TOD Sync acroos platforms (IBM and non-IBM) - STP?
Dear folks; We have a necessity to synchronize our IBM Parallel Sysplex with other platforms, i.e. HP Non Stop, Sun, PSeries, etc. I'm aware that PS does not support NTP. I was thinking if with STP we can achieve this but with the information I have, I think the answer seems to be a big NO. Any suggestion?. How other has addresed this requirement?. TIA, Giovanni -- 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: TOD Sync acroos platforms (IBM and non-IBM) - STP?
To what level of precision is synchronization required? To a 0.1 / second? To a 0.01 / second? To a 0.001 / second? If the first, or perhaps the second, if all servers sync to NIST in Boulder CO, then they'll all be the same. "NTP synchronization" amongst all the platforms you cite will effectively be accomplished. If a higher level of precision is required, then that's a different class of problem. In my experience, in general data processing environments, the problem identified is that server 1 says 10:05 am, server 2 says 10:07 am, router 3 says 09:44 am, and the correct time is actually 10:04:45 am. Getting servers all synced to NIST (or whatever your local government's time source might be) solves most problems. Again, if all servers sync to NIST (to the level of accuracy each server supports), the problem might well be solved. Both the Sysplex Timer and the new STP protocol will support synchronizing with the government time source. Brian On Thu, 16 Mar 2006 17:17:00 -0600, Giovanni Cerquone <[EMAIL PROTECTED]> wrote: >Dear folks; > >We have a necessity to synchronize our IBM Parallel Sysplex with other >platforms, i.e. HP Non Stop, Sun, PSeries, etc. I'm aware that PS does not >support NTP. > >I was thinking if with STP we can achieve this but with the information I >have, I think the answer seems to be a big NO. > >Any suggestion?. How other has addresed this requirement?. > >TIA, > >Giovanni > -- 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: TOD Sync acroos platforms (IBM and non-IBM) - STP?
Run the NTP server on the parallel sysplex and have the other platforms use it as the time source. Jerry Whitteridge Safeway Inc > -Original Message- > From: IBM Mainframe Discussion List > [mailto:[EMAIL PROTECTED] On Behalf Of Giovanni Cerquone > Sent: Thursday, March 16, 2006 3:17 PM > To: IBM-MAIN@BAMA.UA.EDU > Subject: TOD Sync acroos platforms (IBM and non-IBM) - STP? > > Dear folks; > > We have a necessity to synchronize our IBM Parallel Sysplex > with other platforms, i.e. HP Non Stop, Sun, PSeries, etc. > I'm aware that PS does not support NTP. > > I was thinking if with STP we can achieve this but with the > information I have, I think the answer seems to be a big NO. > > Any suggestion?. How other has addresed this requirement?. > > TIA, > > Giovanni > > -- > 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 > > "MMS " made the following annotations. -- Warning: All e-mail sent to this address will be received by the Safeway corporate e-mail system, and is subject to archival and review by someone other than the recipient. This e-mail may contain information proprietary to Safeway and is intended only for the use of the intended recipient(s). If the reader of this message is not the intended recipient(s), you are notified that you have received this message in error and that any review, dissemination, distribution or copying of this message is strictly prohibited. If you have received this message in error, please notify the sender immediately. == -- 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: TOD Sync acroos platforms (IBM and non-IBM) - STP?
What release of z/OS are you? Either Keith Clapp's SNTP code or the SNTP server that is part of z/OS. Have all the other platforms synchronize to z/OS. We have been doing that for about 4 years - seems to be working well once we overcame the political/idealogical problem that z/OS could actually do this. Dear folks; We have a necessity to synchronize our IBM Parallel Sysplex with other platforms, i.e. HP Non Stop, Sun, PSeries, etc. I'm aware that PS does not support NTP. I was thinking if with STP we can achieve this but with the information I have, I think the answer seems to be a big NO. Any suggestion?. How other has addresed this requirement?. TIA, Giovanni -- 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: TOD Sync acroos platforms (IBM and non-IBM) - STP?
