we should be able to do more work on less processors. Please show me a site that has migrated off a modern day mainframe to Intel using the same number of CPU's and same amount of Central Storage/RAM as they had on the mainframe. You know
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
- Original Message - From: john gilmore [EMAIL PROTECTED] Newsgroups: bit.listserv.ibm-main To: IBM-MAIN@BAMA.UA.EDU Sent: Friday, July 20, 2007 1:04 PM Subject: Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article) Unless you two are prepared to formulate your positions carefully enough so that we can have a clear notion of what you are disagreeing about, this thread is going to continue to generate much heat and no light (and should be killed). Alright, excellent suggestion - though I'm not sure I will be able to provide as much specific detail as some may want. I have been involved with discussion about non-mainframe CPUs for several years on various forums and newsgroups, such as comp.arch, Ace's Hardware, etc.. Mostly lurking, but occasionally contributing. During this time, I have repeatedly seen the opinion that mainframe processors are slower than x86, Itanium and RISC processors (which include PA-RISC, SPARC, POWER and the now-defunct Alpha). The reasoning given is that in those markets, processors are routinely compared based upon their integer and floating point computing speed, and that mainframe processors simply cannot compete in these metrics. There is general agreement that mainframes are very good at what they do - which is business data processing that consists of a lot of I/O.It is also generally believed that the mainframe does OK at transaction oriented computing but that distributed platforms are much better for this. In fact, there are some who believe that clustered computing is better than the mainframe even in the traditional mainframe applications. So, when I saw Steve Thompson's question about why people believe mainframe processors are slower, I posted that they are. This was based upon information gathered from people who are involved with chip design, and whom I presumed had better information than I. As I indicated in my recent reply to Ed Jaffe - if anyone has data that would contradict that perception, I would be very happy to hear it. I would love to be able to prove that mainframe processors are not slow, even if they are not faster. To this point, however, what I have heard is a lot of orthogonal discussion that does nothing at all to address the specific question of processor speed. In other words, whether it actually is important or not to mainframes, it is a question that many *do* believe is important. Since they do (and they are either decision makers, or influence decision makers), that makes it important in my opinion. And therefore, is useful to discuss. And, while it may sound like a wonderful debating technique, simply saying You made the assertion, now prove it doesn't further the discussion at all. In fact, it is akin to saying If you can't prove it, then it isn't true, which is, of course a fallacy. Just because something can't be proven does not mean it isn't true. You establish its falsehood by providing the facts showing it to be false. Otherwise, it is *still* a valid opinion to hold, however annoying it may be. Regards, Dean -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
- Original Message - From: Edward Jaffe [EMAIL PROTECTED] Newsgroups: bit.listserv.ibm-main To: IBM-MAIN@BAMA.UA.EDU Sent: Friday, July 20, 2007 11:54 AM Subject: Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article) I really wasn't paying any attention to this thread. But, I happened to read the non sequitur you posted (see below), and thought I should respond. And I appreciate the information. I wasn't aware of it before, so all I had was the article's information about installed MIPs. I realize that this is not a true comparison, and I stated that - but it was all I had at the time. I'm happy that I did, as it seemingly encouraged you to provide a link to much better information. First, you compare installed MIPS with performance. Nothing need be said about that. Non sequitur. Then you asked what the highest performing mainframe was in 2000 vs today. I believe I answered that question correctly. I didn't say your information was wrong, just that it didn't address the original point. I'm not sure I follow your reasoning. Part of mainframe processing power lies in its ability to do effectively use SMP (up to 54-ways now ... more later). It's part of the equation. Good benchmarks are based on computing throughput ... not individual chip speed. If you are comparing processors (which was the basis of the original question), then the compute speed of the chip *is* relevant. If you are comparing systems, then your benchmarks simply have to be testing the same thing (memory bandwidth/latency, I/O rate, transaction rate, etc.). Again - the question was Why do people think that mainframe processors are slower?. . Read your own statement quoted up above. You originally said x86 had an 8-fold increase over seven years. (Is that chip speed or actual server speed?) Actually, I was rounding a bit. My original post on the subject indicated that the span for x86 comparisons was Nov 2000 thru June 2006 (as I was trying to use roughly the same period as in the post I was responding to at the time), and the increase was from 5.8 to 63.6, so the numbers should be a 7.95x increase in just over 5 1/2 years. Now you're saying it had an 8-fold increase in just 4 1/2 years -- the time frame over which the mainframe had a 6.6:1 processor speed increase (or 2.5:1 individual CP speed increase if that's what you choose to focus on). However you slice it, x86 processor performance has increased at roughly twice the rate as mainframe processor performance. I know that some here will take this as an insult, or a put down, or some attempt to make the mainframe look bad. It is, however, simply a reasonable conclusion based upon the available evidence. If someone has different data, please present it rather than just taking umbrage and arguing about little details that are not really important. If we want to talk fallacies, that's known as a Straw Man. I wasn't aware that people think mainframe processors are slow compared to others. Is this from a survey of some kind? Or perhaps from that IT Jungle article you referenced? Can you post the URL? I'm including below the full text of the original post by Steve Thompson that got me onto (into?) this whole thread: --- -Original Message- From: IBM Mainframe Discussion List [mailto:[EMAIL PROTECTED] On Behalf Of Tim Hare Sent: Friday, July 13, 2007 1:52 PM To: IBM-MAIN@BAMA.UA.EDU Subject: Links to decent 'why the mainframe thrives' article http://arstechnica.com/news.ars/post/20070711-legacy-matters-why-the-ibm -mainframe-continues-to-thrive.html referenced article http://www.itjungle.com/big/big071007-story01.html SNIP I keep seeing references to the mainframe processor is slower than those used in other platforms. Seriously, in an effort to compare processor power, IF one were to take a COBOL program that would process 1000 records from a data base and produce a report (let's say a payroll check register), which system would process this in the least amount of elapsed time? I ask this question in this fashion, because I know that Fujitsu has a COBOL compiler that produces code that runs under Windows (I know, because I have used it to do batch reporting at one time). I understand that a similar compiler is available for other of the platforms. So, if we run the data base and applications system on a self-contained system, which one will run with the lowest wall time? This is the kind of benchmark that needs to be done. It, in my opinion, is the only way to get close to a valid comparison. Anyone else have any ideas? [I still have the compiler, and I still have w/2K that I fixed it to operate with if anyone would like to try to build a benchmark.] Regards, Steve Thompson -- For IBM-MAIN subscribe / signoff / archive access instructions, send
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
To this point, however, what I have heard is a lot of orthogonal discussion that does nothing at all to address the specific question of processor speed. I don't think it's orthogonal. IMO, MIPS is just one component in a complex environment. If I could pull out and plug in the processor in a working (real) environment, then maybe I could be convinced that we are measuring something meaningful. The true metrics are: 1. Are we delivering business need? 2. Are we delivering this need effectively? We wonder why management gets hooked on speeds and feeds! I think it is because we do! - Too busy driving to stop for gas! -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
x86 processor performance has increased at roughly twice the rate as mainframe processor performance. Yes, but. We run mainframes at 100%. We run wintels at under 20%. - Too busy driving to stop for gas! -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
I for one would be interested in just why one would buy a mainframe if not for the processing power, and scalability, not to mention IO. I am involved with a data center planning activity, and the Intel servers and RS6000's consume far more cooling and power than the little old z/BC we have. Sooner or later you have to limit the proliferation of servers; I suspect that is the allure of virtualization. Just my $.02 Doug At 03:41 PM 7/20/2007, you wrote: On Fri, 20 Jul 2007 11:05:52 -0700, Dean Kent wrote: I don't think there's anyone who buys a mainframe for its sheer processing power. These servers *are* purchased, in part, for their processing power. snip -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html Doug Fuerst Consultant BK Associates Brooklyn, NY (718) 921-2620 (Office) (718) 921-0952 (Fax) (917) 572-7364 (Cell) [EMAIL PROTECTED] -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
Dean Kent wrote: I'm not sure I follow your reasoning. Part of mainframe processing power lies in its ability to do effectively use SMP (up to 54-ways now ... more later). It's part of the equation. Good benchmarks are based on computing throughput ... not individual chip speed. If you are comparing processors (which was the basis of the original question), then the compute speed of the chip *is* relevant. If you are comparing systems, then your benchmarks simply have to be testing the same thing (memory bandwidth/latency, I/O rate, transaction rate, etc.). I think part of the confusion comes from a terminology problem. Precise up-front definitions of what's meant by the words processor and system are important. Perhaps better (i.e., more universally-understood terms) should be chosen. To get a feel for how IBM uses these terms in hardware manuals, I picked up System z9 Business Class Installation Manual for Physical Planning GC28-6855-01 and found: The use of the terms server, processor, system and all models in this publication refer to the IBM System z9 Business Class. IBM seems to mostly use the term Central Processor (CP) to refer to what you seem to simply calling a processor. And, it sounds like you use the word system to refer to the CEC. I've not heard that usage before. Of course, in my world, a system is usually a z/OS image... -- Edward E Jaffe Phoenix Software International, Inc 5200 W Century Blvd, Suite 800 Los Angeles, CA 90045 310-338-0400 x318 [EMAIL PROTECTED] http://www.phoenixsoftware.com/ -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
Actually, I was rounding a bit. My original post on the subject indicated that the span for x86 comparisons was Nov 2000 thru June 2006 (as I was trying to use roughly the same period as in the post I was responding to at the time), and the increase was from 5.8 to 63.6, so the numbers should be a 7.95x increase in just over 5 1/2 years. Sorry, but this kind of comparison is complete rubbish. Since the upper bound is the speed of an electron, then clearly the closer one gets to that speed the smaller the incremental improvement. Correspondingly, the only way to achieve large increases in chip speed is to have large gains to make. Any other increases in speed will become dependent on architecture, in which case to discuss chips alone is fundamentally meaningless. -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
-Original Message- From: IBM Mainframe Discussion List [mailto:[EMAIL PROTECTED] On Behalf Of Dean Kent Sent: Friday, July 20, 2007 4:18 PM To: IBM-MAIN@BAMA.UA.EDU Subject: Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article) snip Frankly, I think it is a very commonly held belief amongst many who are very knowledgable about processors. In fact, my original comment that they are slower was based largely upon discussions to this effect on various fora where CPU architecture is a common topic - often by those who are involved with chip design. I would be *very* happy to be able to report to these individuals that I have contradictory information... in the form of factual data, rather than just dismissive comments, of course. :-). Regards, Dean SNIP I am amazed that those who know processors would be that ignorant of data bus size on effective speeds or microcode word sizes, etc. Think about this for a moment. If your data bus is 32 bits wide, a move instruction (assuming boundary alignments here) can only move 4 bytes per unit of operation. Now, the G6 (was it?) went to 256 BYTES wide. How long to move data from one record to another, assuming 4K records, between those two situations (as in number of cycles)? Next we have arithmetic. Ok, up until recently the registers were the same size, and so fixed binary arithmetic is done faster on a system with a faster clock. But the majority of IBM mainframe arithmetic is done with PACKED DECIMAL (that Decimal Feature from S/360 days). While a bit slower than Fixed Point, it is much faster than Floating Point. As a result of the Decimal Feature, edit of numbers for placement on a report is done in microcode on IBM's mainframe platforms. I do not know of a similar instruction for the x86 chips (but then, I'm not an x86 ALC programmer). [Any body want to write a C++ class (?) to emulate this?] Memory access is yet another thing. Mainframes, in order to access 2GB of memory had to have expanded registers under the covers (ever since you could have more than 2GB on a machine). Virtual storage (DAT) has been fine tuned on mainframes for years. Then there are the cache line sizes and life in cache, etc. All of this plays to the ability to do LPARS and now 64bit addressing. Microcode and/or circuitry issues. Next, with DAT, are the algorithms for MP communication (E.g. for mainframe: SIGP, CD, CSD, PLO, PLT, IPTE, SSKE, etc.), etc., that all have to be there for CPU and CEC communication so that integrity of data/operations are maintained. All of this involve both circuitry and microcode -- or the SCP must provide routines to do it. A dual core CPUs actually cause themselves a bottleneck if they only have ONE bus to communicate to/with the outside world. This becomes another design issue that while producing apparent processor speed has to balance out against the ability of the bus (or buses) to handle data flow (which gives effective processor speed). Multi-CPU (multiple CPU chips) communications require motherboards and other chips to help solve simultaneous update issues (corrupted data) because of race conditions. These are more circuitry and/or microcode issues. The other thing that is not looked at (and I do not know the answer for IBM's chips), is the apparent speed of the clock vs. the microcode clock speed. There are machine designs where the microcode is running 2-8 times the speed of the system clock, so that you have an apparent speed. These things are also part of the RISC vs. CISC designs and arguments. Obviously the IBM mainframes have been Complicated Instruction Set based systems for a very long time (in the grand scheme of things). And it has been years since I worked at Amdahl where some of this was discussed from time to time. So I'm starting to forget issues. Regards, Steve Thompson -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
