On Fri, Jan 03, 2003 at 10:51:51AM -0500, David Boyes wrote: > I haven't looked at the code in a few weeks (the last time I tried to hunt > this down), but I was referring to the fact that with xntp, the > incremental adjustments of the timer cause the guest to wake up fairly > frequently for checks on timer drift, which effectively negates a lot of > the "go to sleep until I have something real to do" goals of the timer > patch -- you've just given it something legitmate to do on a fairly > periodic basis.
ntpd normally wakes up once per second to do its bookkeeping (EVENT_TIMEOUT=0). This can of course be adjusted (within reasonable limits) by building with a different EVENT_TIMEOUT if one is careful. Though, still, waking up once per second should be much less of an impact than the 100Hz timer interrupt. Usually more annoying is the statistics collection that many distributions enable by default, which causes actual disk I/O to be done rather frequently (/var/log/ntpstats or similar). -- - mdz