On 16 Mar 2006 15:34:50 -0800, [EMAIL PROTECTED] (Jerry Whitteridge) wrote: >Run the NTP server on the parallel sysplex and have the other platforms >use it as the time source. That's certainly a viable solution, but... In some environments, running the NTP server on the mainframe will do a very good job of keeping all your other systems consistent...consistently wrong. I can't speak to STP, since I know very little about this new feature. But traditionally, synchronizing a mainframe environment to an external time source such as NIST or GPS requires an ETR[1]. I don't know what the current price of that device runs, but I've seen figures as high as $35K bandied about in this list for what is essentially a glorified PC. What's your response to the CIO after you've requested an ETR and he/she replies along the lines of: "You want me to spend $xxx for something that every other system in my shop can do for free???!!!" Eric [1] IIRC, the Sysplex Timer does *not*, by itself, synch to an external time source. It merely keeps the TOD clock in all systems in the plex synchronized. -- Eric Chevalier E-mail: [EMAIL PROTECTED] Web: www.tulsagrammer.com Is that call really worth your child's life? HANG UP AND DRIVE! -- 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: TOD Sync acroos platforms (IBM and non-IBM) - STP?
The original poster signaled that they had a parallel sysplex which to me implies he already has an ETR. I must admit I forgot about those shops running on a single footprint with no ETR. On Fri, 2006-03-17 at 09:10 -0600, Eric Chevalier wrote: > That's certainly a viable solution, but... > > In some environments, running the NTP server on the mainframe will do a > very good job of keeping all your other systems consistent...consistently > wrong. I can't speak to STP, since I know very little about this new > feature. But traditionally, synchronizing a mainframe environment to an > external time source such as NIST or GPS requires an ETR[1]. I don't know > what the current price of that device runs, but I've seen figures as high > as $35K bandied about in this list for what is essentially a glorified PC. > > What's your response to the CIO after you've requested an ETR and he/she > replies along the lines of: "You want me to spend $xxx for something that > every other system in my shop can do for free???!!!" > > Eric > > [1] IIRC, the Sysplex Timer does *not*, by itself, synch to an external > time source. It merely keeps the TOD clock in all systems in the plex > synchronized. > > -- > Eric Chevalier E-mail: [EMAIL PROTECTED] >Web: www.tulsagrammer.com > Is that call really worth your child's life? HANG UP AND DRIVE! > > -- > 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 -- Jerry Whitteridge Safeway Inc. (925) 951 4184 "MMS " made the following annotations. -- Warning: All e-mail sent to this address will be received by the Safeway corporate e-mail system, and is subject to archival and review by someone other than the recipient. This e-mail may contain information proprietary to Safeway and is intended only for the use of the intended recipient(s). If the reader of this message is not the intended recipient(s), you are notified that you have received this message in error and that any review, dissemination, distribution or copying of this message is strictly prohibited. If you have received this message in error, please notify the sender immediately. == -- 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: TOD Sync acroos platforms (IBM and non-IBM) - STP?
In a recent note, Alan C. Field said: > Date: Fri, 17 Mar 2006 08:11:42 -0600 > > Either Keith Clapp's SNTP code or the SNTP server that is part of z/OS. Have > all the > other platforms synchronize to z/OS. > > We have been doing that for about 4 years - seems to be working well once we > overcame the political/idealogical problem that z/OS could actually do this. > And: In a recent note Jerry Whitteridge said: > Date: Thu, 16 Mar 2006 16:34:33 -0700 > > Run the NTP server on the parallel sysplex and have the other platforms > use it as the time source. > This suggestion is effective but arrogant. Not everyone can be the boss. If IBM hopes to increase its market share, IBM must sell into shops that have not previously been using IBM mainframes. And the cultural norm (Alan says "political/idealogical problem") is that the new kid in the neighborhood doesn't instantly get to be the gang leader. But I believe that there's real hope that STP will relieve this problem. At least the latest PoOp acknowledges the existence of NTP. -- gil -- StorageTek INFORMATION made POWERFUL -- 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: TOD Sync acroos platforms (IBM and non-IBM) - STP?