A dual core CPUs actually cause themselves a bottleneck if they only have ONE bus to communicate to/with the outside world. This becomes another design issue that while producing apparent processor speed has to balance out against the ability of the bus (or buses) to handle data flow (which gives effective processor speed) According to somebody I know from IBM Canada's Lab, all mainframe chips are dual core, but better. They don't only duplicate the core, but all connections to the world outside the chip. Including buses. I guess they learned after AP's in the late 1970's. - Too busy driving to stop for gas! -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
For example, the highest performing single-core Opteron (HP Proliant System) gets a 14.0 in SPECint peak, and 12.7 base. The best single-core Xeon result (a Bull system), is 17.4 peak and 16.8 base. The best dual-core Opteron submission is 14.9 peak and 13.5 base. The best dual-core score for Xeon is 18.1 and 17.5 respectively. SPECint_rate is a throughput benchmark vs. raw integer computing performance (meaning, it is intended for multi-CPU systems). The best Opteron dual-core score is 27.0 peak, 24.3 base. The best Xeon score is 30.8 and 29.3 respectively. With 4 cores (two dual-cores, since AMD doesn't have a native quad-core result), Opteron (in a Sun system) gets 60.4 peak and 51.6 base, whereas Xeon has a 59.4 peak and 56.7 base. Once again, I think these numbers are meaningless when compared to mainframes. From the SPECint site: The CPU2006 benchmarks (code + workload) have been designed to fit within about 1GB of memory (when compiled with 32 bit pointers), i.e. within the capabilities of systems that allow user applications to use 32 bits (4GB). (SPEC is aware that some systems that are commonly described as 32-bit may provide a smaller number of bits to user applications, for example if one or more bits are reserved to privileged code. SPEC is also aware that there are many ways to spend profligate amounts of virtual memory. Therefore, although 32-bit systems are within the design center for the CPU2006 suites, SPEC does not guarantee any particular memory size for the benchmarks, nor that they will necessarily fit on all systems that are described as 32-bit.) -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
-Original Message- From: IBM Mainframe Discussion List [mailto:[EMAIL PROTECTED] On Behalf Of Ted MacNEIL Sent: Friday, July 20, 2007 5:59 PM To: IBM-MAIN@BAMA.UA.EDU Subject: Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article) A dual core CPUs actually cause themselves a bottleneck if they only have ONE bus to communicate to/with the outside world. This becomes another design issue that while producing apparent processor speed has to balance out against the ability of the bus (or buses) to handle data flow (which gives effective processor speed) According to somebody I know from IBM Canada's Lab, all mainframe chips are dual core, but better. They don't only duplicate the core, but all connections to the world outside the chip. Including buses. I guess they learned after AP's in the late 1970's. SNIP Oh yeah. Between AP [AP Short], AP Long (kinda Dyadic Processor vs. Dual Processor, as I recall) that got us to MP, a few lessons were learned. And if I remember correctly, Amdahl (before my time there) had started what became XA I/O with MDF (Multiple Domain Feature) which answered even more problems in this area. Oh, and before I forget it, PR/SM was IBM's answer to MDF, not vice versa. I think IBM was actually the second paying customer for MDF. Later, Steve Thompson -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
Oh, and before I forget it, PR/SM was IBM's answer to MDF, not vice versa. I think IBM was actually the second paying customer for MDF. I worked at an alpha site of MDF. It was about three years before PR/SM. The problem(s) back then were simple. MDF did not talk to RMF, until RMF could handle PR/SM reporting. MDF was originally implemented with wait-complete=YES. Once those were cleaned up, and HLPF came along, all three competed well. I once wrote an article about PR/SM: April 2005: LPAR and PR/SM http://tinyurl.com/caf3x - Too busy driving to stop for gas! -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
- Original Message - From: Ted MacNEIL [EMAIL PROTECTED] Newsgroups: bit.listserv.ibm-main To: IBM-MAIN@BAMA.UA.EDU Sent: Friday, July 20, 2007 2:52 PM Subject: Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article) x86 processor performance has increased at roughly twice the rate as mainframe processor performance. Yes, but. We run mainframes at 100%. We run wintels at under 20%. I know this is not a good comparison, but...I am reminded of a SHARE session I attended many years ago about TCP/IP. I don't recall all of the details, but it went something like this: The networking/communication folks from IBM were giving a session, and started talking about how IBM had once upon a time insisted that more efficient networks were simply better. SNA and Token Ring were *s* much more efficient, and could run at 100% utilization, while Ethernet started getting performance problems at a mere 40%. What they realized eventually, was that the performance improvements in Ethernet were happening so fast, and that the technology was so cheap - that Ethernet and TCP/IP were going to kill SNA and Token Ring. They realized that IBM had great technology, but lost the war because they failed to consider the economics of it. As you can see, SNA and Token Ring are no more. I hope that this lesson has not been lost on IBM... Regards, Dean - Too busy driving to stop for gas! -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
Re: Supposed factor improvements over time in the integer performance of processors, there are some faulty numbers in this discussion, or at least misleading. It has to do with cores versus chips. Dean, with all due respect, no matter how much you try to fuzz it, they're not directly comparable. The X86 architecture going to two cores and now quad cores was the only way X86 engineers could simulate a Moore's Law improvement. But the dirty little secret is that two cores most definitely does not equal doubling the clock speed of a single core. I think your math is pretending otherwise, but that's not the real world of business computing. Another dirty little secret is that today's typical X86 software is lousy at taking advantage of multi-cores. And yet another dirty little secret is that almost all software vendors charge more for multi-core, so moving to the supposedly higher performance multi-core design might actually raise your cost of computing. (This is a very real problem now. Single core processors are still in demand, especially for light duty test servers, development servers, branch servers, and education/training servers, in order to minimize the cost of the software.) Oh, there's another dirty little secret. Execution errors are becoming more frequent as clock speeds increase, temperatures rise, and densities shrink. Keeping those electronics flowing in the right places is getting tougher, and more often they're leaping out of their little cages resulting in two plus two not equalling four. This is most unacceptable in the financial transaction processing world, for example, which is why IBM mainframes protect against execution errors. It's yet another metric SPECint doesn't seem to report, the long-term processor error rate. If you want to look at integer performance on a benchmark, stick to per core numbers if you're comparing cores. And you'll discover that processor engineers are struggling to increase core speeds, and Moore's Law has probably stalled already. Maybe that's why Intel is cutting back on RD? :-) This multi-core problem is not new to IBM. The solutions (plural) require a total system design perspective, including software. Re: Token-Ring and Ethernet, yes, really lousy analogy. The progression in networking technology mainly had to do with the emergence of network switching, effectively obsoleting both Ethernet and Token-Ring. It had nothing in particular to do with Ethernet getting faster, because Token-Ring did, too (4, 16, 100 Mbps). Both camps had the wrong formula originally, and bits of each survived and converged. The Token-Ring camp was entirely correct that Ethernet collision handling sucked, and so that got bypassed and completely replaced with something at least closer to Token-Ring's idea (i.e. switching) along with full-duplex. Ethernet won the logical signalling over the now non-colliding short hop to the switch. And neither camp got the wiring right at first, something of great import to actual, real businesspeople like facilities managers. Both camps changed the wiring to something else, though IBM's 1984 wiring foundation was over-engineered and could be recycled. I've been in so many buildings that very successfully installed RJ-45 adapters and kept right on going, to this very day, while traditional Ethernet's wiring usually had to be chucked completely. There's another case like Token-Ring v. Ethernet: the contemporaneous battle between Micro Channel Architecture (MCA) and AT bus (ISA/EISA), both invented by IBM (except the EISA extension), oddly enough. Which one won? Answer: neither. They're both gone. Bits of each got folded into PCI, Cardbus, ExpressCard Except maybe MCA did win in spirit. PCI has much more in common with MCA. When PCI first came out, people even spotted that, remarking that it was simply a reinterpretation of MCA. And now we come full circle, because guess what's inside even the latest System z9 mainframe? Yes, PCI, albeit far enhanced from the original. You can buy CryptoExpress2 adapters, for example, to go in those slots. - - - - - Timothy Sipples IBM Consulting Enterprise Software Architect Specializing in Software Architectures Related to System z Based in Tokyo, Serving IBM Japan and IBM Asia-Pacific E-Mail: [EMAIL PROTECTED] -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
If you run Linux for your OS and Apache for your web server and PostgreSQL for your database, you won't have licensing costs. You won't get paid support, of course, but that's a compromise that many sites seem willing to make. -- Jack Hamilton Management Information Analysis - Analytic Information Services Kaiser Foundation Health Plan, Inc. 1950 Franklin Street, Oakland, California 94612 +1 510 987-1556 (KP tieline 8-427-1556) NOTE: This email document and attachments are covered by CA Evidence Code §1157 and CA Health and Safety Code §1370. NOTICE TO RECIPIENT: If you are not the intended recipient of this e-mail, you are prohibited from sharing, copying, or otherwise using or disclosing its contents. If you have received this e-mail in error, please notify the sender immediately by reply e-mail and permanently delete this e-mail and any attachments without reading, forwarding or saving them. Thank you. From: IBM Mainframe Discussion List On Behalf Of Ted MacNEIL If your point 1 were accurate, management would dump those servers. Two points that substantiate the TCA argument (or, to be more presice ICA (initial cost of acquisition)): 1. Mainframe DASD costs more than midrange disk. (Even IBM's DS series). 2. IFL's cost arounnd 100K (US). (I can buy a lot of servers for that price -- a manager I know). Yep And each one of those servers requires a physical connection to electrical power; each requires an operating system license; each one that runs a database requires a DBMS license, etc., etc. Oh, did you want Test/Development and QA copies too? More $. -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
- Original Message - From: Chase, John [EMAIL PROTECTED] Newsgroups: bit.listserv.ibm-main To: IBM-MAIN@BAMA.UA.EDU Sent: Wednesday, July 18, 2007 11:21 AM Subject: Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article) Yep And each one of those servers requires a physical connection to electrical power; each requires an operating system license; each one that runs a database requires a DBMS license, etc., etc. Oh, did you want Test/Development and QA copies too? More $. It is also true, assuming you already have a z/Box, that you can run a lot of servers on one IFL, with ONE operating system license, ONE DBMS license, ZERO physical connections to electrical power, etc., etc. Depending on the application set, the break-even crossover point could occur at as low as 5 servers (or server images). And for T/D and QA copies, just copy a few files, share a few others, create user IDs and profiles, and XAUTOLOG them on. No extra $. Is this actually true? If the servers are running Linux, then the OS cost is the same (except, perhaps, for support fees). Also, isn't it typical that you run a database server, and connect to it rather than have all the servers run a copy of the database... otherwise, how are you getting away with only one DBMS license for a lot of Linux instances? The admin (people) costs are certainly greater, but I am confused as to why the software costs would be. Regards, Dean -jc- -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
Dean Kent wrote: I am comparing the pace of improvement in x86 with the mainframe. The IT Jungle article says that installed mainframe MIPS has increased 4 fold over 7 years. I showed that in the same time period, x86 performance has increased over 8 fold. We still aren't getting an apples to apples comparison because it doesn't tell us what the highest performing mainframe processor was in 2000 vs today. Installed MIPS has nothing to do with performance. I'm not sure why it's been mentioned here. In any case, raw MIPS improvements are easily researched. For example, the charts at http://www.tech-news.com/publib/ show the largest zSeries processor in 2000, made available in 1Q00, was a 2064-116, capable of delivering 2694 MIPS; the largest System z processor today, made available in 4Q05, was a 2094-754, capable of delivering 17,802 MIPS. That's a 6.6:1 improvement in less than five years. (Not bad!) -- Edward E Jaffe Phoenix Software International, Inc 5200 W Century Blvd, Suite 800 Los Angeles, CA 90045 310-338-0400 x318 [EMAIL PROTECTED] http://www.phoenixsoftware.com/ -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
