Re: Entry point for a Mainframe?
The following message is a courtesy copy of an article that has been posted to bit.listserv.ibm-main,alt.folklore.computers as well. re: http://www.garlic.com/~lynn/2010e.html#80 Entry point for a Mainframe? http://www.garlic.com/~lynn/2010e.html#81 Entry point for a Mainframe? http://www.garlic.com/~lynn/2010e.html#83 Entry point for a Mainframe? http://www.garlic.com/~lynn/2010f.html#0 Entry point for a mainframe http://www.garlic.com/~lynn/2010f.html#2 Entry point for a mainframe of course, I had managed to offend the communication group as undergraduate in the 60s working on plug-compatible controller (originally done on interdata/3 ... eventually acquired by Perkin-Elmer and marketing at least thru much of the 80s ... P/E fed/gov marketing manager commented that channel interface board looked like it may have been the original wire-wrap design done at the univ. in the 60s). misc. past posts http://www.garlic.com/~lynn/subtopic.html#360pcm there was writeup blaming four of us for the clone controller business. in the late 90s, touring a merchant acquiring (very) large mainframe installation (multiple max'ed out CECs), there were these boxes handling calls from a significant percentage of the merchant point-of-sale terminals in the US. -- 42yrs virtualization experience (since Jan68), online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: Entry point for a Mainframe?
test -Original Message- From: IBM Mainframe Discussion List [mailto:ibm-m...@bama.ua.edu] On Behalf Of Anne Lynn Wheeler Sent: Sunday, March 14, 2010 12:08 PM To: IBM-MAIN@bama.ua.edu Subject: Re: Entry point for a Mainframe? The following message is a courtesy copy of an article that has been posted to bit.listserv.ibm-main,alt.folklore.computers as well. re: http://www.garlic.com/~lynn/2010e.html#80 Entry point for a Mainframe? http://www.garlic.com/~lynn/2010e.html#81 Entry point for a Mainframe? http://www.garlic.com/~lynn/2010e.html#83 Entry point for a Mainframe? http://www.garlic.com/~lynn/2010f.html#0 Entry point for a mainframe http://www.garlic.com/~lynn/2010f.html#2 Entry point for a mainframe of course, I had managed to offend the communication group as undergraduate in the 60s working on plug-compatible controller (originally done on interdata/3 ... eventually acquired by Perkin-Elmer and marketing at least thru much of the 80s ... P/E fed/gov marketing manager commented that channel interface board looked like it may have been the original wire-wrap design done at the univ. in the 60s). misc. past posts http://www.garlic.com/~lynn/subtopic.html#360pcm there was writeup blaming four of us for the clone controller business. in the late 90s, touring a merchant acquiring (very) large mainframe installation (multiple max'ed out CECs), there were these boxes handling calls from a significant percentage of the merchant point-of-sale terminals in the US. -- 42yrs virtualization experience (since Jan68), online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html CyberDefender has scanned this email for potential threats. Version 2.0 / Build 4.03.29.01 Get free PC security at www.cyberdefender.com -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: Entry point for a Mainframe?
The following message is a courtesy copy of an article that has been posted to bit.listserv.ibm-main,alt.folklore.computers as well. Anne Lynn Wheeler l...@garlic.com writes: the 3725 pieces of the numbers came from official corporate HONE configurator (sales marketing use for selling to customers) ... part of the presentation to fall '86 SNA architecture review board meeting in Raleigh http://www.garlic.com/~lynn/99.html#67 System/1 ? part of spring '86 common presentation on pu4/pu5 support in series/1 http://www.garlic.com/~lynn/99.html#70 Series/1 as NCP (was: Re: System/1 ?) re: http://www.garlic.com/~lynn/2010e.html#80 Entry point for a Mainframe? http://www.garlic.com/~lynn/2010e.html#81 Entry point for a Mainframe? http://www.garlic.com/~lynn/2010e.html#83 Entry point for a Mainframe? Comparison with 3725NCP System: * Higher availability * More reliable * More function * Improved Useability * Non-IBM Host Support * Much better connectivity * Much better performance * Fewer components * Easier to tune * Easier to tailor * Easier to manage * Less expensive ... snip ... SNA RUs were carried within real networking traffic. Part of the ease of implementing all of the above ... was whole infrastructure was real networking ... only dropping down to SNA emulation at boundary interfaces when necessary. One of the internal parties that was really interested was IMS hot-standby. Some of the IMS hot-standby configurations with tens of thousands of terminals was clocking at over 90minuts (sometimes a lot more) to re-establish all the sessions in fall-over scenario. IMS hot-standby wanted slight tweak to the high availability sessions (aka network session information was replicated within the distributed network) where shadow SNA sessions were created with the VTAM on the fall-over processor(s) ... instead of taking potentially hrs to get everything back up and running ... the (SNA/VTAM) session fall-over was as fast as the IMS hot-standby fall-over (all being spoofed by having a real non-SNA networking environment to do it from). -- 42yrs virtualization experience (since Jan68), online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: Entry point for a Mainframe?
A Wheeler writes: a lot of the software as service ... and cloud computing ... very analogous to oldtime online timesharing ... is partially being driven by super-efficient megadatacenters (coupled with ubiquitous high-speed connectivity) Agreed. There are a lot of similarities, but one difference is the ubiquity of the Internet. It's really an accident of history (telco monopolies) that the price-per-carried bit collapsed *after* the prices of CPU and storage did. So we went through (suffered?) an intermediate phase when computing architectures were principally constrained by high priced long distance networking (the PC revolution and then Client/Server). It's interesting viewing those phases through the rear view mirror. In many ways it's back to the future now. Of course, each phase leaves its marks, some of which last forever. To carry this thread back to the MP3000, it's worth noting that the z10 BC is an extremely Internet-friendly server. The MP3000 (sadly) wasn't. The z10 BC supports huge SSL handshake rates and high volume IPSec with its crypto capabilities, 64-bit addressing (for multiple big Java application serving and Apache HTTP z/OS workloads), highly concentrated Linux virtualization, a raft of modern Internet-related software (WebSphere Commerce, WebSphere Portal, WebSphere Dashboard Framework, Lotus ActiveInsight, etc., etc.), the latest (and uncompromised) OSA-Express functions, remote Internet-friendly hardware console support, and so on. Regardless of the size of business, it's a far better server for this new reality of the ubiquitous Internet. CNNIC is a good example of a new Internet-related customer, and they came into the fold with a z9 BC: http://www.ibm.com/press/us/en/pressrelease/27768.wss They wouldn't have been able to join the community of mainframe owners with an MP3000 (even solution time-adjusted to 1999), unfortunately. - - - - - Timothy Sipples IBM Consulting Enterprise Architect for New, Advanced, and/or Innovative Solutions (VCT) Based in Singapore Serving the Growth Markets E-Mail: timothy.sipp...@us.ibm.com -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: Entry point for a Mainframe?