Paul, Given that the poster asked how he could do it, not what would be the best if things were different I think our answers were not arrogant but accurate. If this, If that doesn't fix the problem. If IBM might do something it's still in the future, and were not discussing what IBM should do but solving an existing question with the tools CURRENTLY available. On Fri, 2006-03-17 at 09:59 -0700, Paul Gilmartin wrote: > This suggestion is effective but arrogant. Not everyone can be the boss. > If IBM hopes to increase its market share, IBM must sell into shops that > have not previously been using IBM mainframes. And the cultural norm > (Alan says "political/idealogical problem") is that the new kid in the > neighborhood doesn't instantly get to be the gang leader. > > But I believe that there's real hope that STP will relieve this problem. > At least the latest PoOp acknowledges the existence of NTP. > > -- gil -- Jerry Whitteridge Safeway Inc. (925) 951 4184 "MMS " made the following annotations. -- Warning: All e-mail sent to this address will be received by the Safeway corporate e-mail system, and is subject to archival and review by someone other than the recipient. This e-mail may contain information proprietary to Safeway and is intended only for the use of the intended recipient(s). If the reader of this message is not the intended recipient(s), you are notified that you have received this message in error and that any review, dissemination, distribution or copying of this message is strictly prohibited. If you have received this message in error, please notify the sender immediately. == -- 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: TOD Sync acroos platforms (IBM and non-IBM) - STP?
On 17 Mar 2006 08:54:05 -0800, [EMAIL PROTECTED] (Jerry Whitteridge) wrote: >The original poster signaled that they had a parallel sysplex which to >me implies he already has an ETR. I must admit I forgot about those >shops running on a single footprint with no ETR. I may be mistaken (it happens a couple times a year :-) ), but I was under the impression that a parallel sysplex did *not* require an ETR, only a sysplex timer. ETR != Sysplex Timer Eric -- Eric Chevalier E-mail: [EMAIL PROTECTED] Web: www.tulsagrammer.com Is that call really worth your child's life? HANG UP AND DRIVE! -- 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: TOD Sync acroos platforms (IBM and non-IBM) - STP?
On Fri, 17 Mar 2006 09:10:20 -0600, Eric Chevalier <[EMAIL PROTECTED]> wrote: > >[1] IIRC, the Sysplex Timer does *not*, by itself, synch to an external >time source. It merely keeps the TOD clock in all systems in the plex >synchronized. > The synchronization to an external time source is enabled by something called a CLO card in the timer. We had this same issue. Bought the timer and it kept the machine synched to the timer quite nicely. Except the timer is really rotten at keeping time itself. Once we got the CLO card installed and synching, it works very well. -- 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: TOD Sync acroos platforms (IBM and non-IBM) - STP?
Then I'm confused. ETR = External Time Reference = Sysplex Timer ( If I want to see the status of my sysplex timer I issue the command D ETR) What leads you to write ETR != Sysplex Timer ? (I'm curious -- with the exception of a parallel sysplex running on a single footprint (which I agreed I forgot) I know of no other way to run a parallel sysplex(with or without VM)) I would accept that there are other providers of an ETR but none (that I am aware of) that can connect to zSeries hardware and be useful. On Fri, 2006-03-17 at 11:15 -0600, Eric Chevalier wrote: > On 17 Mar 2006 08:54:05 -0800, > [EMAIL PROTECTED] (Jerry Whitteridge) wrote: > > >The original poster signaled that they had a parallel sysplex which to > >me implies he already has an ETR. I must admit I forgot about those > >shops running on a single footprint with no ETR. > > I may be mistaken (it happens a couple times a year :-) ), but I was under > the impression that a parallel sysplex did *not* require an ETR, only a > sysplex timer. > > ETR != Sysplex Timer > > Eric > > -- > Eric Chevalier E-mail: [EMAIL PROTECTED] >Web: www.tulsagrammer.com > Is that call really worth your child's life? HANG UP AND DRIVE! > > -- > 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 -- Jerry Whitteridge Safeway Inc. (925) 951 4184 "MMS " made the following annotations. -- Warning: All e-mail sent to this address will be received by the Safeway corporate e-mail system, and is subject to archival and review by someone other than the recipient. This e-mail may contain information proprietary to Safeway and is intended only for the use of the intended recipient(s). If the reader of this message is not the intended recipient(s), you are notified that you have received this message in error and that any review, dissemination, distribution or copying of this message is strictly prohibited. If you have received this message in error, please notify the sender immediately. == -- 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: TOD Sync acroos platforms (IBM and non-IBM) - STP?
