On Sat, 19 May 2007 08:40:14 -0500, Eric Bielefeld wrote: >I have a question that I've never seen answered. With all the >discussions of sysplex timers and now SNTP lately, I don't understand >one thing about SNTP. When you go out on the internet to a trusted >time source, how do you account for the time between the time source >getting the request, and the correct time coming back to your machine? >I know there is latency there. Depending on traffic, the time over >the network can vary. How is that accounted for in setting the exact >time? > NIST dialup (used to?) automate this. If the caller echoed every received character, NIST would advance its time signal by half the measured delay. NIST was aware of the delay ranges for any even number of satellite hops. If the delay appeared to correspond to an odd nuumber of satellite hops (although they stated they consdered sucn an unsymmetrical path unlikely), they'd report an error.
NIST also stated, at least previously, that the most accurate setting was available at the lowest rate, 300 BPS, because such archaic modems performed little or no buffering or DSP. NIST also documented that the center of a certain bit (I forget which) in the asterisk character marking the time was the exact time. Obviously this is better than TCP/IP. Otherwise, there's WWV. There's GPS. I believe Sysplex Timer will use WWV. I don't think GPS. How many readers of this list believe their z/Series TOD clock is closer to correct than their PC's? How many less close? Is the z/Series clock as good or better only because the PC uses it as a reference? -- gil ---------------------------------------------------------------------- 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