To repeat an already over-used analogy, there really are trucks in this world, and they're quite popular, even if they do consume only diesel fuel. There are also bicycles. If you're starting a bicycle messenger service, buy a bicycle. There's nothing wrong with that. Thank goodness there are different vehicles for different mission profiles. To extend the above analogy, you can buy Fedex, UPS, or USPS service and avoid renting, leasing, or owning your own trucks or bicycles. Actually, UPS started on bicycles in the far Northwest, then went to motor bycycles, then trucks which, until sometime in the 70's or 80's, they actually manufactured themselves. Finally they went to jets with one of the largest fleets of jets in the world, thousands. I guess somewhere along the line, when they went to jets, they decided to stop manufacturing their delivery equipment. They were private until somewhere around Y2K. When they went public, they were very generous with their employees, many of whom became instant millionaires overnight. It is amazing what a little efficiency, courtesy, and consideration to their employees, their business model, can do in such a mundane world as package delivery. On Wed, Mar 10, 2010 at 8:47 PM, Timothy Sipples timothy.sipp...@us.ibm.com wrote: Steve Thompson writes: And the power requirements for this z10? Can it be plugged into a wall beside my desk lamp on a 30A circuit? No. So you can deduct a point for that if you like. That said, you wouldn't be able to dry your laundry in most U.S. homes if you insisted on 120 volts only. :-) To repeat an already over-used analogy, there really are trucks in this world, and they're quite popular, even if they do consume only diesel fuel. There are also bicycles. If you're starting a bicycle messenger service, buy a bicycle. There's nothing wrong with that. Thank goodness there are different vehicles for different mission profiles. Now, I happen to think that the System z10 BC as a (more) entry-level mainframe is in every way superior to the Multiprise 3000, with perhaps two exceptions: physical space (which includes weight) and the electric circuit requirement. In every other respect I can think of, it scores a lot higher. (Including economics, which is typically a nice way to compensate for those other two criteria.) If you're asking, Could the System z10 BC be even better? -- in the categories of space and electric circuit requirements, for example -- well sure, theoretically. But then it might be compromised in other dimensions. Again, I'm very fond of the MP3000, but its design required a number of compromises. By the way, an awful lot of small businesses are opting for Software as a Service offerings and choosing not to own or host their own servers, of any type. If you want a zero-footprint z/OS machine -- that sure beats the MP3000! -- it's available. To extend the above analogy, you can buy Fedex, UPS, or USPS service and avoid renting, leasing, or owning your own trucks or bicycles. If the world is already heading in that SaaS direction -- and it sure looks that way -- then a z10 footprint makes even more sense. Also (and you alluded to it, Steve), has anyone visited a data center lately? Think about those 1980s narratives: Years ago, the computer was so big, it filled an entire room Well, nowadays it's worse: The racks of servers are so numerous, they fill football fields, consume prodigious amounts of electricity, and run so hot it's getting impossible to cool them Progress! :-) The smallest, coolest running server in most data centers is the System z10. It's the *answer* to server sprawl. And perhaps you'd be surprised how many small businesses suffer from server sprawl. - - - - - Timothy Sipples IBM Consulting Enterprise Architect for New, Advanced, and/or Innovative Solutions (VCT) Based in Singapore Serving the Growth Markets E-Mail: timothy.sipp...@us.ibm.com -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html -- George Henke (C) 845 401 5614 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: Entry point for a Mainframe?
The following message is a courtesy copy of an article that has been posted to bit.listserv.ibm-main,alt.folklore.computers as well. timothy.sipp...@us.ibm.com (Timothy Sipples) writes: Agreed. There are a lot of similarities, but one difference is the ubiquity of the Internet. It's really an accident of history (telco monopolies) that the price-per-carried bit collapsed *after* the prices of CPU and storage did. So we went through (suffered?) an intermediate phase when computing architectures were principally constrained by high priced long distance networking (the PC revolution and then Client/Server). It's interesting viewing those phases through the rear view mirror. In many ways it's back to the future now. re: http://www.garlic.com/~lynn/2010e.html#78 Entry point for a Mainframe? recent post/thread in tcp/ip n.g. http://www.garlic.com/~lynn/2010e.html#73 NSF to Fund Future Internet Architecture (FIA) and similar comments in this (mainframe) post/thread http://www.garlic.com/~lynn/2010e.html#64 LPARs: More or Less? about telcos having very high fixed costs/expenses and significant increase in available bandwdith with all the dark fiber in the ground represented difficult chicken/egg obstacle (disruptive technology). The bandwidth hungry applications wouldn't appear w/o significant drop in use charges (but could still take a decade or more) ... and until the bandwidth hungry applications appeared, any significant drop in the useage charges would mean that they would operate deeply in the red during the transition. in the mid-80s, the hsdt project had a very interesting datapoint with communication group ... where we were deploying and supporting T1 and faster links. http://www.garlic.com/~lynn/subnetwork.html#hsdt The communication group then did a corporate study that claimed that there wouldn't be customer use of T1 until mid-90s (aka since they didn't have product that supported T1, the study supported customers not needing T1 for another decade). The problem was that 37x5 boxes didn't have T1 support ... and so what the communication group studied was fat pipes ... support for being able to operate multiple 56kbit links as single unit. For their T1 conclusions they plotted the number of fat pipes with 2, 3, 4, ..., etc 56kbit links. They found that number of fat pipes dropped off significantly at four or five 56kbit links and there were none above six. There is always the phrase about statistics lie ... well, what the communication group didn't appear to realize was that most telcos had tariff cross-over about five or six 56kbit links being about the same as a single T1 link. What they were seeing, was when customer requirement reached five 56kbit links ... the customers were moving to single T1 link supported by other vendors products (which was the reason for no fat pipes above six). The communication groups products were very oriented towards to the legacy dumb terminal paradigm ... and not the emerging peer-to-peer networking operation. In any case, a very quick, trivial survey by HSDT turned up 200 customers with T1 links (as counter to the communication group survey that customers wouldn't be using T1s until mid-90s ... because they couldn't find any fat pipes with more than six 56kbit links). this is analogous to communication group defining T1 as very high speed in the same period (in part because their products didn't support T1) ... mentioned in this post: http://www.garlic.com/~lynn/2010e.html#11 Crazed idea: SDSF for z/Linux the various internal politics all contributed to not letting us bid on the NSFNET backbone RFP ... even when the director of NSF wrote a letter to corporation ... and there were observations that what we already had running was at least five years ahead of RFP bid responses (to build something new). misc. old NSFNET related email from the period http://www.garlic.com/~lynn/lhwemail.html#nsfnet -- 42yrs virtualization experience (since Jan68), online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: Entry point for a Mainframe?
