Date: Thu, 29 Mar 2007 10:58:22 -0500 From: Eric Anderson <anderson@freebsd.org> To: Ivan Voras <ivoras@fer.hr> Cc: freebsd-geom@freebsd.org Subject: Re: raid3 is slow Message-ID: <460BE21E.7070700@freebsd.org> In-Reply-To: <eugncf$93l$1@sea.gmane.org> References: <003401c7712a$f71ebb60$6502a8c0@peteruj> <eudlg8$pm4$1@sea.gmane.org> <005c01c77134$28e0fce0$6502a8c0@peteruj> <86zm5xph7o.fsf@dwp.des.no> <005301c771e4$bb0a3900$6502a8c0@peteruj> <86lkhg5oz5.fsf@dwp.des.no> <007c01c771fe$805b2fc0$6502a8c0@peteruj> <86odmc42mh.fsf@dwp.des.no> <eugncf$93l$1@sea.gmane.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On 03/29/07 10:52, Ivan Voras wrote: > Dag-Erling Smørgrav wrote: > >> You don't seem to understand what the load averages mean. They are >> the average number of runnable threads in the scheduler queue over the >> last one, five and fifteen seconds. Certain workloads will drive up >> the load averages without consuming all available CPU time. This is >> particularly the case for workloads where small chunks of data (e.g. >> RAID3 stripes) are passed around between multiple threads. > > But, in his case the threads DO seem to consume much more CPU time than > they should - especially the g_down thread. In this case, load avg ( as > an approximation of real system load) is useful, and he's not concerned > without cause. > > It would be good to see output of a ps -auxl next time he's running the test. Eric
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?460BE21E.7070700>