On Oct 20, 2011, at 12:42, Miguel Gonçalves wrote: > Please don't get me wrong as this is just my humble opinion but > perhaps 100 ms whas a good target when the project started. Why don't > we get this down to 25 ms? Since today everyone is requesting time > from servers close by I believe 25 ms (or even 10) would be a nice > target.
Most servers are much more accurate than 100ms (including the one you pointed out had ~38ms inaccuracy -- according to the logs that was a temporary condition for an hour or two). Just targeting 100ms is to make sure network latency to the monitoring server never causes trouble and to make sure we don't exclude servers that are slightly inaccurate but good enough for sntp clients. Proper ntp daemons will sync with the two or three other servers instead if one of them is relatively "bad". - ask _______________________________________________ pool mailing list [email protected] http://lists.ntp.org/listinfo/pool