- Original Message - From: Scott Rowe [EMAIL PROTECTED] Newsgroups: bit.listserv.ibm-main To: IBM-MAIN@BAMA.UA.EDU Sent: Wednesday, July 18, 2007 11:50 AM Subject: Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article) Well of course, Dean, Timothy makes perfect sense. You should believe everything your vendors say, without any verification, right? ;-) At the last shop I worked at (a government shop), they had a study done to evaluate what platform they should be running on, and the study proposed Oracle running on Sun servers. I later found out (to my utter lack of surprise), that the study had been done free of charge. I will give you exactly one guess who it was that did the study gratis. Of course, the conclusion was the one that was desired by those that asked for the study (upper management), who are convinced that the mainframe is obsolete and expensive. The study is being used as a justification to spend 10's of millions of dollars to replace the mainframe. Of course, the complete lack of public benchmarks for the mainframe make it impossible to refute the performance conclusions of the so called study, so I think it actually hurts IBM, even if the benchmarks would not be in favor of the mainframe. IBM's refusal to submit (or allow others to do so) mainframe benchmark scores allows others to make wild claims about how slow mainframes are, without any way to refute such claims, leaving us (the mainframe proponents) completely unarmed. This is actually sort of how I view it. Even though stating that the lack of performance numbers means it must be slower is a fallacy, many people *do* fall victim to such fallacies, even otherwise intelligent people. Marketing people are obviously very good at pointing at the positives of their own product, while highlighting the negatives of the competition - and why they make the 'big bucks' (or at least, get the big budget). For example, let us consider the concept that performance does not matter on the mainframe to be absolutely true and valid. So what does it matter then? I can think of one reason - and it has to do with business. What customers generally want is a good price/performance ratio. Benchmarks like TPC-C provide such a figure. Once you get into that type of competitive environment, your profits run the risk of being squeezed as the competitors start pointing to your price/performance and ignore the other factors. So, from a marketing perspective, I would not expect IBM (nor any IBM employee) to support the idea of having a benchmark comparing the performance of mainframe processors to others (including their own POWER). OTOH, people *do* have a perception that the mainframe is slower. Therefore, not addressing it allows the marketing people from the competition to take advantage of that perception. So, while your profits remain high your installed base shrinks. The way that is being addressed is to offer specialty engines, which have successfully increased the installed base - but as you can see, those offerings have to be at a low price to compete with those they are directly compared against. So, I understand IBM's situation, but as a 'techie' I just want to be able to make the comparisons, as do other techies when discussing issues with their decision makers. Regards, Dean -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
Tom Marchant [EMAIL PROTECTED] wrote in message news:[EMAIL PROTECTED] On Tue, 17 Jul 2007 23:51:48 -0700, Dean Kent wrote: ...As I said, Intel, for example, does not use SOI and seems to be producing very high clockrate devices without it, which (currently) outperform their rival AMD, Another assertion without data to back it up. I hope you are not talking about clock rates. The only one who has *ever* talked about clock rates has been you. If, as you have asserted, you are familiar with various platforms then this information should not come as any surprise to you at all. I've already provided a link to SPEC, which has industry standard processor benchmark submissions. There are other benchmarks that people use, and this one has its detractors - but it is considered an industry standard for CPU performance comparisons (and IBM belongs to SPEC). For example, the highest performing single-core Opteron (HP Proliant System) gets a 14.0 in SPECint peak, and 12.7 base. The best single-core Xeon result (a Bull system), is 17.4 peak and 16.8 base. The best dual-core Opteron submission is 14.9 peak and 13.5 base. The best dual-core score for Xeon is 18.1 and 17.5 respectively. SPECint_rate is a throughput benchmark vs. raw integer computing performance (meaning, it is intended for multi-CPU systems). The best Opteron dual-core score is 27.0 peak, 24.3 base. The best Xeon score is 30.8 and 29.3 respectively. With 4 cores (two dual-cores, since AMD doesn't have a native quad-core result), Opteron (in a Sun system) gets 60.4 peak and 51.6 base, whereas Xeon has a 59.4 peak and 56.7 base. Just for information's sake, the base score is with the benchmark compiled using 'typical' compiler options. The peak score allows any compiler options - and some compilers may be targeted for SPEC benchmarks. Peak scores are usually looked at a little more skeptically because of this. There was an interesting submission by Sun (with their own SPARC processors) a few years back with SPEC CPU FP 2000 (not integer) where they had discovered how to 'break' one of the subtests, giving them a *huge* score in that one (they optimized a particular data structure so it would fit into cache). It was considered a 'cheat' because it was not something that would help in any case except that one. http://www.spec.org/osg/cpu2000/results/res2001q4/cpu2000-2009-01131.html I've not seen any data regarding performance improvements for mainframe processors over that time except for the 4-fold increase in installed MIPS between 2000 and 2007. There is no way to know how that relates to individual processors. If you have any data, that would be nice to hear. http://www-03.ibm.com/servers/eserver/zseries/lspr/ Thank you. I don't think there's anyone who buys a mainframe for its sheer processing power. These servers *are* purchased, in part, for their processing power. In part? I really have no idea what point you are trying to argue - except just to be arguing. I've repeatedly stated that the original question was processor performance, not system throughput, reliability or anything else. It may suit your particular position to try and steer it away from that, but it remains a fact that this is what the original post I responded to was questioning. Now, since I have provided benchmark results showing Intel processors (without SOI) outperform AMD processors (using IBM's SOI), that Intel's process size is comparable, at the least, with IBMs, and that performance improvements for x86 (8-fold) appear to have exeeded mainframe processor performance improvments (2.5x in the same time period), I would like to see support for your assertions about IBM mainframe technology *leading* rather than just keeping pace (if it, indeed, does that). Since you have made the challenge of backing up assertions, I would expect you would follow that yourself. This is not meant to be argumentative, nor to disparage - I am trying to find out if, in fact, you have information that would convincingly rebut the data I've shown and dispel the perception that mainframe processors are slower than those used in other platforms. I only included the Intel vs. AMD data to show that your claim of 'technology leadership' (process size, copper, SOI, whatever) is not an indicator of processor performance. I'm interested in this from what you might call an 'academic' interest - I just find this sort of thing interesting (hence my pursuit of the information). Regards, Dean -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
I replied to this directly to the newsgroup yesterday when email delivery was interrupted. I know some don't read the newsgroup, so I just want to reply again with the pertinent data in case anyone is interested. I hope I am not breaching ettiquette by doing this - but if so, I apologize in advance. - Original Message - From: Edward Jaffe [EMAIL PROTECTED] Newsgroups: bit.listserv.ibm-main To: IBM-MAIN@BAMA.UA.EDU Sent: Wednesday, July 18, 2007 4:31 PM Subject: Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article) In any case, raw MIPS improvements are easily researched. For example, the charts at http://www.tech-news.com/publib/ show the largest zSeries processor in 2000, made available in 1Q00, was a 2064-116, capable of delivering 2694 MIPS; the largest System z processor today, made available in 4Q05, was a 2094-754, capable of delivering 17,802 MIPS. That's a 6.6:1 improvement in less than five years. (Not bad!) That isn't per-CPU MIPS, though (which is what was originally being discussed). IOW, there was a discussion about how PSI claimed 350MIPS for an 8-core Itanium vs. 585 MIPS for a single CPU z9, with the comment that it took 8 cores to get half of what a single z9 could do. In any case, I see that the 2064-1C1 (single processor) achieved about 250 MIPS, while a 2094-701 achieved about 608 (single OS image) MIPS. So in reality, the *processor* improvement was about 2.5x - still OK, but not quite as impressive. Which brings me back to the point about 'keeping pace' with processor improvements of other architectures. x86 had about an 8-fold performance increase during the same period (I'm not talking clock rates). For whatever that is worth. And, I am aware that the system performance is more important than processor peformance for end users- however, I'm just trying to keep the focus on the question of why people think mainframe processors are slow compared to others. Regards, Dean -- Edward E Jaffe Phoenix Software International, Inc 5200 W Century Blvd, Suite 800 Los Angeles, CA 90045 310-338-0400 x318 [EMAIL PROTECTED] http://www.phoenixsoftware.com/ -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
Dean Kent wrote: In any case, raw MIPS improvements are easily researched. For example, the charts at http://www.tech-news.com/publib/ show the largest zSeries processor in 2000, made available in 1Q00, was a 2064-116, capable of delivering 2694 MIPS; the largest System z processor today, made available in 4Q05, was a 2094-754, capable of delivering 17,802 MIPS. That's a 6.6:1 improvement in less than five years. (Not bad!) That isn't per-CPU MIPS, though (which is what was originally being discussed). I really wasn't paying any attention to this thread. But, I happened to read the non sequitur you posted (see below), and thought I should respond. I am comparing the pace of improvement in x86 with the mainframe. The IT Jungle article says that installed mainframe MIPS has increased 4 fold over 7 years. I showed that in the same time period, x86 performance has increased over 8 fold. We still aren't getting an apples to apples comparison because it doesn't tell us what the highest performing mainframe processor was in 2000 vs today. First, you compare installed MIPS with performance. Nothing need be said about that. Non sequitur. Then you asked what the highest performing mainframe was in 2000 vs today. I believe I answered that question correctly. In any case, I see that the 2064-1C1 (single processor) achieved about 250 MIPS, while a 2094-701 achieved about 608 (single OS image) MIPS. So in reality, the *processor* improvement was about 2.5x - still OK, but not quite as impressive. I'm not sure I follow your reasoning. Part of mainframe processing power lies in its ability to do effectively use SMP (up to 54-ways now ... more later). It's part of the equation. Good benchmarks are based on computing throughput ... not individual chip speed. Which brings me back to the point about 'keeping pace' with processor improvements of other architectures. x86 had about an 8-fold performance increase during the same period (I'm not talking clock rates). For whatever that is worth. Read your own statement quoted up above. You originally said x86 had an 8-fold increase over seven years. (Is that chip speed or actual server speed?) Now you're saying it had an 8-fold increase in just 4 1/2 years -- the time frame over which the mainframe had a 6.6:1 processor speed increase (or 2.5:1 individual CP speed increase if that's what you choose to focus on). And, I am aware that the system performance is more important than processor peformance for end users- however, I'm just trying to keep the focus on the question of why people think mainframe processors are slow compared to others. I wasn't aware that people think mainframe processors are slow compared to others. Is this from a survey of some kind? Or perhaps from that IT Jungle article you referenced? Can you post the URL? -- Edward E Jaffe Phoenix Software International, Inc 5200 W Century Blvd, Suite 800 Los Angeles, CA 90045 310-338-0400 x318 [EMAIL PROTECTED] http://www.phoenixsoftware.com/ -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
-Original Message- From: IBM Mainframe Discussion List [mailto:[EMAIL PROTECTED] On Behalf Of Edward Jaffe Sent: Friday, July 20, 2007 1:55 PM To: IBM-MAIN@BAMA.UA.EDU Subject: Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article) SNIP I wasn't aware that people think mainframe processors are slow compared to others. Is this from a survey of some kind? Or perhaps from that IT Jungle article you referenced? Can you post the URL? SNIP Actually, I have had this very argument with some previous clients. But they wanted to get rid of the mainframe and did not want to be confused with the facts. They saw GHz for clock speeds for xx machine and saw that the mainframe(s) they had were in MHz. They weren't interested in architectural differences (buss width vs. buss width), number of simultaneous operations (compute or I/O). They (PHBs) could only focus on processor speeds. And somebody save me from all these people with the title of Architect. Most of them have NO IDEA about architecture. Regards, Steve Thompson -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
On Fri, 20 Jul 2007 11:05:52 -0700, Dean Kent wrote: I don't think there's anyone who buys a mainframe for its sheer processing power. These servers *are* purchased, in part, for their processing power. In part? I really have no idea what point you are trying to argue - except just to be arguing. I've repeatedly stated that the original question was processor performance, not system throughput, reliability or anything else. It may suit your particular position to try and steer it away from that, but it remains a fact that this is what the original post I responded to was questioning. But, On Mon, 16 Jul 2007 06:07:07 -0700, Dean Kent wrote: As I said in my first post on this topic, the comparison that would be valid is system thoughput, not processor speed. That's the third time you've accused me of arguing just to be arguing. I still don't think anyone would buy a System z because of it's sheer processor speed. And I don't mean, In part. -- Tom Marchant -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