Also (and you alluded to it, Steve), has anyone visited a data center lately? Think about those 1980s narratives: Years ago, the computer was so big, it filled an entire room Well, nowadays it's worse: The racks of servers are so numerous, they fill football fields, consume prodigious amounts of electricity, and run so hot it's getting impossible to cool them Progress! :-) The smallest, coolest running server in most data centers is the System z10. It's the *answer* to server sprawl. And perhaps you'd be surprised how many small businesses suffer from server sprawl. Actually a previous client, a large Wall St investment house that survived the recent crisis, has so many blade servers in its data center they can't fit anymore in. So they have a pilot project to bring them up on LINUX under z/VM. Hurray for server consolidation, finally, a software solution beating a hardware solution. Amen On Thu, Mar 11, 2010 at 9:43 AM, Anne Lynn Wheeler l...@garlic.comwrote: The following message is a courtesy copy of an article that has been posted to bit.listserv.ibm-main,alt.folklore.computers as well. timothy.sipp...@us.ibm.com (Timothy Sipples) writes: Agreed. There are a lot of similarities, but one difference is the ubiquity of the Internet. It's really an accident of history (telco monopolies) that the price-per-carried bit collapsed *after* the prices of CPU and storage did. So we went through (suffered?) an intermediate phase when computing architectures were principally constrained by high priced long distance networking (the PC revolution and then Client/Server). It's interesting viewing those phases through the rear view mirror. In many ways it's back to the future now. re: http://www.garlic.com/~lynn/2010e.html#78 Entry point for a Mainframe? recent post/thread in tcp/ip n.g. http://www.garlic.com/~lynn/2010e.html#73 NSF to Fund Future Internet Architecture (FIA) and similar comments in this (mainframe) post/thread http://www.garlic.com/~lynn/2010e.html#64 LPARs: More or Less? about telcos having very high fixed costs/expenses and significant increase in available bandwdith with all the dark fiber in the ground represented difficult chicken/egg obstacle (disruptive technology). The bandwidth hungry applications wouldn't appear w/o significant drop in use charges (but could still take a decade or more) ... and until the bandwidth hungry applications appeared, any significant drop in the useage charges would mean that they would operate deeply in the red during the transition. in the mid-80s, the hsdt project had a very interesting datapoint with communication group ... where we were deploying and supporting T1 and faster links. http://www.garlic.com/~lynn/subnetwork.html#hsdt The communication group then did a corporate study that claimed that there wouldn't be customer use of T1 until mid-90s (aka since they didn't have product that supported T1, the study supported customers not needing T1 for another decade). The problem was that 37x5 boxes didn't have T1 support ... and so what the communication group studied was fat pipes ... support for being able to operate multiple 56kbit links as single unit. For their T1 conclusions they plotted the number of fat pipes with 2, 3, 4, ..., etc 56kbit links. They found that number of fat pipes dropped off significantly at four or five 56kbit links and there were none above six. There is always the phrase about statistics lie ... well, what the communication group didn't appear to realize was that most telcos had tariff cross-over about five or six 56kbit links being about the same as a single T1 link. What they were seeing, was when customer requirement reached five 56kbit links ... the customers were moving to single T1 link supported by other vendors products (which was the reason for no fat pipes above six). The communication groups products were very oriented towards to the legacy dumb terminal paradigm ... and not the emerging peer-to-peer networking operation. In any case, a very quick, trivial survey by HSDT turned up 200 customers with T1 links (as counter to the communication group survey that customers wouldn't be using T1s until mid-90s ... because they couldn't find any fat pipes with more than six 56kbit links). this is analogous to communication group defining T1 as very high speed in the same period (in part because their products didn't support T1) ... mentioned in this post: http://www.garlic.com/~lynn/2010e.html#11 Crazed idea: SDSF for z/Linux the various internal politics all contributed to not letting us bid on the NSFNET backbone RFP ... even when the director of NSF wrote a letter to corporation ... and there were observations that what we already had running was at least five years ahead of RFP bid responses (to build something new). misc. old NSFNET related email from the period http://www.garlic.com/~lynn/lhwemail.html#nsfnet
Re: Entry point for a Mainframe?
The following message is a courtesy copy of an article that has been posted to bit.listserv.ibm-main,alt.folklore.computers as well. gahe...@gmail.com (George Henke) writes: Actually a previous client, a large Wall St investment house that survived the recent crisis, has so many blade servers in its data center they can't fit anymore in. So they have a pilot project to bring them up on LINUX under z/VM. Hurray for server consolidation, finally, a software solution beating a hardware solution. re: http://www.garlic.com/~lynn/2010e.html#80 Entry point for a Mainframe? the claim was that in the 90s ... to have multiple applications co-exist in the same operating system required scarce, high-level skills (allocation, co-existance, capacity planning, etc) ... it was much easier and cheaper to throw hardware at the problem ... giving each application (and even application instance) its own dedicated hardware. rolling forward to couple years ago and organizations found themselves with thousands, tens of thousands, or even hundreds of thousands of these dedicated services ... all running at 5-10% utilization. virtualization, dynamic load balancing and some other technologies came together to support server consolidation (sometimes 10 or 20 to one running on essentially the identical hardware). part of the issue was that there was only very modest incremental skill level required for server consolidation (as compared to trying to getting lots of disparent applications to co-exist in the same system). lots of technologies are being pump into virtualization environment ... like dynamic migration of virtual machine to different hardware (even in different datacenters) for capacity reasons and/or continuous operation reasons. other posts in this thread: http://www.garlic.com/~lynn/2010e.html#68 Entry point for a Mainframe? http://www.garlic.com/~lynn/2010e.html#70 Entry point for a Mainframe? http://www.garlic.com/~lynn/2010e.html#71 Entry point for a Mainframe? http://www.garlic.com/~lynn/2010e.html#72 Entry point for a Mainframe? http://www.garlic.com/~lynn/2010e.html#78 Entry point for a Mainframe? -- 42yrs virtualization experience (since Jan68), online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: Entry point for a Mainframe?