On 17 Mar 2006 10:22:46 -0800, [EMAIL PROTECTED] (Jerry Whitteridge) wrote: >Then I'm confused. ETR = External Time Reference = Sysplex Timer >( If I want to see the status of my sysplex timer I issue the command D >ETR) >What leads you to write ETR != Sysplex Timer ? (I'm curious -- with the >exception of a parallel sysplex running on a single footprint (which I >agreed I forgot) I know of no other way to run a parallel sysplex(with >or without VM)) I'll trust the experts on this list to correct me if I'm wrong, but here's a diagram that illustrates the relationship between the systems in the plex, the sysplex timer and the ETR (as I understand the architecture): (The diagram is best viewed with a fixed-pitch font.) +--++--++--+ | || || | | zOS A || zOS B || zOS C | | || || | +--++--++--+ | | | | | | +---+---+ | | +-+ | Sysplex | | Timer | +-+ | | +-+ | ETR | +-+ | | +-+ | NIST | | GPS | | etc | +-+ The z/OS systems might be LPARs on the same physical box, or LPARs on completely separate boxes; I believe the architecture is the same in either case. The sysplex timer is the device that makes sure the TOD clocks in all the systems in the plex are in synch. However, that synchronization has no direct correlation with the outside world. If the plex has no ETR, then over time the TOD clocks may drift and no longer accurately match real-world time of day. However, the sysplex timer insures that ALL of the TOD clocks in the plex will be off BY THE SAME AMOUNT. If you then acquire an ETR, it serves as an interface between your external time source and the sysplex timer. Somehow, through magic I don't know anything about, the ETR will feed correct time-of-day information to the sysplex timer. The sysplex timer, in turn, will feed corrections to the systems in the plex so that the TOD clocks are not only in sych, but also match real-world time. As Dave Kopischke mentioned in a previous post, it sounds like current zServer systems can use "something called a CLO card in the timer" to maintain synchronization with an external time source. Sounds like the CLO card performs the functions of the ETR; I'll have to defer to Dave on this, since I don't have any knowledge of CLO cards. Eric -- Eric Chevalier E-mail: [EMAIL PROTECTED] Web: www.tulsagrammer.com Is that call really worth your child's life? HANG UP AND DRIVE! -- 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: TOD Sync acroos platforms (IBM and non-IBM) - STP?
See SG24-2070 S/290 Time Management and IBM 9037 Sysplex Timer 300 plus pages. My interpretation is that the Sysplex Timer IS the ETR. You connect it to an ETS (External Time Source), modem dialing NIST (what we do here), GPS or WWV receiver and that gets the timer in sync with the rest of the world. The timer then causes the clock(s) in the processor to be adjusted to this time. Coming in z/OS 1.7 is a feature called STP (if I remember right) that makes the Sysplex timer obsolete and allows getting time from the network. Alan -- 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: TOD Sync acroos platforms (IBM and non-IBM) - STP?
On Fri, 2006-03-17 at 11:15 -0600, Eric Chevalier wrote: > I may be mistaken (it happens a couple times a year :-) ), but I was under > the impression that a parallel sysplex did *not* require an ETR, only a > sysplex timer. > > ETR != Sysplex Timer Increment the count. ETR *IS* the Sysplex timer. Which (of course) may not be external at all, and may not be *too* external, and is a lousy time keeper. But it *is* used as a common reference - one out of three will have to suffice. To run a Parallel Sysplex you *must* have an ETR (amongst other things) which may be emulated. And (of course) if you wish to use an external time source to synch your ETR you have to plug another bit of hardware in to do the job. IBM obviously finally got sick of customers actually trying to tie their dinos into real networks and appear to have relented with STP. Shane ... -- 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: TOD Sync acroos platforms (IBM and non-IBM) - STP?