- Original Message - From: Thompson, Steve [EMAIL PROTECTED] Newsgroups: bit.listserv.ibm-main To: IBM-MAIN@BAMA.UA.EDU Sent: Friday, July 20, 2007 12:18 PM Subject: Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article) Actually, I have had this very argument with some previous clients. But they wanted to get rid of the mainframe and did not want to be confused with the facts. They saw GHz for clock speeds for xx machine and saw that the mainframe(s) they had were in MHz. Frankly, I think it is a very commonly held belief amongst many who are very knowledgable about processors. In fact, my original comment that they are slower was based largely upon discussions to this effect on various fora where CPU architecture is a common topic - often by those who are involved with chip design. I would be *very* happy to be able to report to these individuals that I have contradictory information... in the form of factual data, rather than just dismissive comments, of course. :-). Regards, Dean -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
John S. Giltner writes: Right now the plan is to ONLY have 96 total cores (some systems will have single core processors and some will have dual core) CPU's and 456 GB of RAM. This is a est. and they beleive that they may need to increase this by as much as 50%. So the head to head comparsion is 7 CPU's and 20GB to 96 CPU's and 456 GB of RAM. Doesn't seem to head to head to me. So let's suppose it's 130 cores, for sake of argument. Just curious, but do you then get to double that to support disaster recovery? (Will the disaster recovery work?) Dean Kent writes: Yes, reliability, fault-tolerance, data integrity, etc. are all factors too - but the mainframe does not have a lock on these features, other platforms do as well, including those based on x86. I'm sorry, Dean, but that statement borders on malpractice. If you focus solely on the chip you're totally missing the point, because business outcome two chips. Why are we still focused on the chip in this discussion? John Giltner makes an excellent point, and SPEC wouldn't give you a clue about his situation. Also, the chips, though still important, are probably the least important architectural component in delivering reliability, fault tolerance, data integrity, and other service qualities. As just one example among many, if there's another way to have an active-active highest availability MQ configuration without using z/OS and shared queues in an IBM coupling facility, I'd be interested to know what it is. To pick another example, there's just nothing else like DB2 data sharing. Even Larry Ellison said so. Are there any other general purpose business servers constitutionally capable of delivering five-9s business service availability, no excuses (i.e. including both planned and unplanned outages)? General purpose here means, in particular, running middleware that's actually market relevant? There are niche systems, perhaps. There are also general purpose servers that don't meet the business service availability levels. Is the combination of general purpose and highest service qualities unique to the IBM mainframe? I think so. Let me try to make it simple again. An IBM mainframe is, quite literally, an entire data center in one box. (The earlier server farm comment is quite correct.) If you can benchmark an entire data center all at once you might be on to something. As it turns out some of the outsourcing companies are relatively good at this. You can also get IBM to do the work independently, dispassionately. One good example: http://www.ibm.com/servers/library/pdf/scorpion.pdf - - - - - Timothy Sipples IBM Consulting Enterprise Software Architect Specializing in Software Architectures Related to System z Based in Tokyo, Serving IBM Japan and IBM Asia-Pacific E-Mail: [EMAIL PROTECTED] -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
- Original Message - From: Tom Marchant [EMAIL PROTECTED] Newsgroups: bit.listserv.ibm-main To: IBM-MAIN@BAMA.UA.EDU Sent: Tuesday, July 17, 2007 10:08 PM Subject: Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article) Process size is a limiting factor for performance. One can find devices manufactured in a process one or two generations back that perform just as well as one manufactured on a current process generation, depending upon the design (Itanium vs. x86, for example). So process size is not an indication of performance, particularly across architectures. How can you say that? Haven't you read the literature? The fact is that both of these do provide performance benefits. Yes, I have read them. My understanding is that it allows for lower leakage and power consumption. This, in turn, *may* allow for higher clocked devices. However, performance at a given clock speed does not increase and there may be other limiting factors in clockspeed.As I said, Intel, for example, does not use SOI and seems to be producing very high clockrate devices without it, which (currently) outperform their rival AMD, who *does* use IBM's SOI process. and has nothing at all to do with feature size. Didn't you just say that process (feature) size is not an indicator of performance? Yes, I mistyped. It was supposed to help with leakage, but Intel seems to be doing quite will without these. As Timothy pointed out, Intel is in fact using copper. Has been for years. Lest you misunderstand me - I am not trying to say that Intel is 'better' than IBM, nor the other way around. That's right, you didn't say better. You said faster. More precisely, you said, The mainframe MPU *is* slower than other processor platforms. Yes, and I recanted if you recall. Now I am looking for any data that would help identify whether the perception is, in fact, true or not. Simply stating that IBM is a technology leader, and pointing to their manufacturing process provides no insight into this - and that is what I was pointing out in my rebuttal to you. ... However, IBM is positioning the mainframe to compete in some of the same markets that x86 competes. What do you mean by that? Are you talking about Linux on z? Or are you talking about the larger servers that are being constructed from 86 processors in the hope of competing with mainframes? I'm talking about positioning the mainframe as a web/database server, and a Linux platform. Using the argument that IBM is a leader in technology, and therefore z9 must be better than x86 is ludicrous, if that was your point. I most certainly didn't say that, and I think you know it. Red herrings are not rational arguments. Then I have no idea what your point was in making the statement about IBM's process. TSMC is a huge foundry, and uses advanced manufacturing process - but this has nothing at all to do with whether they are a 'technology leader' able to design and produce advanced microprocessors that can compete with IBM or Intel. I mentioned that I find it hard to believe that IBM would invest in mainframe performance to the extent that x86 manufacturers would, considering the difference in the competitivness of the markets. IBM invests where the money is. The mainframe business is a profitable one. I am comparing the pace of improvement in x86 with the mainframe. The IT Jungle article says that installed mainframe MIPS has increased 4 fold over 7 years. I showed that in the same time period, x86 performance has increased over 8 fold. We still aren't getting an apples to apples comparison because it doesn't tell us what the highest performing mainframe processor was in 2000 vs today. My point was that in a market where there isn't the intense competition that there is less incentive to pour money into it. Intel and AMD *have* to, lest the other one grab huge market share (as Intel saw AMD do with the Opteron, and now AMD is seeing Intel do with the Core 2 processors). IBM seems to be in an enviable position with the mainframe where there isn't any real direct competition, yet. It was stated that IBM invests $1.2B annually on mainframe RD (hardware, software and services). Intel, on the other hand, spends almost $6B on their semiconductor business alone. It should not be surprising that Intel is also a leader in technology - even if their primary product is the lowly x86 based processors. Yes, Intel is another leader in the semiconductor industry. Not, IMO, in computer architectures, though. The iAPX 432 was a notable exception. Agreed. But my intent was not to compare IBM and Intel, but z9 with other architectures - such as x86, Itanium and POWER. x86 being the most common, and Itanium only because it was used in the PSI systems. As for fault-tolerant systems, Stratus and NEC offer them (and likely others). Tandem was first with real fault tolerance as we know
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
- Original Message - From: John S. Giltner, Jr. [EMAIL PROTECTED] Newsgroups: bit.listserv.ibm-main To: IBM-MAIN@BAMA.UA.EDU Sent: Tuesday, July 17, 2007 7:21 PM Subject: Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article) Can the mainframe z900/z990/z9 compete head to head with Intel? IMHO, yes. Well, I think that depends upon the definition of 'head-to-head'. Head to head. You are talking about the same number of processors and same amount of RAM/Central Storage. Not really. For example, I could compare the slowest Sun SPARC against the fastest Itanium, and while one might call that a 'head-to-head' comparison, it really would be skewed in favor of the Itanium. In addition, if you have two z990s vs. 48 x86 machines, you will necessarily require more memory because you will be running at least 46 more copies of an operating system. So no, I wouldn't call this a head-to-head comparison of processors. Are we? NOPE. In the end to replace 80% of 7 z990 CPU's and 20GB of RAM the Intel side will have Right now the plan is to ONLY have 96 total cores (some systems will have single core processors and some will have dual core) CPU's and 456 GB of RAM. This is a est. and they beleive that they may need to increase this by as much as 50%. So the head to head comparsion is 7 CPU's and 20GB to 96 CPU's and 456 GB of RAM. Doesn't seem to head to head to me. It isn't, but it is really cool information. What it seems to indicate to me is not that the z9 CPU is faster (which is may or may not be), but that the z990 system design has much better throughput. Of course, without any details on the Intel based system, it is hard to determine what the real limiting factor is. For example, whoever made the decision might have gone with the cheapest solution (hardwarewise), rather than a more robust and more expensive solution. Since the application may be very I/O bound (typical business application), the CPU speed may not even be the bottleneck (and probably isn't in this case). This would mean you need more systems to handle the same transaction load, which means more CPUs just because you need at least one for each system, and of course, more memory. I mentioned that Stratus makes very robust, fault-tolerant x86 based systems but they are really expensive. Sun makes systems that are able to handle more users/transactions as well, but still aren't as cheap as commodity server systems. A true head to head comparison would have to be done holding as many factors constant as possible. If one were to determine how many users a single, well-designed x86 box could handle, and measure performance for both it and a z990 with the same number of users/transactions, that would be a better indicator of processor speed... but that is also presuming they are running the same application. For example, if you are moving from z/OS to Windows, you are likely not running the same application code and probably not even from the same vendor. So the OS and application differences can account for quite a bit of performance difference. However, if both are running Linux and the same application - then you have a better head-to-head test. So, I suggest that a real head-to-head CPU performance comparison would be to put the exact same load on a single processor system (for example) and measure the response time, or how fast a problem is solved, etc. That is what CPUs do (calculations). However, where the mainframe shines is its ability to support many users and perform massive amount of I/O without bogging down the system. This is not a processor issue, but a system design issue, as has been mentioned before. That is a about a 13:1 ratio on CPU's and 23:1 on RAM. If Intel was faster then we should be able to do more work on less processors. No. As explained above. It would be true if all other factors were the same, but they are not. Please show me a site that has migrated off a modern day mainframe to Intel using the same number of CPU's and same amount of Central Storage/RAM as they had on the mainframe. You know head to head. I would like to see just that kind of comparison myself, but we probably won't see it. But, as I said, the information you posted is really interesting. If you can provide more details about the Intel systems, that would be useful (yes, I am curious. No, I am not working for a hardware vendor of any kind). My conclusion regarding this situation is that it seems obvious that the processor speed isn't really the issue. And in that case, whether the z9 is slower than an x86 (or RISC) processor is not even important. This is where appropriate benchmarks are useful (such as the SPECjbb or TPC benchmarks I've mentioned, or any other that can be identified) - to show where the real strengths of a system are. Regards, Dean -- For IBM-MAIN subscribe /
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
- Original Message - From: Timothy Sipples [EMAIL PROTECTED] Newsgroups: bit.listserv.ibm-main To: IBM-MAIN@BAMA.UA.EDU Sent: Tuesday, July 17, 2007 11:12 PM Subject: Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article) I'm sorry, Dean, but that statement borders on malpractice. If you focus solely on the chip you're totally missing the point, because business outcome two chips. Why are we still focused on the chip in this discussion? John Giltner makes an excellent point, and SPEC wouldn't give you a clue about his situation. Also, the chips, though still important, are probably the least important architectural component in delivering reliability, fault tolerance, data integrity, and other service qualities. Because the original question was why there is a perception that mainframe processors are slow, relatively speaking. All of these other factors are very well known, and discussing them here is preaching to the choir. This is a mainframe list, after all, so I expect people to know the attributes of mainframes, but not necessarily those of other platforms (particularly when I see the comments about PCs being just for word processing). I sense that the defensiveness about this subject is a reaction to the perception here that the mainframe market is an embattled one, with mounting losses and so it has to be vigorously defended against all heretics. The discussion was about processor performance, and my comments have been intended to address that question, and nothing else. When all of the other attributes of the mainframe have been brought up, I've tried to just point out that other platforms have, or are acquiring, those features as well and get back to the performance issue. This seems to be considered a put down or a threat. The closest thing I've seen to useful information related to the original question was the data on how many instructions were required to emulate the instruction set, and the MIPS for the PSI systems vs. z9. That was interesting, and helps to answer the question about peformance, which I personally find interesting. Regards, Dean -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