From: Anne Lynn Wheeler l...@garlic.com To: IBM-MAIN@bama.ua.edu Sent: Thu, March 11, 2010 8:43:50 AM Subject: Re: Entry point for a Mainframe? SNIP The communication group then did a corporate study that claimed that there wouldn't be customer use of T1 until mid-90s (aka since they didn't have product that supported T1, the study supported customers not needing T1 for another decade). The problem was that 37x5 boxes didn't have T1 support ... and so what the communication group studied was fat pipes ... support for being able to operate multiple 56kbit links as single unit. For their T1 conclusions they plotted the number of fat pipes with 2, 3, 4, ..., etc 56kbit links. They found that number of fat pipes dropped off significantly at four or five 56kbit links and there were none above six. There is always the phrase about statistics lie ... well, what the communication group didn't appear to realize was that most telcos had tariff cross-over about five or six 56kbit links being about the same as a single T1 link. What they were seeing, was when customer requirement reached five 56kbit links ... the customers were moving to single T1 link supported by other vendors products (which was the reason for no fat pipes above six). The communication groups products were very oriented towards to the legacy dumb terminal paradigm ... and not the emerging peer-to-peer networking operation. In any case, a very quick, trivial survey by HSDT turned up 200 customers with T1 links (as counter to the communication group survey that customers wouldn't be using T1s until mid-90s ... because they couldn't find any fat pipes with more than six 56kbit links). this is analogous to communication group defining T1 as very high speed in the same period (in part because their products didn't support T1) ... mentioned in this post: http://www.garlic.com/~lynn/2010e.html#11 Crazed idea: SDSF for z/Linux the various internal politics all contributed to not letting us bid on the NSFNET backbone RFP ... even when the director of NSF wrote a letter to corporation ... and there were observations that what we already had running was at least five years ahead of RFP bid responses (to build something new). misc. old NSFNET related email from the period http://www.garlic.com/~lynn/lhwemail.html#nsfnet Ann In the mid 70's we had a T1 and we muxed it and IIRC we had 1 256K chunk and another chunk (sorry do not remember the speed) connected up to our 3745 and it worked really well (except a really strange bug which took us with the help of chance to figure out what the issue was). We were exercising it and kept it busy at least 20 out of 24 hours a day. I vaguely remember talking about the bug with IBM at the time (we were a small minority user of something like this at the time as IBM apparently only had a few people that seemed to know this part of NCP). Its not too surprising I guess that IBM really did not support a full T1 but if my memory (its iffy here) is correct it had something to do with the speed of the 3745 as to why IBM couldn't support it. SInce memory fades with time and I only remember small pieces we did seem to be on the bleeding edge at that time. Our bug turned out to not to have anything to do with NCP (per se) but I think if IBM would have had more experience they would have helped us find the issue sooner. IIRC there was semi documented information about lic weights (???) and you had to read it closely or you ended up with bad information. Sorry about the sketchiness but we are talking 35 years ago. Ed -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: Entry point for a Mainframe?
The following message is a courtesy copy of an article that has been posted to bit.listserv.ibm-main,alt.folklore.computers as well. ps2...@yahoo.com (Ed Gould) writes: In the mid 70's we had a T1 and we muxed it and IIRC we had 1 256K chunk and another chunk (sorry do not remember the speed) connected up to our 3745 and it worked really well (except a really strange bug which took us with the help of chance to figure out what the issue was). We were exercising it and kept it busy at least 20 out of 24 hours a day. I vaguely remember talking about the bug with IBM at the time (we were a small minority user of something like this at the time as IBM apparently only had a few people that seemed to know this part of NCP). Its not too surprising I guess that IBM really did not support a full T1 but if my memory (its iffy here) is correct it had something to do with the speed of the 3745 as to why IBM couldn't support it. SInce memory fades with time and I only remember small pieces we did seem to be on the bleeding edge at that time. re: http://www.garlic.com/~lynn/2010e.html#80 Entry point for a Mainframe? http://www.garlic.com/~lynn/2010e.html#81 Entry point for a Mainframe? 3705 in the 70s, 3725 in the 80s, 3745 later. this has (some?) 3745 withdrawn from marketing in sep2002 http://www.networking.ibm.com/nhd/webnav.nsf/pages/375:375prod.html 3745 wiki page http://en.wikipedia.org/wiki/IBM_3745 3745 wasn't announced until 1988 http://www-03.ibm.com/ibm/history/history/year_1988.html in mid-80s, Laguade had an experimental 3725 that was dedicated to running single T1. corporation did have 2701 in the 60s that supported T1 ... but the communication group in the 70s acquired increasingly narrow myopic focus on dumb terminals; they also leveraged corporate politics to keep other business units out of areas that thot even remotely touched on what they believed was their responsibility. this shows up, at least in the constant battles my wife had with the communication group when she was in POK responsible for loosely-coupled architecture ... and only temporary truces that she could use her own protocol for machine-to-machine communication within datacenter walls. some past posts mentioning her peer-coupled shared data architecture ... which except for IMS hot-standby, saw little uptake until sysplex. http://www.garlic.com/~lynn/submain.html#shareddata the narrow focus on dumb terminal became increasingly rigid in the 80s ... even tho early on real 3270s were being replaced with more capable ibm/pcs and terminal emulation http://www.garlic.com/~lynn/subnetwork.html#emulation 2701s were becoming increasingly long in the tooth during this period. in the mid-80s, federal systems division did come up with zirpel card for S/1 that supported T1 (for special gov. bids). however, for the most part, if other business units couldn't be kept out purely using the argument that only communication group could produce communication related products ... then there was always studies like the fat pipe argument that would be presented to corporate hdqtrs ... showing that customers didn't even want such products. it was also motivation for senior engineer from disk division getting presentation slippped into the internal world-wide communication group annual conference ... where the opening statement was that the communication group was going to be responsible for the demise of the disk division. I got HSDT involved with babybell that had done NCP emulation on series/1 ... and I was deep into trying to put it out as a corporate product (and really got involved in interesting politics ... this is scenario where the truth is really stranger than fiction). In any case, I gave a presentation on the work at fall '86, SNA architecture review board meeting in Raleigh ... quickly putting out a series/1 based version while quickly porting to RIOS chip (aka rs/6000). the 3725 pieces of the numbers came from official corporate HONE configurator (sales marketing use for selling to customers) ... part of the presentation to fall '86 SNA architecture review board meeting in Raleigh http://www.garlic.com/~lynn/99.html#67 System/1 ? part of spring '86 common presentation on pu4/pu5 support in series/1 http://www.garlic.com/~lynn/99.html#70 Series/1 as NCP (was: Re: System/1 ?) -- 42yrs virtualization experience (since Jan68), online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: Entry point for a Mainframe?
