Date: Thu, 27 Oct 2005 09:23:27 +0100 (BST) From: Robert Watson <rwatson@FreeBSD.org> To: Poul-Henning Kamp <phk@phk.freebsd.dk> Cc: David Xu <davidxu@freebsd.org>, Marian Hettwer <MH@kernel32.de>, current@freebsd.org Subject: Re: MySQL Performance 6.0rc1 Message-ID: <20051027092153.D31152@fledge.watson.org> In-Reply-To: <21137.1130401220@critter.freebsd.dk> References: <21137.1130401220@critter.freebsd.dk>
next in thread | previous in thread | raw e-mail | index | archive | help
On Thu, 27 Oct 2005, Poul-Henning Kamp wrote: > In message <43607DD5.3020708@freebsd.org>, David Xu writes: > >> Check gettimeofday syscall, it follows every I/O syscall, I think >> our gettimeofday is tooooooo expensive, if we can directly get time from >> memory, the performance will be improved further. > > Why would anybody take a timestamp at all I/O syscalls ? > > "I wonder why my car can only go 30 km/h with the trunk full of > concrete" ? > > In a data base application I could possibly understand a timestamp after > every write. > > But after _all_ I/O syscalls ? That's just plain stupid... This is actually a problem that came up with netperf as well -- it turned out we built it by default with -DHISTOGRAM, which meant that each syscall was followed by a time check. However, the issue of expensive time keeping will presumably keep coming up. It would be interesting to know which time counter is in use on the system is question. Robert N M Watson
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20051027092153.D31152>