In a recent note, Jerry Whitteridge said: > Date: Fri, 17 Mar 2006 09:09:34 -0800 > > Paul, Given that the poster asked how he could do it, not what would be > the best if things were different I think our answers were not arrogant > but accurate. > On re-reading what I wrote, I see that it bears an unintended flavor of personal characterization. My apologies. The arrogance should be attributed to the unilateral presumption by any particular network component that it is entitled to be the sole master of some function. An analogy: When I first read the owners' manual for the VT-100 terminal, I noticed that it had the option of NTSC video output. However, when used in a network of video sources, it was required that the VT-100 supply the time base for all the others (before the era of digital time base correction, most video sources had sync inputs; VT100 had only a sync output). I deem this arrogance on the part of the VT-100 design, but not anyone who finds it acceptable to use the VT-100 as a sync source. (But what if someone wanted _two_ VT-100s in his network?) > If this, If that doesn't fix the problem. If IBM might do something it's > still in the future, and were not discussing what IBM should do but > solving an existing question with the tools CURRENTLY available. > Charter notwithstanding, it's typical of threads in this forum, on Friday, to drift to what might be, what may soon be (such as STP), and especially, what used to be. > On Fri, 2006-03-17 at 09:59 -0700, Paul Gilmartin wrote: > > > This suggestion is effective but arrogant. Not everyone can be the boss. -- gil -- StorageTek INFORMATION made POWERFUL -- 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: TOD Sync acroos platforms (IBM and non-IBM) - STP?
On 17 Mar 2006 12:49:15 -0800, [EMAIL PROTECTED] (Shane) wrote: >And (of course) if you wish to use an external time source to synch your >ETR you have to plug another bit of hardware in to do the job. OK, I stand corrected, at least on my terminology, and possibly on the cost of that other "bit of hardware" required to keep the plex in synch with real-world time. I believe that the z800 where I work does not synch to some external time source; issuing: ro *all,d t shows a local time from all systems that's about 15-20 seconds slow compared to my Casio Waverunner (wristwatch that synchs to NIST). If that other "bit of hardware" were only a few hundred dollars, I'm pretty sure we'd be taking advantage of that feature. Eric -- Eric Chevalier E-mail: [EMAIL PROTECTED] Web: www.tulsagrammer.com Is that call really worth your child's life? HANG UP AND DRIVE! -- 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: TOD Sync acroos platforms (IBM and non-IBM) - STP?
On 17 Mar 2006 09:34:16 -0800, in bit.listserv.ibm-main I wrote: >I may be mistaken (it happens a couple times a year :-) ), but I was under >the impression that a parallel sysplex did *not* require an ETR, only a >sysplex timer. > >ETR != Sysplex Timer Based on feedback from Alan Field, Shane and Giovanni Cerquone, I admit being egregiously mistaken in this statement. (This group has a great, and very tolerant, collection of teachers!) I think the following diagram more accurately summarizes the architecture I was writing about in earlier messages: +--++--++--+ | || || | | zOS A || zOS B || zOS C | | || || | +--++--++--+ | | | | | | +---+---+ | | +-+ | Sysplex | | Timer | | (ETR) | +-+ | | +-+ | External| | Time| | Source | | Interface | | (CLO Card?) | +-+ | | +-+ | NIST | | GPS | | etc | +-+ I hope this version is a more accurate representation of reality! Relating back to the original suggestions of running an NTP server on the mainframe: that solution will keep all of your shop's systems pretty much in synch, but without that "External Time Source Interface", your systems still may not reflect real-world time. -- Eric Chevalier E-mail: [EMAIL PROTECTED] Web: www.tulsagrammer.com Is that call really worth your child's life? HANG UP AND DRIVE! -- 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: TOD Sync acroos platforms (IBM and non-IBM) - STP?