Steve Thompson writes: And the power requirements for this z10? Can it be plugged into a wall beside my desk lamp on a 30A circuit? No. So you can deduct a point for that if you like. That said, you wouldn't be able to dry your laundry in most U.S. homes if you insisted on 120 volts only. :-) To repeat an already over-used analogy, there really are trucks in this world, and they're quite popular, even if they do consume only diesel fuel. There are also bicycles. If you're starting a bicycle messenger service, buy a bicycle. There's nothing wrong with that. Thank goodness there are different vehicles for different mission profiles. Now, I happen to think that the System z10 BC as a (more) entry-level mainframe is in every way superior to the Multiprise 3000, with perhaps two exceptions: physical space (which includes weight) and the electric circuit requirement. In every other respect I can think of, it scores a lot higher. (Including economics, which is typically a nice way to compensate for those other two criteria.) If you're asking, Could the System z10 BC be even better? -- in the categories of space and electric circuit requirements, for example -- well sure, theoretically. But then it might be compromised in other dimensions. Again, I'm very fond of the MP3000, but its design required a number of compromises. By the way, an awful lot of small businesses are opting for Software as a Service offerings and choosing not to own or host their own servers, of any type. If you want a zero-footprint z/OS machine -- that sure beats the MP3000! -- it's available. To extend the above analogy, you can buy Fedex, UPS, or USPS service and avoid renting, leasing, or owning your own trucks or bicycles. If the world is already heading in that SaaS direction -- and it sure looks that way -- then a z10 footprint makes even more sense. Also (and you alluded to it, Steve), has anyone visited a data center lately? Think about those 1980s narratives: Years ago, the computer was so big, it filled an entire room Well, nowadays it's worse: The racks of servers are so numerous, they fill football fields, consume prodigious amounts of electricity, and run so hot it's getting impossible to cool them Progress! :-) The smallest, coolest running server in most data centers is the System z10. It's the *answer* to server sprawl. And perhaps you'd be surprised how many small businesses suffer from server sprawl. - - - - - Timothy Sipples IBM Consulting Enterprise Architect for New, Advanced, and/or Innovative Solutions (VCT) Based in Singapore Serving the Growth Markets E-Mail: timothy.sipp...@us.ibm.com -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: Entry point for a Mainframe?
timothy.sipp...@us.ibm.com (Timothy Sipples) writes: By the way, an awful lot of small businesses are opting for Software as a Service offerings and choosing not to own or host their own servers, of any type. If you want a zero-footprint z/OS machine -- that sure beats the MP3000! -- it's available. To extend the above analogy, you can buy Fedex, UPS, or USPS service and avoid renting, leasing, or owning your own trucks or bicycles. If the world is already heading in that SaaS direction -- and it sure looks that way -- then a z10 footprint makes even more sense. a lot of the software as service ... and cloud computing ... very analogous to oldtime online timesharing ... is partially being driven by super-efficient megadatacenters (coupled with ubiquitous high-speed connectivity) Microsoft: PUE of 1.22 for Data Center Containers http://www.datacenterknowledge.com/archives/2008/10/20/microsoft-pue-of-122-for-data-center-containers/ Google: The World's Most Efficient Data Centers http://www.datacenterknowledge.com/archives/2008/10/01/google-the-worlds-most-efficient-data-centers/ i was at presentation that had a claim about google's very careful crafting of their servers ... results in them having price/performance about 1/3rd that of their next closest competitor 215 Data Centers to Participate in EPA Study http://www.datacenterknowledge.com/archives/2008/06/30/215-data-centers-to-participate-in-epa-study/ ibm's entry in some of this http://www.datacenterknowledge.com/archives/2009/12/02/ibm-steps-up-its-partner-driven-container-game/ http://www.datacenterknowledge.com/archives/2008/06/11/ibm-launches-modular-data-centers-containers/ http://www.datacenterknowledge.com/archives/2008/06/11/a-look-at-ibms-data-center-container/ http://www-05.ibm.com/lt/ibmforum/pdf/ibm_data_center_family_maite_frey.pdf http://www-03.ibm.com/press/us/en/pressrelease/29058.wss http://www.itjungle.com/tfh/tfh120709-story08.html http://datacenterjournal.com/content/view/3392/41/ http://www.theregister.co.uk/2009/12/07/ibm_data_center_containers/ the rack references in the above seems to imply that it isn't a mainframe play some survey here of emerging operations http://perspectives.mvdirona.com/2009/04/25/RandyKatzOnHighScaleDataCenters.aspx -- 42yrs virtualization experience (since Jan68), online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: Entry point for a Mainframe?
-Original Message- From: IBM Mainframe Discussion List [mailto:ibm-m...@bama.ua.edu] On Behalf Of Timothy Sipples Sent: Tuesday, March 09, 2010 6:36 AM To: IBM-MAIN@bama.ua.edu Subject: Re: Entry point for a Mainframe? In contrast, today's smallest capacity IBM mainframe is the System z10 = BC model 2098-A01, rated at approximately 26 MIPS and a mere 3 MSUs. If yo= u need to upgrade, you can even move to a 4 MSU model, then 5, etc., and = way past 10 times the capacity of a maxed out MP3000. (Very nice!) SNIPPAGE And the power requirements for this z10? Can it be plugged into a wall beside my desk lamp on a 30A circuit? The entry point for a mainframe needs to be compared to the entry point of a x86 server. It can contain 2TB of RAID5 storage (I know, it could be more or less). It can contain 4 CPUs (2 physical with dual core) -- yes, it could be more or less. But this is where the FLEX/ES box came into this. And IBM's offering was the MP2000 which was replaced by the MP3000 (which us Tier 2 guys could sell, but not the next level up -- which meant we were on a death march, but I digress). But what were the environmental requirements for these boxes? If you go to the next level of server, and leave the clusters of floor models, you have to go to a rack and the blades. I don't think this is an entry level. And look at the increased environmentals. So are we saying, that after 20+ years, the entry level mainframe is similar to the S/360-20, where we have the CPU, the Tape drives and controller the two (or 4) 2311 drives, 1 or two printers and a 2501 card reader with a 1442 punch/reader? All requiring 220 or 208 3Phase? And the Wintell system is a monitor, keyboard, mouse, CPU box, Ethernet cable to a hub/switch and plugs into a wall at 125VAC 60Hz. So the z10 requires a raid box and what other externals? And what power requirements? And how much in the way of environmentals? See, entry level systems need a bit more thought than our current box can be configured with 3 MSUs and ... If you made no provision for entry level, and sales people are incented to kick small mainframes out because you can sell more x86 hardware... then you will not be planting seeds that will become the large systems in the future. So the entry point for a mainframe is rather high, because IBM says so. Regards, Steve Thompson -- Opinions expressed by this poster may not reflect those held by poster's employer -- -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: Entry point for a Mainframe?
