Zero - time drift
Zero - time drift in your OS
Time and time drift is an interesting concept. Beside the "time is relative" ideas derived from Einsteinian relativity, we must also realise that "timekeeping" requires methods of communication (e.g.. wireless, wired internet) that have inherent latency and other delays. Interestingly, when you begin to think about these issues, its a wonder that "the time" is regarded with such reverence. Who decides "the" time? You might feel a little uncomfortable when you learn that "the time" is an average agreement between a number of atomic clocks around the world and that if one (or more) drifts too much, it is taken out of the voting pool until it can be tweaked to track the others in the pool before it can be brought back to "accurate". Well, no need to worry, our lives are normally lived one minute to the next, not one femto-second to the next.
But what about the Internet? At this accuracy, clocks can tell the difference in day length from one day to the next (it's not 24 hours!), and the differences in the period of rotation about the Sun (it's not 365 1/4 days). There are wobbles in both.
One of the most forgotten aspects of computer maintenance (because it is usually part of a system update) is time co-ordination. Computers communicating all over the world need to maintain time to within milliseconds or better over a long period of timekeeping to ensure that their data is in a time consistent state. I am sure that a few issues with operating systems derive from machines not keeping (up with) time.
The CPU clock is not usually part of the timekeeping chain because timekeeping requires battery backup of an independent oscillator. All modern operating systems maintain two major sources of time: the local system time and a process whereby system time can be compensated for drift (mainly due to regular and random temperature variations) by reference to highly accurate atomic standard clocks. So how can you ensure your computer takes full advantage of these super time keepers?
A simple way is to point your computer's time keeping program accessed through the Date and Time control panel) to one of:
- time-a.nist.gov -- possibly most accurate
- time-b.nist.gov -- possibly most accurate
- time-a.timefreq.bldrdoc.gov
- time-b.timefreq.bldrdoc.gov
- time-c.timefreq.bldrdoc.gov
- utcnist.colorado.edu time.nist.gov
- time-nw.nist.gov
- nist1.datum.com -- preferred for access from Australia, California
- nist1.dc.glassey.com
- nist1.ny.glassey.com
- nist1.sj.glassey.com
- nist1.aol-ca.truetime.com -- preferred for access from Australia, California
- nist1.aol-va.truetime.com
Most server software: Mac OS X included can act as timer servers for a LAN, and this is probably better than having each LAN client directly access a Network Time Protocol Server. By using a single local server access a NTP server and the LAN clients access the local time server they would acquire a more precise time. The relative latencies and delays are much less.
if you are more interested in this subject then look at