Eric Chevalier wrote: OK, I stand corrected, at least on my terminology, and possibly on the cost of that other "bit of hardware" required to keep the plex in synch with real-world time. Depends on whether your "plex" is just LPARs on the same box or different boxes connected together. A sysplex timer is required to sync a z800 to other physical machines in a sysplex. It's not inexpensive. If your plex is just LPARs on a single z800, then you don't need a sysplex timer at all. All of your z/OS systems will specify SIMETRID in CLOCKxx. A z800 cannot synchronize to an external time source by itself. It depends on the sysplex timer to do that. Only a shop connecting different boxes together would invest in a sysplex timer. With just a single footprint, a sysplex timer would be gratuitous. STP will allow different physical z990 and System z9 processors to synchronize with each other, thus eliminating the need for a sysplex timer. With the sysplex timers gone, the processors will need to be able to natively synchronize the plex with an external time source. I believe that the z800 where I work does not synch to some external time source; issuing: ro *all,d t shows a local time from all systems that's about 15-20 seconds slow compared to my Casio Waverunner (wristwatch that synchs to NIST). If that other "bit of hardware" were only a few hundred dollars, I'm pretty sure we'd be taking advantage of that feature. Just a single footprint z800. No ETR. No ETS. -- 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: TOD Sync acroos platforms (IBM and non-IBM) - STP?
In a recent note, Edward E. Jaffe said: > Date: Fri, 17 Mar 2006 14:29:16 -0800 > > Eric Chevalier wrote: > > OK, I stand corrected, at least on my terminology, and possibly on the > > cost of that other "bit of hardware" required to keep the plex in synch > > with real-world time. > > A z800 cannot synchronize to an external time source by itself. It > depends on the sysplex timer to do that. Only a shop connecting > different boxes together would invest in a sysplex timer. With just a > single footprint, a sysplex timer would be gratuitous. > That's an economic decision. If keeping your single footprint in synch with real-world time without periodic IPLs is worth $30K, then you need a sysplex timer. --gil -- StorageTek INFORMATION made POWERFUL -- 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: TOD Sync acroos platforms (IBM and non-IBM) - STP?
FWIW. We got our 3 9037-1 approx 3 years ago for USD1000 a piece. Two online, one in case the other breaks. None have broken. They are set to synchronize with Boulder NIST once a day. We've had zero problems and zero issues. Consequentially, the NTPD daemon on z/OS always serves up the correct time to all who ask. -- Jeff Eric Chevalier said the following on 03/17/2006 10:10 AM: On 16 Mar 2006 15:34:50 -0800, [EMAIL PROTECTED] (Jerry Whitteridge) wrote: Run the NTP server on the parallel sysplex and have the other platforms use it as the time source. That's certainly a viable solution, but... In some environments, running the NTP server on the mainframe will do a very good job of keeping all your other systems consistent...consistently wrong. I can't speak to STP, since I know very little about this new feature. But traditionally, synchronizing a mainframe environment to an external time source such as NIST or GPS requires an ETR[1]. I don't know what the current price of that device runs, but I've seen figures as high as $35K bandied about in this list for what is essentially a glorified PC. What's your response to the CIO after you've requested an ETR and he/she replies along the lines of: "You want me to spend $xxx for something that every other system in my shop can do for free???!!!" Eric [1] IIRC, the Sysplex Timer does *not*, by itself, synch to an external time source. It merely keeps the TOD clock in all systems in the plex synchronized. -- Eric Chevalier E-mail: [EMAIL PROTECTED] Web: www.tulsagrammer.com Is that call really worth your child's life? HANG UP AND DRIVE! -- 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: TOD Sync acroos platforms (IBM and non-IBM) - STP?