The following message is a courtesy copy of an article that has been posted to bit.listserv.ibm-main,alt.folklore.computers as well. Anne Lynn Wheeler l...@garlic.com writes: has anybody gotten hands on intel 6core gulftown with two threads per core? ... there is reference that some chips might be sold with only four cores enabled (lower price?). http://en.wikipedia.org/wiki/Gulftown_%28microprocessor%29 what is the chance of beating 1000MIPS?? re: http://www.garlic.com/~lynn/2010e.html#68 Entry point for a Mainframe? http://www.garlic.com/~lynn/2010e.html#70 Entry point for a Mainframe? http://www.garlic.com/~lynn/2010e.html#71 Entry point for a Mainframe? possibly $5/370-mip??? vendors have been sorting chips ... chips failing higher speed test, being classed at lower rate for lower price. however, there seems to be some additional sorting ... apparently oriented towards overclocking gaming marketing ... that pushes higher rates ... and are sold at premium price. brand names are starting to offer boxes with such chips ... when it use to be just niche, offbrand players. some of the reduced core chips aren't necessarily just pricing ... sometimes it may be chip defects that would ordinarily have the whole chip going to trashbin ... localized defects may be core specific ... rest of the chip still being useable. other recent chip/foundary posts http://www.garlic.com/~lynn/2010e.html#62 z9 / z10 instruction speed(s) http://www.garlic.com/~lynn/2010e.html#66 z9 / z10 instruction speed(s) late 70s early 80s, single chip processors were starting to appear that drastically reduced cost of building computer systems ... and saw lots of vendors starting to move into the market. However, the cost of producing proprietary operating system hadn't come done ... so overall costs weren't reduced that much and therefor the price that the system could be offered to customers wouldn't come down. I've frequently commented those economics significantly contributed to the move to unix offerings ... vendors could ship unix on their platform for enormously lower cost (similar to the cost reduction offered by single chip processors) compared to every vendor doing their own proprietary operating system from scratch. A similar argument was used in the IBM/ATT effort moving higher level pieces of UNIX to stippred down TSS/370 base (the cost of adding mainframe ras, erep, device support, etc ... being several times larger than plain unix port). reference in this recent post mentioning adtech conference i did (that including presentations on both the unix/ssup activity as well as running cms applications on mvs): http://www.garlic.com/~lynn/2010e.html#17 Senior Java Developer vs. MVS Systems Programmer (warning: Conley rant) conference was also somewhat the origins for the VM/XB (ZM) effort (also mentioned in the above). the effort was then declared strategic, hundreds of people writting specs, and then collapsed under its own weight (somewhat a mini-FS). The strategic scenario was doing microkernel (somewhat akin to tss/370 ssup effort for ATT/unix) that had (at least) all the mainframe ras, erep and device support ... that could be used as common base for all the company's operating system offerings (the costs to the company in this area was essentially fully replicated for every operating system offering). in later 80s, having aix/370 (project that ported UCLA's Locus unix-clone to both 370 386) run under vm370 was aix/370 being able to rely on vm370 RAS (cost of adding that RAS directly to aix/370 was many times larger than the simple port of locus to 370). In recent years, increasing amounts of RAS is moving into intelligent hardware ... somewhat mitigating the duplication of effort in the operating systems. -- 42yrs virtualization experience (since Jan68), online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: Entry point for a Mainframe?
(Other) Steve: the smallest Multiprise 3000, model 7030-H30, was rated at approximately 60 MIPS and 11 MSUs. It did not offer sub-capacity IBM software licensing. GOLC was available but not zNALC nor Solution Edition offerings. You could get an IFL, but unfortunately it wasn't very exciting, and I don't think you could get an IFL-only machine. Parallel Sysplex, GDPS, Coupling Facilities, and FICON/FCP (at any speed) were flat out unavailable. Networking was...well, it had networking at least. Crypto? No, afraid not, nor any decimal floating point or 64-bit instructions. The next upgrade was a giant jump to an H50 with approximately 117 MIPS and exactly 20 MSUs. (Yowza!) I can't remember for sure, but I think it was a disruptive upgrade requiring an outage. IBM wasn't able to offer a same-serial upgrade to a subsequent model, unfortunately, but IBM always tries to offer attractive upgrades despite that limitation. (Please give us a ring.) It did have (slow) internal disk storage, but that topped out at about 200 GB, or less than a single EAV these days. It had limited availability-related design features. In contrast, today's smallest capacity IBM mainframe is the System z10 BC model 2098-A01, rated at approximately 26 MIPS and a mere 3 MSUs. If you need to upgrade, you can even move to a 4 MSU model, then 5, etc., and way past 10 times the capacity of a maxed out MP3000. (Very nice!) You can even upgrade for a day (Capacity On Demand) for a small amount of money, to handle a temporary surge, and there are lots of other new CoD features available. Also, 3 MSU sub-capacity licensing is available for practically all IBM software on the z10 BC, and you've got zELC, EWLC, MWLC, zNALC, and Solution Edition offerings on top of that. Not to mention full speed engine zAAPs and zIIPs in addition to beefy IFLs (with an advertised price of 47.5K in 2010 dollars not 125K in 1999 dollars). And of course it's a vastly better machine in so many other respects. Oh, and it has an advertised entry price lower than the Multiprise 3000's, especially in inflation-adjusted dollars. Yes, the z10 BC is physically bigger (and with its disk added), so let's subtract a point for that. But I'd say the z10 BC offers a very compelling set of improvements. Don't get me wrong: I'm quite fond of the Multiprise 3000, and I somewhat understand the hagiographic memories. But the z10 BC is an amazing machine and definitely represents significant progress. - - - - - Timothy Sipples IBM Consulting Enterprise Architect for New, Advanced, and/or Innovative Solutions (VCT) Based in Singapore Serving the Growth Markets E-Mail: timothy.sipp...@us.ibm.com -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Entry point for a Mainframe?
On Tue, Mar 9, 2010 at 7:35 AM, Timothy Sipples timothy.sipp...@us.ibm.com wrote: (Other) Steve: the smallest Multiprise 3000, model 7030-H30, was rated at approximately 60 MIPS and 11 MSUs. It did not offer sub-capacity IBM software licensing. GOLC was available but not zNALC nor Solution Edition offerings. You could get an IFL, but unfortunately it wasn't very exciting, and I don't think you could get an IFL-only machine. You could get an IFL-only machine if you twisted IBM's arm. We had one at Linuxcare/Levanta but I suspect it was the only one on the planet :-) ...phsiii (no, this isn't important to your point; added for historical purposes!) -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: Entry point for a Mainframe?
-Original Message- From: IBM Mainframe Discussion List [mailto:ibm-m...@bama.ua.edu] On Behalf Of Anne Lynn Wheeler Sent: Tuesday, 9 March 2010 3:50 PM To: IBM-MAIN@bama.ua.edu Subject: Re: Entry point for a Mainframe? note that the supercomputer market is starting to latch onto the GPUs developed for high-end graphics in the gaming market (starting to push thousand cores/GPU) ... it would be interesting to see if any emulator pieces could be mapped to GPU with hundreds/thousands of cores. Taking advantage of GPU parallelism... I think this is the paper that won a computing prize a few years back? http://www.cs.virginia.edu/~skadron/Papers/bakkum_sqlite_gpgpu10.pdf -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Entry point for a Mainframe?
In various threads there has been mention of IBM neglecting the entry level or small z/OS shops. Taking a look at the smallest of the z10s available z10-BC 2098-A01 3-MSU (29 MIPS) z10-EC 2097-401 27-MSU (217 MIPS) z10-EC 2097-701 115-MSU (924 MIPS) I wonder what is considered an entry level or small shop these days. I get along on an older z990-302 132-MSUs (832 MIPS) and it seems pretty small to me. Oddly enough the 2097-701 is less MSUS but more MIPS (yes I know it is marketing and has no basis in reality). And yes I know that historically the world ran on a dozen or so 1 MIPS boxes but where would progress be without bloat. Could you actually get any productive work done on a 3 MSU box today? z/OS, TSO, BATCH, CICS, DB2, Websphere, and associated monitors et. al. And if you wanted more than one LPAR? (Please do not continue the LPARS more or less thread here). What would be considered the smallest usable (FSVO usable) box? (no ZIIP/ZAAP allowed just a new z/OS workload). I know it depends but if I was going to get a Mainframe for the first time and I was a growing company I would think I would have at least some sort of CICS or DB2 (or equivalent) workload. I could of course be migrating from another platform and my entry point could be larger (after all I KNOW a z/OS based Mainframe IS the best platform in the world ;-) ). I guess the question is where do we expect a new z/OS shop to come from and what would the entry point be? -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: Entry point for a Mainframe?
-Original Message- From: IBM Mainframe Discussion List [mailto:ibm-m...@bama.ua.edu] On Behalf Of Ken Porowski Sent: Monday, March 08, 2010 4:50 PM To: IBM-MAIN@bama.ua.edu Subject: Entry point for a Mainframe? In various threads there has been mention of IBM neglecting the entry level or small z/OS shops. Taking a look at the smallest of the z10s available =20 z10-BC 2098-A01 3-MSU (29 MIPS) z10-EC 2097-401 27-MSU (217 MIPS) z10-EC 2097-701 115-MSU (924 MIPS) I wonder what is considered an entry level or small shop these days. I get along on an older z990-302 132-MSUs (832 MIPS) and it seems pretty small to me. Oddly enough the 2097-701 is less MSUS but more MIPS (yes I know it is marketing and has no basis in reality). And yes I know that historically the world ran on a dozen or so 1 MIPS boxes but where would progress be without bloat. =20 Could you actually get any productive work done on a 3 MSU box today? z/OS, TSO, BATCH, CICS, DB2, Websphere, and associated monitors et. al.=20 And if you wanted more than one LPAR? (Please do not continue the LPARS more or less thread here). SNIPPAGE I could, and have, run a department off of a P390 with OS/390 V1R1, of 10 people and still had sub second response time. With a MP/3000 type machine (which I think is the truly entry level), having hard drives on board, Ethernet adapter on board (that works, Dave), and external tape drives. With this type of set up (at 30MIPS), I'm quite sure you can handle 100 CICS users, plus a few programmers using TSO. This is the level of machine IBM killed when they pulled the plug on the FLEX/ES boxes. And those boxes (FLEX/ES) were upgradeable (as I understand it) to be able to connect to the standard RAID boxes, and even have CTCA between them (once they had ESCON capability), so that you could grow into a sysplex. And what did such a box cost compared to the z10-BC? That would have been a drop, plug and play environment (pretty much a turn-key system). Regards, Steve Thompson -- Opinions expressed by this poster may not reflect those of poster's employer -- -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: Entry point for a Mainframe?