On Tue, 17 Jul 2007 23:51:48 -0700, Dean Kent wrote: ...As I said, Intel, for example, does not use SOI and seems to be producing very high clockrate devices without it, which (currently) outperform their rival AMD, Another assertion without data to back it up. I hope you are not talking about clock rates. I've not seen any data regarding performance improvements for mainframe processors over that time except for the 4-fold increase in installed MIPS between 2000 and 2007. There is no way to know how that relates to individual processors. If you have any data, that would be nice to hear. http://www-03.ibm.com/servers/eserver/zseries/lspr/ I don't think there's anyone who buys a mainframe for its sheer processing power. These servers *are* purchased, in part, for their processing power. In part? -- Tom Marchant -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
-Original Message- From: IBM Mainframe Discussion List [mailto:[EMAIL PROTECTED] On Behalf Of Timothy Sipples Sent: Wednesday, July 18, 2007 1:13 AM To: IBM-MAIN@BAMA.UA.EDU Subject: Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article) SNIP Let me try to make it simple again. An IBM mainframe is, quite literal= ly, an entire data center in one box. SNIP Not any more (remember IBM v PSI). As of this past fall, when the contracts with FLEX were not renewed, it now takes a z/box, a raid unit, probably a tape unit or three, etc. While IBM was shipping the MP200, MP3000, R/390, and/or P/390 this was true. All the DASD in the box, with ethernet, etc. Same held true with the FLEX boxes and the z/Laptop systems. But now IBM has killed off those drop and play systems, and so this is no longer true. HOWEVER, the TCO and TCO (ownership, operation) IS known (regardless of physical config). The ROI can be plotted. There is very little HIDDEN costs (HCO) in a mature system such as the z/Architecture. Regards, Steve Thompson -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
HOWEVER, the TCO and TCO (ownership, operation) IS known (regardless of physical config). The ROI can be plotted. There is very little HIDDEN costs (HCO) in a mature system such as the z/Architecture. Yes, but: 1. TCA (acquisition) is what scares management. 2. People costs are often a political, rather than a technical argument. - Too busy driving to stop for gas! -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
On 17 Jul 2007 12:09:37 -0700, [EMAIL PROTECTED] (Dean Kent) wrote: I agree with your first point, but not your second. There *is* a reason that SPEC (and other benchmarking organizations) exist. These customers want a common performance metric to identify the value they are getting for the money they spend. Yes, reliability, fault-tolerance, data integrity, etc. are all factors too - but the mainframe does not have a lock on these features, other platforms do as well, including those based on x86. If I was making a decision about my shop's hardware, a Standard benchmarking test is a start, as long as they are benchmarking real data processing - but I'm really interested in benchmarking my particular needs. The System is what determines my throughput, response time, and capacity. If a PC has a faster chip than a mainframe, I cannot assume that it will give me better response time with a thousand concurrent users. -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
-Original Message- From: IBM Mainframe Discussion List [mailto:[EMAIL PROTECTED] On Behalf Of Ted MacNEIL Sent: Wednesday, July 18, 2007 9:01 AM To: IBM-MAIN@BAMA.UA.EDU Subject: Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article) HOWEVER, the TCO and TCO (ownership, operation) IS known (regardless of physical config). The ROI can be plotted. There is very little HIDDEN costs (HCO) in a mature system such as the z/Architecture. Yes, but: 1. TCA (acquisition) is what scares management. 2. People costs are often a political, rather than a technical argument. SNIP If your point 1 were accurate, management would dump those servers. However, because they are pleasing to the eye and would make one wise... Later, Steve Thompson -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
If your point 1 were accurate, management would dump those servers. Two points that substantiate the TCA argument (or, to be more presice ICA (initial cost of acquisition)): 1. Mainframe DASD costs more than midrange disk. (Even IBM's DS series). 2. IFL's cost arounnd 100K (US). (I can buy a lot of servers for that price -- a manager I know). - Too busy driving to stop for gas! -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
Dean Kent wrote: After all, IBM does use these benchmarks for POWER and x86 based systems that they sell into those markets. Regards, Dean -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html A different benchmark is described here: http://www.ibm.com/press/us/en/pressrelease/21044.wss Would be interesting to see such an application with this number of transactions on other platforms as well. Birger Heede IBM Denmark -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
Two points that substantiate the TCA argument (or, to be more presice ICA (initial cost of acquisition)): 1. Mainframe DASD costs more than midrange disk. (Even IBM's DS series). I ascribe this to you get what you pay for, to a large degree. 2. IFL's cost arounnd 100K (US). (I can buy a lot of servers for that price -- a manager I know). FWIW, I was quoted $41k earlier this month, at WAVV it was mentioned they could be had in the 30s (although I haven't found one that cheap). Neither of which alters your point appreciably. -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
-Original Message- From: IBM Mainframe Discussion List [mailto:[EMAIL PROTECTED] On Behalf Of Ted MacNEIL Sent: Wednesday, July 18, 2007 10:43 AM To: IBM-MAIN@BAMA.UA.EDU Subject: Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article) If your point 1 were accurate, management would dump those servers. Two points that substantiate the TCA argument (or, to be more presice ICA (initial cost of acquisition)): 1. Mainframe DASD costs more than midrange disk. (Even IBM's DS series). 2. IFL's cost arounnd 100K (US). (I can buy a lot of servers for that price -- a manager I know). What kind and size of Server? How much business processing can it do by itself? I know that the Enterprise class Intel/AMD servers here are very expensive, compared to desktop class. Unfortunately, I don't know even the approximate cost. I have have been told that the combined hardware+software cost of the Open Systems servers (mainly Windows) costs more than the hardware+software cost of the z9BC we have. I don't know the personnel costs. Oh, and the z9BC is still doing at least 80% of the core business. And people are constantly complaining about server response time. CICS response time stays sub-second. Batch, on the other hand, sometimes gets bogged down. -- John McKown Senior Systems Programmer HealthMarkets Keeping the Promise of Affordable Coverage Administrative Services Group Information Technology The information contained in this e-mail message may be privileged and/or confidential. It is for intended addressee(s) only. If you are not the intended recipient, you are hereby notified that any disclosure, reproduction, distribution or other use of this communication is strictly prohibited and could, in certain circumstances, be a criminal offense. If you have received this e-mail in error, please notify the sender by reply and delete this message without copying or disclosing it. -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
It was my experience that the main cost difference is not the hardware, it's software. Open system software was lisenced by the processor. So doing a little math. 1 IFL = 1 processor = $100,000 AP Websphere = $10,000 (very old price) Total = $110,000 30 servers ( 4 way ) @ $3000 each = $90,000 AP WS = $10,000 30 servers * 4 processor each * 10,000 = $1,200,000 Total = $1,290,000 So where is the cost savings from using the server farm? We won't even get into personel costs or environmentals. Jimmy -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
Depending on the deal, you can get an IFL in the low $30's. I agree too, the RS6000's that we are installing have LPAR's and cost as much as the z9 in some cases. If you load up a four or eight way Xeon server box, you are looking at $30-50K in many instances. I think that to some degree, many new managers equate mainframe and expensive, and servers and inexpensive, when it is simply not true in many cases. Let us not begin to talk about security and viruses either.. Doug snip What kind and size of Server? How much business processing can it do by itself? I know that the Enterprise class Intel/AMD servers here are very expensive, compared to desktop class. Unfortunately, I don't know even the approximate cost. I have have been told that the combined hardware+software cost of the Open Systems servers (mainly Windows) costs more than the hardware+software cost of the z9BC we have. I don't know the personnel costs. Oh, and the z9BC is still doing at least 80% of the core business. And people are constantly complaining about server response time. CICS response time stays sub-second. Batch, on the other hand, sometimes gets bogged down. -- John McKown Senior Systems Programmer HealthMarkets Keeping the Promise of Affordable Coverage Administrative Services Group Information Technology snip Doug Fuerst Consultant BK Associates Brooklyn, NY (718) 921-2620 (Office) (718) 921-0952 (Fax) (917) 572-7364 (Cell) [EMAIL PROTECTED] -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
- Original Message - From: Birger Heede [EMAIL PROTECTED] Newsgroups: bit.listserv.ibm-main To: IBM-MAIN@BAMA.UA.EDU Sent: Wednesday, July 18, 2007 8:55 AM Subject: Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article) A different benchmark is described here: http://www.ibm.com/press/us/en/pressrelease/21044.wss Would be interesting to see such an application with this number of transactions on other platforms as well. One that I have wondered about is SAP, as it runs on mainframes, minis and micros (to use the legacy terms), and is a fairly widely used application. Might be license issues, of course... Regards, Dean Birger Heede IBM Denmark -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
-Original Message- From: IBM Mainframe Discussion List On Behalf Of Ted MacNEIL If your point 1 were accurate, management would dump those servers. Two points that substantiate the TCA argument (or, to be more presice ICA (initial cost of acquisition)): 1. Mainframe DASD costs more than midrange disk. (Even IBM's DS series). 2. IFL's cost arounnd 100K (US). (I can buy a lot of servers for that price -- a manager I know). Yep And each one of those servers requires a physical connection to electrical power; each requires an operating system license; each one that runs a database requires a DBMS license, etc., etc. Oh, did you want Test/Development and QA copies too? More $. It is also true, assuming you already have a z/Box, that you can run a lot of servers on one IFL, with ONE operating system license, ONE DBMS license, ZERO physical connections to electrical power, etc., etc. Depending on the application set, the break-even crossover point could occur at as low as 5 servers (or server images). And for T/D and QA copies, just copy a few files, share a few others, create user IDs and profiles, and XAUTOLOG them on. No extra $. -jc- -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
Yep And each one of those servers requires a physical connection to electrical power; each requires an operating system license; each one that runs a database requires a DBMS license, etc., etc. Oh, did you want Test/Development and QA copies too? More $. I did state that management had the issues, not me. I pointed out their concerns and saw the arguments re-iterated for the n-th time. We are not the problem. IBM is pointing this all out to the techies! Nobody is properly presenting this to management. All we are doing is b*tching about the fact that they are not listening. - Too busy driving to stop for gas! -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
Well of course, Dean, Timothy makes perfect sense. You should believe everything your vendors say, without any verification, right? ;-) At the last shop I worked at (a government shop), they had a study done to evaluate what platform they should be running on, and the study proposed Oracle running on Sun servers. I later found out (to my utter lack of surprise), that the study had been done free of charge. I will give you exactly one guess who it was that did the study gratis. Of course, the conclusion was the one that was desired by those that asked for the study (upper management), who are convinced that the mainframe is obsolete and expensive. The study is being used as a justification to spend 10's of millions of dollars to replace the mainframe. Of course, the complete lack of public benchmarks for the mainframe make it impossible to refute the performance conclusions of the so called study, so I think it actually hurts IBM, even if the benchmarks would not be in favor of the mainframe. IBM's refusal to submit (or allow others to do so) mainframe benchmark scores allows others to make wild claims about how slow mainframes are, without any way to refute such claims, leaving us (the mainframe proponents) completely unarmed. Dean Kent 7/17/2007 3:03:45 PM Instead, Timothy Sipples suggests (and I paraphrase from his reply to me) if you don't know, talk to your IBM rep - he'll tell you what you need. Sure, he'll tell me I need a Sun system instead of an IBM system - right? Or perhaps I should go talk to Sun or HP or Dell to find out what best suits my needs. If you care about the platform, you should care about the problem... or so it seems to me. Regards, Dean
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
I was not going to comment on this as I am finding it quite enjoyable and I am learning some things. I would like to interject a performance comparison between an IBM 3081 and a Sun System and a PC. I don't know if this is a good one or not, I will leave it up to the reader. Approximately 15 years ago we had an application that during its processing needed to create a random number for every options trade. This was written in COBOL (the random number was a psuedo random number) A performance person was asked to look into it and figured out that SAS had a better random number generator (the COBOL was OK for the designer ie it didn't have to be truly random). At the time there were the usual wars between the pc and the MF as to which was better. SAS beat COBOL hands down. The performance person was able to duplicate the test on a PC and a SUN system. Funny thing in just raw cpu time the 3081 was 1/3 less than the Sun system and 50 percent less than the PC. The only I/O involved was to read a file that had the options trade in it, no output was written. Again I was not the person involved in the test but he was partial to the Sun system going into the test. I will take his word that all other things were equal. He was getting his timing from the standard ways. I looked at the source briefly and it seemed on the up and up, I did not have any reason to doubt him. Ed -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
Dean Kent writes: [Re: SOI technology.] It has not been shown that either of these provides any benefit in performance, and has nothing at all to do with feature size. It was supposed to help with leakage, but Intel seems to be doing quite will without these. We're getting way, way far afield here, but there are a few points to tidy up. Indeed Intel has not widely deployed SOI, though their primary objection is cost and not value. (Patent royalties?) And Intel does concede they see value in an SOI variant called FD-SOI. Also, Wikipedia happens to disagree with you re: the usefulness of SOI. I realize Intel has its point of view, but Intel the semiconductor market. Re: Copper, IBM invented it and now virtually all semiconductor companies, including Intel, use it. However, IBM is positioning the mainframe to compete in some of the same markets that x86 competes. Some, yes. IBM also sells lots of X86 servers (and software and services): http://www.ibm.com/systems/x An IBM rep would be happy to work with you to understand which server(s) is(are) appropriate for your particular business. And it'll have extremely little to do with SPECint. :-) As far as the largest semiconductor manufacturers in the world, Intel is #1, with Samsung, TI, Toshiba and STmicro rounding out the top 5. IBM isn't listed in the top 10 list. AMD was listed as #7 as of Dec 2006, so the two leading x86 manufacturers are in the top 10. This, of course, has absolutely nothing to do with whether any given product is a better performer than any other. Big problems with that ranking you've got, though. For one thing it excludes foundries -- you know, the places where chips are actually made. :-) IBM is a substantial foundry for other companies on the list. Also, that list excludes royalties and RD services, areas where IBM happens to make lots of money -- and which are the most relevant for purposes of this discussion thread. It was stated that IBM invests $1.2B annually on mainframe RD (hardware, software and services). Intel, on the other hand, spends almost $6B on their semiconductor business alone. Please do note that that advertised figure is *direct* investment. If IBM invents SOI or, more recently, airgap nano-assembly, that RD money is not counted in the z bucket even though it has a huge beneficial impact. And, no, Intel does not spend nearly $6B on its semiconductor business alone. That figure is the company's *total* corporate-wide RD budget -- and Intel expects that figure will dip to roughly $5.3B in 2007, by the way. Maybe the X86 market isn't so competitive this year. :-) I'm sure it's mostly semiconductors, since that's Intel's core business, but some portion is not (e.g. compilers). It also includes such things as wireless chip, graphics chip, and flash memory RD, areas largely irrelevant to business server development. RD related to notebook computers, PDAs, mobile phones, and other small devices has only partial relevance to business servers, and mobile computing is really Intel's focus at the moment. Now, that's still a lot of RD money, but for comparison IBM's annual corporate RD budget exceeds $6B. Re: Fault tolerance, software plays no role? Of course it does. Could you lash together two X86 or Itanium chips and have them execute in parallel? Yes, you could. But the software above it is the huge problem. The IBM mainframe design is unique in pushing RAS down deeper into the hardware and in having the most mature, popular, business-oriented operating system and middleware products riding shotgun. Granted, I worry about software a lot, but it really is extremely important. The goal is to achieve a business result (continuous business service, including planned events), remember. I notice nobody ever mentioned multi-level cache memory structures in massively SMP systems. You know, something IBM mainframes have. It's a very difficult piece of engineering. It also happens to be the most useful system attribute for consolidation and virtualization. And that's a huge problem with the benchmarks you've listed. SPECjbb is fine if you're running a single Java 3-tier application. What if you're running 50 of them, in 3 different programming languages, with 100 batch jobs? With varying service classes? How do you benchmark that, which is exactly the real world use of these special systems -- data center in a box? Tough problem. Never mind computing resources for DR, training, education, testing, development, etc. By the way, the reason that total MIPS shipment figures are relevant, at least to some extent, is that mainframe MIPS are actually *used* and, as close as anything in the world of computing, that translates into increased business activity with the systems. Nobody with a budget ever buys extra MIPS just because: there's no purpose and no advantage. In contrast, the vast bulk of X86 MHz/GHz shipments (99+% I'd guess) simply wait faster, because the vast majority of
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
On 16 Jul 2007 19:27:37 -0700, [EMAIL PROTECTED] (Dean Kent) wrote: Lest you misunderstand me - I am not trying to say that Intel is 'better' than IBM, nor the other way around. I am not trying to say that x86 processors are 'better' than z9. Each has its strengths, and weaknesses. There is no 'one processor to rule them all'. However, IBM is positioning the mainframe to compete in some of the same markets that x86 competes. This means that customers will expect a direct comparison on performance - and rightly so. IBM is not positioning the mainframe to compete with a computer chip. Apples and bird seed. Customers of computer chips are computer manufacturers. IBM is positioning the mainframe to compete with server farms. That is something very different. And the x86 speed of a PC is not what customers care about when looking at alternatives for these business needs. -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
In a message dated 7/17/2007 9:29:45 A.M. Central Daylight Time, [EMAIL PROTECTED] writes: IBM is positioning the mainframe to compete with server farms. That is something very different. And the x86 speed of a PC is not what customers care about when looking at alternatives for these business needs. _http://www.embedded-computing.com/news/db/?6516_ (http://www.embedded-computing.com/news/db/?6516) ** Get a sneak peek of the all-new AOL at http://discover.aol.com/memed/aolcom30tour -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
- Original Message - From: Howard Brazee [EMAIL PROTECTED] Newsgroups: bit.listserv.ibm-main To: IBM-MAIN@BAMA.UA.EDU Sent: Tuesday, July 17, 2007 7:29 AM Subject: Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article) IBM is not positioning the mainframe to compete with a computer chip. Apples and bird seed. Customers of computer chips are computer manufacturers. IBM is positioning the mainframe to compete with server farms. That is something very different. And the x86 speed of a PC is not what customers care about when looking at alternatives for these business needs. I agree with your first point, but not your second. There *is* a reason that SPEC (and other benchmarking organizations) exist. These customers want a common performance metric to identify the value they are getting for the money they spend. Yes, reliability, fault-tolerance, data integrity, etc. are all factors too - but the mainframe does not have a lock on these features, other platforms do as well, including those based on x86. Maybe I can point out the dilemma better this way: People here have asked how managers can justify migrating their mission-critical applications off the mainframe and onto a 'PC'. At the same time, these people will say that there is no common metric to compare the various platforms, that they are just different. So, a manager who must make a business/financial decision is given no tools with which to make that decision - so is it any wonder that those decisions seem, well, random? One would think that *if* the mainframe can compete head-to-head with these other server systems that those in the business would *want* a common metric. They would *strive* to identify something that managers could use to make better decisions. This should, you would think, include the vendor who would benefit most by such information. Continually claiming that there is, and cannot be any comparison seems counter productive. You are *asking* these managers to go with the latest fad because they have nothing else to use as a guideline. If it can't compete, then perhaps it makes sense to claim that no such metric can be identified. As long as the workloads are completely different, then it makes sense. They they overlap, however, you are asking for people to flip a coin to choose unless you give them another tool to use. As for the car analogy in performance, I would suggest the following: I can look at horsepower, top speed, acceleration, luggage capacity, towing capacity, gas mileage and various other factors that are available for *all* vehicles. This allows me to make an intelligent, informed decision about which particular vehicle is best for my needs, whether it be a sportscar, a family vehicle, a farm vehicle or a large commercial vehicle. Instead, Timothy Sipples suggests (and I paraphrase from his reply to me) if you don't know, talk to your IBM rep - he'll tell you what you need. Sure, he'll tell me I need a Sun system instead of an IBM system - right? Or perhaps I should go talk to Sun or HP or Dell to find out what best suits my needs. If you care about the platform, you should care about the problem... or so it seems to me. Regards, Dean -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
-Original Message- From: IBM Mainframe Discussion List [mailto:[EMAIL PROTECTED] On Behalf Of Dean Kent Sent: Tuesday, July 17, 2007 2:04 PM To: IBM-MAIN@BAMA.UA.EDU Subject: Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article) SNIP Instead, Timothy Sipples suggests (and I paraphrase from his reply to me) if you don't know, talk to your IBM rep - he'll tell you what you need. Sure, he'll tell me I need a Sun system instead of an IBM system - right? Or perhaps I should go talk to Sun or HP or Dell to find out what best suits my needs. If you care about the platform, you should care about the problem... or so it seems to me. Regards, Dean SNIP Or, like a former client of mine, they [IBM z/Series sales reps] will sell them on Siebel, Regatta, etc. to the tune of a few US$, because the Mainframe can't do GUI (I kid you not, not all IBM sales people have been to the Mainframe Top Gun classes, and they revenue on their specialty, not mainframes). Caveat Emptor. Regards, Steve Thompson -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
-Original Message- From: IBM Mainframe Discussion List [mailto:[EMAIL PROTECTED] On Behalf Of Thompson, Steve Sent: Tuesday, July 17, 2007 2:26 PM To: IBM-MAIN@BAMA.UA.EDU Subject: Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article) SNIP Or, like a former client of mine, they [IBM z/Series sales reps] will sell them on Siebel, Regatta, etc. to the tune of a few US$, because the Mainframe can't do GUI (I kid you not, not all IBM sales people have been to the Mainframe Top Gun classes, and they revenue on their specialty, not mainframes). SNIP Correction: [IBM z/Series sales reps] was supposed to be [IBM non z/Series sales reps]. Sometimes spelling checkers just don't do what you need... Regards, Steve Thompson -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
-Original Message- From: IBM Mainframe Discussion List [mailto:[EMAIL PROTECTED] On Behalf Of Thompson, Steve Sent: Tuesday, July 17, 2007 2:26 PM To: IBM-MAIN@BAMA.UA.EDU Subject: Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article) -Original Message- From: IBM Mainframe Discussion List [mailto:[EMAIL PROTECTED] On Behalf Of Dean Kent Sent: Tuesday, July 17, 2007 2:04 PM To: IBM-MAIN@BAMA.UA.EDU Subject: Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article) SNIP Instead, Timothy Sipples suggests (and I paraphrase from his reply to me) if you don't know, talk to your IBM rep - he'll tell you what you need. Sure, he'll tell me I need a Sun system instead of an IBM system - right? Or perhaps I should go talk to Sun or HP or Dell to find out what best suits my needs. If you care about the platform, you should care about the problem... or so it seems to me. Regards, Dean SNIP Or, like a former client of mine, they [IBM z/Series sales reps] will sell them on Siebel, Regatta, etc. to the tune of a few US$, because the Mainframe can't do GUI (I kid you not, not all IBM sales people have been to the Mainframe Top Gun classes, and they revenue on their specialty, not mainframes). Caveat Emptor. Regards, Steve Thompson Gee, the mainframe can't do GUI. I guess that I'd better get rid of this Java application which is GUI based and runs on the z/OS system. Granted the terminal is a X server on my desktop. But what does that matter? The 3270 green screens did a lot of their processing off mainframe as well. And if I needed it, I could get a very cheap system which can run Linux and act as an X server. In fact, I know of a box called the Koolu which is only USD 199.00 for the box with Ubuntu Linux (256 Mib RAM, VGA up to 1920x440, 4 USB ports, 16 bit audio). Just add a keyboard, mouse, and monitor. That's even cheaper than the old style 3270 green screen. -- John McKown Senior Systems Programmer HealthMarkets Keeping the Promise of Affordable Coverage Administrative Services Group Information Technology The information contained in this e-mail message may be privileged and/or confidential. It is for intended addressee(s) only. If you are not the intended recipient, you are hereby notified that any disclosure, reproduction, distribution or other use of this communication is strictly prohibited and could, in certain circumstances, be a criminal offense. If you have received this e-mail in error, please notify the sender by reply and delete this message without copying or disclosing it. -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