Eric, Almost I think. The Sysplex Timer is a big clunky box. It connects to a PC as it's "console" We have a modem plugged into the PC that dials out to NIST. I think the CLO (CLOck?) card is the card in the processor that connects to the Sysplex timer. On 17 Mar 2006 09:34:16 -0800, in bit.listserv.ibm-main I wrote: >I may be mistaken (it happens a couple times a year :-) ), but I was under >the impression that a parallel sysplex did *not* require an ETR, only a >sysplex timer. > >ETR != Sysplex Timer Based on feedback from Alan Field, Shane and Giovanni Cerquone, I admit being egregiously mistaken in this statement. (This group has a great, and very tolerant, collection of teachers!) I think the following diagram more accurately summarizes the architecture I was writing about in earlier messages: +--++--++--+ | || || | | zOS A || zOS B || zOS C | | || || | +--++--++--+ | | | | | | +---+---+ | | +-+ | Sysplex | | Timer | | (ETR) | +-+ | | +-+ | External| | Time| | Source | | Interface | | (CLO Card?) | +-+ | | +-+ | NIST | | GPS | | etc | +-+ I hope this version is a more accurate representation of reality! Relating back to the original suggestions of running an NTP server on the mainframe: that solution will keep all of your shop's systems pretty much in synch, but without that "External Time Source Interface", your systems still may not reflect real-world time. -- 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: TOD Sync acroos platforms (IBM and non-IBM) - STP?
On Sat, 18 Mar 2006 23:13:33 -0600, Alan C. Field <[EMAIL PROTECTED]> wrote: >Eric, > >Almost I think. > >The Sysplex Timer is a big clunky box. It connects to a PC as it's >"console" > >We have a modem plugged into the PC that dials out to NIST. > >I think the CLO (CLOck?) card is the card in the processor that connects >to >the Sysplex timer. > > The CLO card is physically a card that installs in the timer. As I understand it, it enables the timer to receive time information from an outside source. I had to go back and dig out the old manuals I used a couple years ago to research this. >From "S/390 Time Management and IBM 9037 Sysplex Timer" SG24-2070-00 CLO = Control Link Oscillator "In the Expanded Basic configuration, a Control Link Oscillator (CLO) card is installed in the 9037. Therefore, this configuration has two advantages over the Basic configuration, described in 2.1.3, Basic Configuration on page 13. With this card installed, the 9037 can track to an External Time Source (ETS). Refer to 2.4, External Time Source Attachment on page 16 for more information on attaching a 9037 to an external time source. Secondly, the Expanded Basic configuration can be enhanced to the fully redundant Expanded Availability configuration concurrently, therefore not requiring any outage of operations." And my understanding of the function of the timer is to steer the onboard clocks on the processor. The processor clocks are always the source of time information. The timer compares the time from the external source to the time on the processor and either slows the clock down or speeds it up until they match. If you don't have a CLO card, the timer itself is the time source (your wristwatch is probably more accurate). You can't get more than four seconds out of synch or the timer won't be able to generate the proper steering impulses to get the clocks back in synch. There's lots of detail in the manuals if you care to attempt to understand 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: TOD Sync acroos platforms (IBM and non-IBM) - STP?
Dave, If by "processor" you mean the sZeries CPC, then your understanding might be a little off... My understanding is that the timer just sends timing signals to the CPC, it is unaware of what the CPC is doing. Basically it is a one way process. Whether there is a CLO or not, the 9037 never gets the time from the CPCs. It just sends the time and stepping signals to control the CPC clock rate. (it also sends offsets for daylight saving/Leap seconds etc). The only thing the 9037 is aware of is that there is a CPC connected at the other end of the port. So long as there is a link established, then it doesn't know what the CPC is doing with the information. If your CPC is connected to 2 9037 ports, then both 9037 ports send data unaware as to which port is being used by the CPC. The only thing the 9037 knows about the CPC is how far away it is. It determines this so that it can "tune" the signal on each port to take account of time delay etc.. If there isn't a CLO card then you're right in that the clock on the 9037 processor board is used as the time source. However, if you have a CLO card (which is required when you hook up 2 timers), then I believe that internally to the 9037 it is the CLO card(s) that are used to maintain the clock and its steering rate. These use some sort of voltage controlled oscillator to maintain syncronisation with two timer units and to enable the clock rate to steer to a corrected time. Whether the corrected time is a manual input, or from an ETS, I don't think makes much difference to the CLO - it just steers the clock to the new time. Regards Roy Dave Kopischke wrote: And my understanding of the function of the timer is to steer the onboard clocks on the processor. The processor clocks are always the source of time information. The timer compares the time from the external source to the time on the processor and either slows the clock down or speeds it up until they match. If you don't have a CLO card, the timer itself is the time source (your wristwatch is probably more accurate). You can't get more than four seconds out of synch or the timer won't be able to generate the proper steering impulses to get the clocks back in synch. There's lots of detail in the manuals if you care to attempt to understand 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 -- 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: TOD Sync acroos platforms (IBM and non-IBM) - STP?
