-----Original Message----- From: IBM Mainframe Discussion List [mailto:[EMAIL PROTECTED] On Behalf Of R.S. Sent: Saturday, May 19, 2007 8:22 AM To: [email protected] Subject: Re: Synchronize Time Between Mainframe and Servers?
As usually in mainframe world, we have a lot of proofs that mainframe is better, we can provide citations from "ESA/390 Holy Principles", we can answer that we have ntp server... But the only honest answer to the question is "NO, unfortunately we are not able to synchronize with atomic clock, like other devices in our server room. Mainframe *can't do it*. We can only use wrist watch as a time source and then rely on mainframe internal clock." Of course, after the confession we can go back to our Principles, talk about our availability, hint that the rest of the world should synchronize with our CPC... -- Radoslaw Skorupka Lodz, Poland P.S. I would like to hear that I'm wrong. Preferrably with details - how to synchronize the mainframe with ntp server or atomic clock. Preferrably without 9037 Timers, which are expensive and EOM (End Of Market). <snip> First, I was answering the original question as the poster asked it, with the understanding that they did NOT have an ETR (v1 or v2). As late as 2003 I had a problem with a client who migrated off VSE to z/OS, and then tried to migrate off the mainframe to other platforms. The need for time sync was demonstrable (DR purposes was secondary!), and we could not seem to get management of the company to recognize the issues (they bought the idea that NTP was sufficient). The technical people handling the non-mainframe platforms did understand, and the DBAs (non-mainframe) recognized that time sync using the mainframe would be much more reliable than what they were trying to do (even with that installation using an operator's view of a clock pending the strike of the <enter> key). So sometimes what you have is a purely political issue. Second, you most certainly can use an external time source to set the mainframe (either via Unix System Services & TCP/IP, or through the HMC -- neither of which I have personal experience doing). The question is, since the time source you will attempt to sync to (absent an ETR connected to a "GPS/UTC" time source) is NOT sufficiently accurate to avoid ambiguities in TOD processing -- why would you desire to use NTP to set the mainframe, as opposed to letting the mainframe be your NTP source? This was my reason for pointing out, from the PoOP, that the TOD cycles bit 51 at the micro-second rate, giving a very small drift (baring a bad power source as someone else pointed out -- but that kind of issue should cause the system to crash because of, shall we call them timer-checks?) compared to Intel Architecture based clocks, which seem to suffer significant drift problems (I have personal experience in this case using Tardis to keep a LAN synchronized and looking at the drift report on the NT server was SCARY - 2-10 seconds in a day, and other LANs with other software/hardware with similar issues). The current thing coming from IBM (which I greatly appreciate and can't wait until we take delivery on the new hardware to use it) is STP. But not having ANY experience with it yet, I can't truly speak to it. Regards, Steve Thompson -- STD disclaimer: poster's posting may not reflect poster's employer's opinions and should not necessarily be construed as such. Deputy Dawg did not approve of this message. YMMV -- ---------------------------------------------------------------------- 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