The following message is a courtesy copy of an article that has been posted to bit.listserv.ibm-main,alt.folklore.computers as well. steve_thomp...@stercomm.com (Thompson, Steve) writes: This is the level of machine IBM killed when they pulled the plug on the FLEX/ES boxes. And those boxes (FLEX/ES) were upgradeable (as I understand it) to be able to connect to the standard RAID boxes, and even have CTCA between them (once they had ESCON capability), so that you could grow into a sysplex. And what did such a box cost compared to the z10-BC? That would have been a drop, plug and play environment (pretty much a turn-key system). this is flex presentation done at mar2005 baybunch meeting (5 yrs ago) http://www.baybunch.org/prezos/zbb.pdf a major FLEX platform was sequent (before ibm bought sequent). we did some consulting for Steve Chen when he was CTO at sequent ... and there were customers that had escon attachments (ibm connectivity) for sequent (numa) box (up to 256 intel shared memory multiprocessor). I know of at least one sequent numa customer (in 90s, before sequent bought by ibm) had escon and 3990 tape drives. sequent numa supported shared disk, raid, cluster, FCS (FCS open systems from early 90s flavor of ibm's proprietary ficon), etc ... all in 90s before being bought by ibm. 2000 competitive analysis of (unix) clusters http://h30097.www3.hp.com/dhba_ras.pdf above includes discussion sequent's cluster implementation (some number of loosely-coupled/clustered 256 processor tightly-coupled machines, say 4*256-way for 1024 processor complex). above ranking/comparion also includes our ha/cmp that we started in late 80s http://www.garlic.com/~lynn/subtopic.html#hacmp this old post discusses jan92 meeting regarding ha/cmp 128-way cluster operation (using FCS) http://www.garlic.com/~lynn/95.html#13 just a couple weeks before project was transferred and we were told we couldn't work on anything with more than four processors. related email on ha/cmp scaleup and fcs: http://www.garlic.com/~lynn/lhwemail.html#medusa sequent had specialized in commercial unix markets. (after departing) we were called in to consult with small client/server startup that wanted to do payment transactions on their server ... the startup had also invented this technology called SSL they wanted to use (the result is now frequently called electronic commerce). one of the things happening during this period was lots of servers were starting to experience heavy processor overload with web operation. the small client/server startup was growing and having to add increasing numbers of servers to handle their various kinds of web traffic. finally they installed a sequent system ... and things settled down. in turned out that sequent had fixed the networking implementation that was absorbing majority of server processing on other platforms. sequent explained that they had encountered the specific problem with commercial accounts supporting 20,000 (terminal) telnet sessions ... long before other platforms started experiencing the same networking problem with large number of HTTP/HTTPS connections (94/95 timeframe). Somewhat later, other platforms started distributing fixes for the tcp/ip processor overhead problem. for other topic drift ... part of the effort for electronic commerce was deploying something called a payment gateway ... that took payment transactions tunneled thru SSL, from webservers on the internet and passed tham to acquiring processor. misc. past posts mentioning payment gateway http://www.garlic.com/~lynn/subnetwork.html#gateway -- 42yrs virtualization experience (since Jan68), online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: Entry point for a Mainframe?
The following message is a courtesy copy of an article that has been posted to bit.listserv.ibm-main,alt.folklore.computers as well. Anne Lynn Wheeler l...@garlic.com writes: a major FLEX platform was sequent (before ibm bought sequent). we did some consulting for Steve Chen when he was CTO at sequent ... and there were customers that had escon attachments (ibm connectivity) for sequent (numa) box (up to 256 intel shared memory multiprocessor). I know of at least one sequent numa customer (in 90s, before sequent bought by ibm) had escon and 3990 tape drives. re: http://www.garlic.com/~lynn/2010e.html#68 Entry point for a Mainframe? small typo ... that is 3590 tape drive ... some old email about large mainframe customer with some sequent multiprocessor machines and sponsoring attaching 3590s to sequent machines. Date: Mon, 30 Oct 1995 09:49:55 -0800 From: wheeler In original sept. meeting we had thot we would have drives from IBM by 11/1/95 ... and would be able to loan (at least) one drive to sequent for development testing. they currently have one 3590 drive for the project attached to a dynix 2.3 system. the 3590 driver (w/o stacker support) will be thru beta test on 11/1/95 ... but will continue thru various kinds of product test/support. We require a 3590 driver (eventually w/stacker support) for a dynix 4.2 system. Sequent estimates approximately 7-10 days to port the 2.3 driver to 4.2 ... after availability of 3590 drive on dynix 4.2 level system. We've tentatively estimated that we might have a loaner 3590 drive for them on or around mid. Dec. ... snip ... more than year later: Date: Wed, 11 Dec 1996 12:31:49 -0800 To: wheeler 1) The 3590 driver SILI handling is wrong. This makes varying sized blocks impossible to implement. 2) The writev() implementation cannot be used for scatter gather. A possible solution would be to write a special interface for direct QCIC DMA gathering of block fragments. To not have this means sever memory/cache overhead in reconstructing new blocks for record inserts and length changes. Sequent was asked to provide a cost for access to the PTX kernel code for purposes of estimating the effort to add code to support a true gather function for whole blocks on tape. The $$ figure was never given. ... snip ... sequent wiki http://en.wikipedia.org/wiki/Sequent_Computer_Systems above mentions oracle parallel server high availability on sequent in 1993. there was some folklore that design for the implementation had come from some other vendors implementation. it also discusses ibm purchase and then seuquent is gone. -- 42yrs virtualization experience (since Jan68), online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
Re: Entry point for a Mainframe?
The following message is a courtesy copy of an article that has been posted to bit.listserv.ibm-main,alt.folklore.computers as well. Mike mham...@bellsouth.net writes: Hey Lynn, I remember giving that presentation to the Bay Bunch!! Just for what it's worth, using today's Intel based processors and IBM's zPDT software based system (some might call it emulator) we are getting well over 100 MIPS per core. (I suspect FLEX-ES version 8 sould also be in this range, if allowed on the market.) With a quad core processor, running 3 enabled processors, thats somewhere in the range of 300 - 350 MIPS in a relatively inexpensive system. zPDT is only available for developers today though. IBM is very cautions about making any comments about possible commercial availability. Mike Hammock m...@hammocktree.us re: http://www.garlic.com/~lynn/2010e.html#68 Entry point for a Mainframe? http://www.garlic.com/~lynn/2010e.html#70 Entry point for a Mainframe? relative off the shelf, white box, over-clocked, quiet (liquid cooling, low-noise fans, low-noise case), 64bit 4-core ... say $3k-$6k (tens of gbytes memory, six disks, terabyte or larger) ... done for gaming market. possibly 25%-50% faster than stock chip. processor cache is important. I've got a large DBMS implementation that runs faster on stock 1.7ghz chip with 2mbyte cache than it does on stock 3.4ghz chip that only has 512kbyte cache. has anybody gotten hands on intel 6core gulftown with two threads per core? ... there is reference that some chips might be sold with only four cores enabled (lower price?). http://en.wikipedia.org/wiki/Gulftown_%28microprocessor%29 what is the chance of beating 1000MIPS?? note that the supercomputer market is starting to latch onto the GPUs developed for high-end graphics in the gaming market (starting to push thousand cores/GPU) ... it would be interesting to see if any emulator pieces could be mapped to GPU with hundreds/thousands of cores. -- 42yrs virtualization experience (since Jan68), online at home since Mar1970 -- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