In a recent note, Roy Hewitt said: > Date: Thu, 23 Mar 2006 12:42:46 + > > My understanding is that the timer just sends timing signals to the CPC, > it is unaware of what the CPC is doing. Basically it is a one way > process. Whether there is a CLO or not, the 9037 never gets the time > from the CPCs. It just sends the time and stepping signals to control > the CPC clock rate. (it also sends offsets for daylight saving/Leap > seconds etc). > What initially sets the content of the TOD clock at IPL (or is it POR) time. IIRC, there is some setting that can only be accomplished at POR. Is there a clock in the HMC, so the data path for that initial setting might be: ETR to HMC (at POR) HMC to TOD (at IPL)? > Dave Kopischke wrote: > > > > And my understanding of the function of the timer is to steer the onboard > > clocks on the processor. The processor clocks are always the source of time > > information. The timer compares the time from the external source to the > > time on the processor and either slows the clock down or speeds it up until > > they match. If you don't have a CLO card, the timer itself is the time > > source (your wristwatch is probably more accurate). You can't get more than > > four seconds out of synch or the timer won't be able to generate the proper > > steering impulses to get the clocks back in synch. There's lots of detail > > in the manuals if you care to attempt to understand it. > > I recall we once got about 15 seconds out of sync, and were able to recover without the dreaded POR by repeatedly issuing manual commands to steer by a couple seconds until after about a week the ETR was within tolerance to sync to NIST dialup. I believe an IBM employee (Greg Dyck?) has supplied instructions for doing this on this list. -- gil -- StorageTek INFORMATION made POWERFUL -- 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: TOD Sync acroos platforms (IBM and non-IBM) - STP?
Paul Gilmartin wrote: What initially sets the content of the TOD clock at IPL (or is it POR) time. IIRC, there is some setting that can only be accomplished at POR. Is there a clock in the HMC, so the data path for that initial setting might be: ETR to HMC (at POR) HMC to TOD (at IPL)? The HMC is not used in any of the time setting process. Remember the HMC is purely optional, its only an operator aid, you can POR/IPL without it by using the Support Element (SE) (the Thinkpad inside the CPC). There is an option to sync the HMC time with the time on the SE As for the POR time, I suppose it depends on the CPC model in use. But with the latest zZeries it gets the TOD time from the 9037 - if you have one. If not, it gets the initial time from the Support Element at POR. That being said if you POR without a 9037, then connect one afterwards, it just gets on with life and uses the 9037 value. This I know as I recently did a lot of work with new CPCs and timers. The timer was connected after the POR and all was ok. Just to add to the loop the SE will sync to the CPC once a day With IPL time, each LPAR has its own virtual TOD clock that is initially set when the partition is activated. The initial value is usually that of the CPC TOD. If there is a 9037 and the Opsys supports it then it will use the 9037 time and can accept offset changes etc. The LPAR profile has options to enable the use of the 9037 and also to specify any offset value I recall we once got about 15 seconds out of sync, and were able to recover without the dreaded POR by repeatedly issuing manual commands to steer by a couple seconds until after about a week the ETR was within tolerance to sync to NIST dialup. I believe an IBM employee (Greg Dyck?) has supplied instructions for doing this on this list. Setting a new 9037 time to drift to can be a real pain due to the 4.999 seconds maximum value at a rate 1 sec per 12 hours - i had to adjust 45 secs recently.I never understood why this maximum value was so low. Then recently I saw some possible explanation for the low value. I was testing the 9037 ETS a few weeks back and for some reason, not sure why, the time that was interpreted from the dialup modem was way off, about 21 hours in advance. As this was over the maximum if ignored the value and disabled the ETS. I suppose a sensible approach. What would have been better would to have the 4.999 secs for ETS, but a manual offset greater than this could prompt for "are you sure". Regards Roy -- 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