Dean Kent wrote: - Original Message - From: Howard Brazee [EMAIL PROTECTED] Newsgroups: bit.listserv.ibm-main To: IBM-MAIN@BAMA.UA.EDU Sent: Tuesday, July 17, 2007 7:29 AM Subject: Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article) IBM is not positioning the mainframe to compete with a computer chip. Apples and bird seed. Customers of computer chips are computer manufacturers. IBM is positioning the mainframe to compete with server farms. That is something very different. And the x86 speed of a PC is not what customers care about when looking at alternatives for these business needs. I agree with your first point, but not your second. There *is* a reason that SPEC (and other benchmarking organizations) exist. These customers want a common performance metric to identify the value they are getting for the money they spend. Yes, reliability, fault-tolerance, data integrity, etc. are all factors too - but the mainframe does not have a lock on these features, other platforms do as well, including those based on x86. Maybe I can point out the dilemma better this way: People here have asked how managers can justify migrating their mission-critical applications off the mainframe and onto a 'PC'. At the same time, these people will say that there is no common metric to compare the various platforms, that they are just different. So, a manager who must make a business/financial decision is given no tools with which to make that decision - so is it any wonder that those decisions seem, well, random? One would think that *if* the mainframe can compete head-to-head with these other server systems that those in the business would *want* a common metric. They would *strive* to identify something that managers could use to make better decisions. This should, you would think, include the vendor who would benefit most by such information. Continually claiming that there is, and cannot be any comparison seems counter productive. You are *asking* these managers to go with the latest fad because they have nothing else to use as a guideline. If it can't compete, then perhaps it makes sense to claim that no such metric can be identified. As long as the workloads are completely different, then it makes sense. They they overlap, however, you are asking for people to flip a coin to choose unless you give them another tool to use. As for the car analogy in performance, I would suggest the following: I can look at horsepower, top speed, acceleration, luggage capacity, towing capacity, gas mileage and various other factors that are available for *all* vehicles. This allows me to make an intelligent, informed decision about which particular vehicle is best for my needs, whether it be a sportscar, a family vehicle, a farm vehicle or a large commercial vehicle. Instead, Timothy Sipples suggests (and I paraphrase from his reply to me) if you don't know, talk to your IBM rep - he'll tell you what you need. Sure, he'll tell me I need a Sun system instead of an IBM system - right? Or perhaps I should go talk to Sun or HP or Dell to find out what best suits my needs. If you care about the platform, you should care about the problem... or so it seems to me. Regards, Dean Can the mainframe z900/z990/z9 compete head to head with Intel? IMHO, yes. Head to head. You are talking about the same number of processors and same amount of RAM/Central Storage. We have two z990, a 304 and 303. A total of 7 (seven) CPU's and 20GB of total central stoage for the z/OS images. Now in our enviroment we do NOT have a test mainframe, we do not have test LPAR's. All production/test/development/QA/user accecptance testing are done with the same LPARs. There is a system programmer sand box on the 303. We are planning to migrate 80% of our workload off the mainframe on to Intel. If the Intel processors were really faster/better than the z990 CPU, then we sould be to get a single Intel box with 6 CPUs to replace our two mainframes. Right? Remember this is head to head. Are we? NOPE. In the end to replace 80% of 7 z990 CPU's and 20GB of RAM the Intel side will have Right now the plan is to ONLY have 96 total cores (some systems will have single core processors and some will have dual core) CPU's and 456 GB of RAM. This is a est. and they beleive that they may need to increase this by as much as 50%. So the head to head comparsion is 7 CPU's and 20GB to 96 CPU's and 456 GB of RAM. Doesn't seem to head to head to me. That is a about a 13:1 ratio on CPU's and 23:1 on RAM. If Intel was faster then we should be able to do more work on less processors. Please show me a site that has migrated off a modern day mainframe to Intel using the same number of CPU's and same amount of Central Storage/RAM as they had on the mainframe. You know head to head. -- For IBM-MAIN subscribe /
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
On Mon, 16 Jul 2007 19:21:28 -0700, Dean Kent wrote: Tom Marchant [EMAIL PROTECTED] wrote in message news:[EMAIL PROTECTED] On Mon, 16 Jul 2007 13:53:39 -0700, Dean Kent wrote: Since Intel is already on 45nm process, I don't think you can call 90nm 'leading in technology'. Already? when will they begin shipments? They say 2H2007. The z9 has been shipping since September, 2005. As I said - you seem to be arguing just to argue. Intel began shipping 90nm products in Feb. 2004, but that really isn't the point. Process size is not an indicator of performance, or feature set. Personal attacks are unwarranted. Process size is a limiting factor for performance. I guess you don't think much of SOI or copper either. It has not been shown that either of these provides any benefit in performance, How can you say that? Haven't you read the literature? The fact is that both of these do provide performance benefits. and has nothing at all to do with feature size. Didn't you just say that process (feature) size is not an indicator of performance? It was supposed to help with leakage, but Intel seems to be doing quite will without these. As Timothy pointed out, Intel is in fact using copper. Has been for years. Lest you misunderstand me - I am not trying to say that Intel is 'better' than IBM, nor the other way around. That's right, you didn't say better. You said faster. More precisely, you said, The mainframe MPU *is* slower than other processor platforms. ... However, IBM is positioning the mainframe to compete in some of the same markets that x86 competes. What do you mean by that? Are you talking about Linux on z? Or are you talking about the larger servers that are being constructed from 86 processors in the hope of competing with mainframes? Using the argument that IBM is a leader in technology, and therefore z9 must be better than x86 is ludicrous, if that was your point. I most certainly didn't say that, and I think you know it. Red herrings are not rational arguments. I mentioned that I find it hard to believe that IBM would invest in mainframe performance to the extent that x86 manufacturers would, considering the difference in the competitivness of the markets. IBM invests where the money is. The mainframe business is a profitable one. It was stated that IBM invests $1.2B annually on mainframe RD (hardware, software and services). Intel, on the other hand, spends almost $6B on their semiconductor business alone. It should not be surprising that Intel is also a leader in technology - even if their primary product is the lowly x86 based processors. Yes, Intel is another leader in the semiconductor industry. Not, IMO, in computer architectures, though. The iAPX 432 was a notable exception. As for fault-tolerant systems, Stratus and NEC offer them (and likely others). Tandem was first with real fault tolerance as we know it today. The I/O performance that was once the realm of the mainframe is now available for other platforms as well. There's another assertion. Have you any data to back it up? What kind of I/O bandwidth can these systems handle? A z9 has from 16 to 64 STI busses, each capable of transferring 2.7 Gbytes/second. So, while the mainframe still enjoys a relatively comfortable niche, I don't think mainframers should be too smug about it. x86 processors are not just good for word processing, despite some comments to that effect. Making snide, derogatory remarks about x86 or other platforms is just as foolish as PC people making derogatory comments about the mainframe. It would be nice if people would post information that would further the dialog rather than simply to defend a position. Please do, if you have any data. So far you have provided precious little. ... I've been working with mainframes since 1976, and x86 based systems since about 1992. I'm certainly not a hardware engineer, and am no authority on all of the ins and outs of the various strengths and weaknesses of each platform. I don't know why you're telling us that. As it happens I have you beat. I started programming on System/360 in 1970. In 1986 I was writing x86 assembler code. Starting in 1976 I was designing and building hardware for a 6502 based system, as well as writing 6502 assembler code. I've spent a lot of time evaluating processor architectures. Today, I work full time developing z/Architecture system code. What I do find a little tiresome are the assertions and derogations about various platforms based upon 'common wisdom' rather than verifiable information. So do I. I'm still waiting for you to back up your assertions. And, by the way, the SPECjbb that you keep mentioning is purely a CPU benchmark. It is designed to run without performing any I/O. I don't think there's anyone who buys a mainframe for its sheer processing power. -- Tom Marchant
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
On Tue, 17 Jul 2007 22:21:23 -0400, John S. Giltner, Jr. [EMAIL PROTECTED] wrote: Can the mainframe z900/z990/z9 compete head to head with Intel? IMHO, yes. Head to head. You are talking about the same number of processors and same amount of RAM/Central Storage. I don't understand why the comparison is always technological. Why not compare what the costs to acquire, provision, operate and maintain the system over, say, 4 years? This is the TCO stuff, including staff, software license costs, electricity, floor space, etc. Head to head where it counts: IT spending. Do people *really* care that much about MHz or GB? I figured folks were primarily interested in affordability. Can you get the function you need, with the qualities of service you require, at a price you can afford? Can your infrastructure absorb/accomodate growth? If all you want is a web server, don't buy a mainframe. If all you want is 5 webservers, don't buy a mainframe. If, on the other hand, you want 400 webservers, buy a mainframe. If all you want is 50 servers, buy a small mainframe. Alan Altmark z/VM Development IBM -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
Interesting, Them guys gave a presentation over here some time ago. The numbers they quoted for a HP Superdome was 80 MIPS per engine, with up to 64 engines. Are they now stating that using Itaniums can generate 350 MIPS per engine, that means that upgrading from the original HP Superdome, to an Itanium based system, is a 400% improvement. (gossip about unqualifiable numbers :-D ) On Sun, 15 Jul 2007 17:13:42 -0500, Tom Marchant m42tom- [EMAIL PROTECTED] wrote: On Sun, 15 Jul 2007 12:44:34 -0700, Dean Kent wrote: - Original Message - From: Timothy Sipples [EMAIL PROTECTED] Dean Kent: Itanium likely could emulate zArch instructions faster than native zSeries systems can execute them No. If you have any published numbers to verify that, it would be very nice to see them. PSI claims that they can provide up to 350 MIPS using Itanium processors. They do not say how many processors they use to attain that A z9 UP is closer to 600 MIPS. -- Tom Marchant Regards Bruce Hewson -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
Bruce Hewson wrote: Interesting, Them guys gave a presentation over here some time ago. The numbers they quoted for a HP Superdome was 80 MIPS per engine, with up to 64 engines. Are they now stating that using Itaniums can generate 350 MIPS per engine, that means that upgrading from the original HP Superdome, to an Itanium based system, is a 400% improvement. (gossip about unqualifiable numbers :-D ) It means Itanium is 4.3 times faster than AMD. I doubt it. BTW: IMHO comparison of CPUs is senseless when we talk about *computer speed*. What about I/O ? I saw Hercules on PC, almost 40MIPS, but TSO logon time was over 5 minutes. Every I/O operation was slooow. IMHO the strengths of mainframe are: a) I/O b) OS -- Radoslaw Skorupka Lodz, Poland -- BRE Bank SA ul. Senatorska 18 00-950 Warszawa www.brebank.pl Sd Rejonowy dla m. st. Warszawy XII Wydzia Gospodarczy Krajowego Rejestru Sdowego, nr rejestru przedsibiorców KRS 025237 NIP: 526-021-50-88 Wedug stanu na dzie 01.01.2007 r. kapita zakadowy BRE Banku SA (w caoci opacony) wynosi 118.064.140 z. W zwizku z realizacj warunkowego podwyszenia kapitau zakadowego, na podstawie uchwa XVI WZ z dnia 21.05.2003 r., kapita zakadowy BRE Banku SA moe ulec podwyszeniu do kwoty 118.760.528 z. Akcje w podwyszonym kapitale zakadowym bd w caoci opacone. -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
R.S. [EMAIL PROTECTED] wrote in message news:[EMAIL PROTECTED] It means Itanium is 4.3 times faster than AMD. I doubt it. That would depend upon what benchmark. Itanium is very strong in floating point vs. anything AMD has to offer, but in integer it is merely adequate. However, I think the mistake might have been that the original number was 80 MIPS per CPU, while the 350 MIPS is apparently across 8. While going from 1 to 8 doesn't provide perfect scaling, and if you don't know what that scaling factor is extrapolating is difficult at best, you can presume that a single Itanium CPU would not provide much more (if not less) than the 80 MIPS mentioned. BTW: IMHO comparison of CPUs is senseless when we talk about *computer speed*. What about I/O ? I saw Hercules on PC, almost 40MIPS, but TSO logon time was over 5 minutes. Every I/O operation was slooow. IMHO the strengths of mainframe are: a) I/O b) OS I fully agree with this. The original question was, however, why people claim that the processor is slow. Obviously, there are many instructions that can't be compared directly because of the different target workloads - but one *can*, with the right numbers, make some kind of comparison. For example, we have seen it mentioned that it requires, on average, approx 17 instructions on Itanium to emulate one for the mainframe. If the 80 MIPS per processor is accurate, that really isn't all that bad considering the number of instructions being executed. Now all we need is to know what it would take for a mainframe to emulate the Itanium instruction set and we could have a nice comparison. ;-). As I said in my first post on this topic, the comparison that would be valid is system thoughput, not processor speed. There *are* benchmarks for this that could be used, I believe, because there are a few workloads that are common between mainframes and traditional Unix/Windows machines. SPECjbb comes to mind, and there might be others. With more emphasis on Linux, the OS becomes less of a differentiator, I think. Regards, Dean -- Radoslaw Skorupka Lodz, Poland -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
