Messages in this thread | | | Date | Fri, 4 Dec 1998 08:41:31 -0500 (EST) | From | "Richard B. Johnson" <> | Subject | Re: Linux timekeeping plans |
| |
On Thu, 3 Dec 1998, Brandon S. Allbery KF8NH wrote: > > PC RTCs are horrible. For all that it lost up to 15 seconds a day, my > 486DX2/66 kept better time in Linux (or OS/2, etc.) than its RTC did. My > P200 dosesn't appear to be much better. >
Here is a log of my Linux file-server's time referenced to NIST at one-hour intervals.
Script started on Thu Dec 3 11:30:48 1998 # while gettime time-a.timefreq.bldrdoc.gov ; do sleep 3600 ; done
Reference time = Thu Dec 3 11:31:37 1998 My time = Thu Dec 3 11:31:37 1998 Difference = 0 Reference time = Thu Dec 3 12:31:38 1998 My time = Thu Dec 3 12:31:38 1998 Difference = 0 Reference time = Thu Dec 3 13:31:38 1998 My time = Thu Dec 3 13:31:38 1998 Difference = 0 Reference time = Thu Dec 3 14:31:38 1998 My time = Thu Dec 3 14:31:38 1998 Difference = 0 Reference time = Thu Dec 3 15:31:38 1998 My time = Thu Dec 3 15:31:38 1998 Difference = 0 Reference time = Thu Dec 3 16:31:37 1998 My time = Thu Dec 3 16:31:37 1998 Difference = 0 Reference time = Thu Dec 3 17:31:37 1998 My time = Thu Dec 3 17:31:37 1998 Difference = 0 Reference time = Thu Dec 3 18:31:37 1998 My time = Thu Dec 3 18:31:37 1998 Difference = 0 Reference time = Thu Dec 3 19:31:37 1998 My time = Thu Dec 3 19:31:37 1998 Difference = 0 Reference time = Thu Dec 3 20:31:36 1998 My time = Thu Dec 3 20:31:33 1998 Difference = -3 seconds Reference time = Thu Dec 3 21:31:36 1998 My time = Thu Dec 3 21:31:33 1998 Difference = -3 seconds Reference time = Thu Dec 3 22:31:35 1998 My time = Thu Dec 3 22:31:34 1998 Difference = -1 seconds gettime connect(): Connection timed out # exit exit
Script done on Fri Dec 4 08:24:22 1998
You can see that the time was quite stable until 20:31:36 where it took a jump of a few seconds, then by 22:31:35 it had gotten close to the original.
I find that most all PCs are quite stable, but do suffer a long-term drift, mostly because of the unidirectional jog shown above. If the time would slowly vary about a mean value, the adjtime mechanism would on, the average, keep the time quite close to NIST.
Instead, I fix by resetting the time to NIST once each morning using a crond script.
Cheers, Dick Johnson ***** FILE SYSTEM WAS MODIFIED ***** Penguin : Linux version 2.1.128 on an i686 machine (400.59 BogoMips). Warning : It's hard to remain at the trailing edge of technology.
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |