On Mon, 2003-06-02 at 15:35, Richard Troth wrote: > I note that the zSeries clock is (800, 900, even 9672s) > is awfully accurate compared to even the newest PCs. > (Have not compared it to Sun or HP mid-range hardware.) > I'm talking about watching the drift file NTPD maintains. > It goes below 1.000 on the mainframe. Not seen that elsewhere. > > I mention this because you *might* get some relief > if you synch the time on your "support element" and then POR. > Linux guests can run NTPD and stay locked in synch with the network. > Other zSeries hosts will be off, but by a surprisingly small amount. > > Would that be enough for your Java guys?
I would hope so but that's sometimes hard to tell. They are developing an application that uses a payment service that is hosted outside of the agency. This service is running on midrange gear and syncs time through ntp. The reports are that the time on the payment servers are very close to our network time and the mainframe is +90 seconds off. Interestingly enough the support element is in sync with the mainframe and at the last POR (roughly 3 weeks ago) the difference was only about 20 seconds. Being the mainframe novice of the group it seems to me that the clock in the 2066 is tracking the support element clock (or vice versa). Am I correct to say that the mainframe time is set from the support element ONLY at POR, or are there periodic updates between the support element and the mainframe clock?