You left out: c) Amazingly good-looking, intelligent, and affable technical staff. Jon snip It means Itanium is 4.3 times faster than AMD. I doubt it. BTW: IMHO comparison of CPUs is senseless when we talk about *computer speed*. What about I/O ? I saw Hercules on PC, almost 40MIPS, but TSO logon time was over 5 minutes. Every I/O operation was slooow. IMHO the strengths of mainframe are: a) I/O b) OS /snip -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
On Mon, 16 Jul 2007 06:07:07 -0700, Dean Kent wrote: ... The original question was, however, why people claim that the processor is slow Since you are one of the people making that claim, perhaps you can answer the question. I would suggest that it is because some people look at raw clock rates. You might want to google for megahertz myth. As I said in my first post on this topic, the comparison that would be valid is system thoughput, not processor speed. Yes, you did say that, but also, On Fri, 13 Jul 2007 23:06:01 -0700, Dean Kent wrote: The mainframe MPU *is* slower than other platforms Itanium likely could emulate zArch instructions faster than native zSeries systems can execute them... Then, On Sun, 15 Jul 2007 12:44:34 -0700, Dean Kent wrote: Timothy Sipples wrote: Dean Kent: Itanium likely could emulate zArch instructions faster than native zSeries systems can execute them No. If you have any published numbers to verify that, it would be very nice to see them. The pace of improvements in the x86 world are quite stunning, because of the competitive nature of things. I find it hard to believe that IBM would spend the money for mainframe processors to keep pace when there isn't really much of an incentive to do so. Again, if there are any reliable comparisons between these processors, it would be great to see them. Otherwise, all we have are assertions. First of all, Itanium is not in the x86 world. Secondly, you have obviously not been keeping up with the improvements in mainframe technology over the last several years. Those improvements are indeed quite stunning. Far from keeping pace, mainframe technology has been leading. -- Tom Marchant -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
Radoslaw, I confess I don't see the connection between what Bruce said about the old versus new superdomes and your comment about Itanium and AMD. I can believe (with a grain of salt) the idea that the new superdomes are 4+ times faster than the originals and it has nothing to do with AMD. The original superdome was running an HP chip - the PA-RISC 8600 running at 552 MHz (HP published specs). I can vouch for that as we still have a couple of the original 552 MHz cell boards in our superdome. Current superdomes are running Itanium2 chips at 1.6 GHz. They've also significantly increased the memory and bus speeds and made other changes over the past 6-7 years. Rex -Original Message- From: IBM Mainframe Discussion List [mailto:[EMAIL PROTECTED] On Behalf Of R.S. Sent: Monday, July 16, 2007 3:35 AM To: IBM-MAIN@BAMA.UA.EDU Subject: Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article) Bruce Hewson wrote: Interesting, Them guys gave a presentation over here some time ago. The numbers they quoted for a HP Superdome was 80 MIPS per engine, with up to 64 engines. Are they now stating that using Itaniums can generate 350 MIPS per engine, that means that upgrading from the original HP Superdome, to an Itanium based system, is a 400% improvement. (gossip about unqualifiable numbers :-D ) It means Itanium is 4.3 times faster than AMD. I doubt it. BTW: IMHO comparison of CPUs is senseless when we talk about *computer speed*. What about I/O ? I saw Hercules on PC, almost 40MIPS, but TSO logon time was over 5 minutes. Every I/O operation was slooow. IMHO the strengths of mainframe are: a) I/O b) OS -- Radoslaw Skorupka Lodz, Poland -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
Tom Marchant [EMAIL PROTECTED] wrote in message news:[EMAIL PROTECTED] Yes, you did say that, but also, On Fri, 13 Jul 2007 23:06:01 -0700, Dean Kent wrote: The mainframe MPU *is* slower than other platforms Itanium likely could emulate zArch instructions faster than native zSeries systems can execute them... Yes, and obviously I was wrong. There was an interesting speculation several years ago about POWER being used in mainframes because it could do emulation as fast, or faster, than a native mainframe processor, hence my own speculation. Part of the problem, as has been mentioned, is that many of the instructions are designed for completely different purposes. However, what I was talking about as far as it being slower is the raw CPU speed for integer/floating point operations. Then, If you have any published numbers to verify that, it would be very nice to see them. The pace of improvements in the x86 world are quite stunning, because of the competitive nature of things. I find it hard to believe that IBM would spend the money for mainframe processors to keep pace when there isn't really much of an incentive to do so. Again, if there are any reliable comparisons between these processors, it would be great to see them. Otherwise, all we have are assertions. I see nothing at all wrong with asking for information. First of all, Itanium is not in the x86 world. I know very well that Itanium is not an x86 processor, however it does compete head-to-head with x86 in many markets so it *is* in the x86 world, and has to keep pace with the performance of those. The fact that it did not early on (and still struggles with it today) is one of the big reasons it has been more-or-less a flop in the market. Secondly, you have obviously not been keeping up with the improvements in mainframe technology over the last several years. Those improvements are indeed quite stunning. Far from keeping pace, mainframe technology has been leading. Now I would like to get some specifics from you, since you've made the statement. Can you point me to references that show mainframe technology has been leading - not in RAS or features, but in performance (which is the context of the discussion)? Is that comment with regards to other platforms, or only within the mainframe market? Does 'mainframe technology' mean something other than performance, or are you including the entire set of platform improvements? Regards, Dean -- Tom Marchant -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
On Mon, 16 Jul 2007 09:49:13 -0700, Dean Kent wrote: Secondly, you have obviously not been keeping up with the improvements in mainframe technology over the last several years. Those improvements are indeed quite stunning. Far from keeping pace, mainframe technology has been leading. Now I would like to get some specifics from you, since you've made the statement. Can you point me to references that show mainframe technology has been leading - not in RAS or features, but in performance (which is the context of the discussion)? I did. On Sun, 15 Jul 2007 21:17:55 -0500, Tom Marchant wrote: Actually, IBM is an acknowledged leader in processor design and microelectronics technology. I would suggest you read some of the articles in http://www.research.ibm.com/journal/rd51-12.html . Particularly, http://www.research.ibm.com/journal/rd/511/mayer.html . For some additional background, see http://www.research.ibm.com/journal/rd50-45.html http://www.research.ibm.com/journal/rd48-34.html http://www.research.ibm.com/journal/rd46-45.html You might think that 1.7 GHz is slow when you hear about Intel processor clock rates, but z/Architecture is considerably more complex than any Intel processor. It is quite remarkable that such a large and complex processor is able to achieve a cycle time that is equivalent to the time that it takes light to travel about 7 inches. Some specifics: 90-nm in the z9 SOI Copper wiring You could do the research if you wanted to. -- Tom Marchant -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
- Original Message - Some specifics: 90-nm in the z9 SOI Copper wiring You could do the research if you wanted to. Now you appear to be arguing just to argue. The question was on comparisons of technology, and specifically performance. Since Intel is already on 45nm process, I don't think you can call 90nm 'leading in technology'. Regards, Dean -- Tom Marchant -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
On Mon, 16 Jul 2007 13:53:39 -0700, Dean Kent wrote: Since Intel is already on 45nm process, I don't think you can call 90nm 'leading in technology'. Already? when will they begin shipments? They say 2H2007. The z9 has been shipping since September, 2005. I guess you don't think much of SOI or copper either. -- Tom Marchant -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
I guess you don't think much of SOI or copper either. As has been said on this thread, I think we are off-topic. All transactions consume: CPU Memory I/O Network Print! Spending too much time on one (small) component is a waste of time, breath, electrons. - Too busy driving to stop for gas! -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
Tom Marchant [EMAIL PROTECTED] wrote in message news:[EMAIL PROTECTED] On Mon, 16 Jul 2007 13:53:39 -0700, Dean Kent wrote: Since Intel is already on 45nm process, I don't think you can call 90nm 'leading in technology'. Already? when will they begin shipments? They say 2H2007. The z9 has been shipping since September, 2005. As I said - you seem to be arguing just to argue. Intel began shipping 90nm products in Feb. 2004, but that really isn't the point. Process size is not an indicator of performance, or feature set. I guess you don't think much of SOI or copper either. It has not been shown that either of these provides any benefit in performance, and has nothing at all to do with feature size. It was supposed to help with leakage, but Intel seems to be doing quite will without these. Lest you misunderstand me - I am not trying to say that Intel is 'better' than IBM, nor the other way around. I am not trying to say that x86 processors are 'better' than z9. Each has its strengths, and weaknesses. There is no 'one processor to rule them all'. However, IBM is positioning the mainframe to compete in some of the same markets that x86 competes. This means that customers will expect a direct comparison on performance - and rightly so. Using the argument that IBM is a leader in technology, and therefore z9 must be better than x86 is ludicrous, if that was your point. AMD is an IBM partner, and they use the same process technology - so this *cannot* be a reason z9 has any advantage over x86, and vice-versa. As far as the largest semiconductor manufacturers in the world, Intel is #1, with Samsung, TI, Toshiba and STmicro rounding out the top 5. IBM isn't listed in the top 10 list. AMD was listed as #7 as of Dec 2006, so the two leading x86 manufacturers are in the top 10. This, of course, has absolutely nothing to do with whether any given product is a better performer than any other. I mentioned that I find it hard to believe that IBM would invest in mainframe performance to the extent that x86 manufacturers would, considering the difference in the competitivness of the markets. It was stated that IBM invests $1.2B annually on mainframe RD (hardware, software and services). Intel, on the other hand, spends almost $6B on their semiconductor business alone. It should not be surprising that Intel is also a leader in technology - even if their primary product is the lowly x86 based processors. As for fault-tolerant systems, Stratus and NEC offer them (and likely others). I have had correspondence with one of the main architects of Stratus systems, and his background includes working for DEC on their fault-tolerant systems. Their website claims 99.9997% uptime, and they include many of the features that mainframers might consider unique to mainframes - using Xeon processors. The I/O performance that was once the realm of the mainframe is now available for other platforms as well. Consider that IBM uses fibre-channel and 3.5 FBA devices, just like everyone else - and emulate CKD for the mainframe. PCI/E and SATA provide error recovery capabilities. So, while the mainframe still enjoys a relatively comfortable niche, I don't think mainframers should be too smug about it. x86 processors are not just good for word processing, despite some comments to that effect. Making snide, derogatory remarks about x86 or other platforms is just as foolish as PC people making derogatory comments about the mainframe. It would be nice if people would post information that would further the dialog rather than simply to defend a position. I don't expect IBM, or IBM employees, to post information that is not already public - which is why I requested published numbers and comparisons, if there are any. I've been working with mainframes since 1976, and x86 based systems since about 1992. I'm certainly not a hardware engineer, and am no authority on all of the ins and outs of the various strengths and weaknesses of each platform. What I do find a little tiresome are the assertions and derogations about various platforms based upon 'common wisdom' rather than verifiable information. Yes, I occasionally find myself repeating some of this nonsense, but I hope I've shown that when the data is presented I'm willing to accept the facts and reform my opinions. I hope others can do likewise. Regards, Dean -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: PSI MIPS (was: Links to decent 'why the mainframe thrives' article)
Ted MacNEIL wrote: I guess you don't think much of SOI or copper either. As has been said on this thread, I think we are off-topic. All transactions consume: CPU Memory I/O Network Print! Spending too much time on one (small) component is a waste of time, breath, electrons. - Too busy driving to stop for gas! Yes we are getting a bit to much down the into the details. I don't think we could even agree with what 'better performing means. Which is better performing, a Corvette, or a semi-tractor? Depends, I would not want to attempt to haul 5 tons in a Corvette. They both can easily top 120 MPH, but the Corvette can get there a lot quicker than the semi and the Corvette can go even faster than 120 MPH. The semi can go a lot further before it has to stop for fueling. So which is better? Depends on your needs. You have 400 people that you must fly someplace. Which would you rather have a Boeing 747 or 40 Lear Jets? It depends, if you need to get 40 groups of 10 people each to 40 different locations, the Lear jets MAY be a better solution. Depends on how diverse the locations are. However if you need to get all 400 people to the same location, the 747 will be much less expensive and more efficient. Is a 2.8 Ghz processor more powerful than a 1.7 Ghz processor? If they are the same architecture, yes. If they are different architectures, then who knows. I the processor speed the only measure you should use in determining how powerful a computer is. Does an engine running at 9,000 RPM always move a car faster than an engine running at 5,000 RPM? Depends on the gearing. So which computer do you need? It depends? Which platform performs better? It depends on your performace requirments. -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html