Date: Wed, 18 Feb 2009 12:16:44 +0100 From: "Alexey Shuvaev" <shuvaev@physik.uni-wuerzburg.de> To: Gary Jennejohn <gary.jennejohn@freenet.de> Cc: FreeBSD Current <freebsd-current@FreeBSD.org> Subject: Re: HEADS UP: More CAM fixes. Message-ID: <20090218111644.GB43516@wep4035.physik.uni-wuerzburg.de> In-Reply-To: <20090218091151.4d9c2bd7@ernst.jennejohn.org> References: <499981AF.9030204@samsco.org> <20090217164203.4c586f48@ernst.jennejohn.org> <20090218073542.E5200@delplex.bde.org> <499B221C.2050804@samsco.org> <20090218091151.4d9c2bd7@ernst.jennejohn.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, Feb 18, 2009 at 09:11:51AM +0100, Gary Jennejohn wrote:
> On Tue, 17 Feb 2009 13:46:20 -0700
> Scott Long <scottl@samsco.org> wrote:
>
> > Bruce Evans wrote:
> > > On Tue, 17 Feb 2009, Gary Jennejohn wrote:
> > >
> > >> I tested this with an Adaptec 29160. I saw no real improvement in
> > >> performance, but also no regressions.
> > >>
> > >> I suspect that the old disk I had attached just didn't have enough
> > >> performance reserves to show an improvement.
> > >>
> > >> My test scenario was buildworld. Since /usr/src and /usr/obj were both
> > >> on the one disk it got a pretty good workout.
> > > ^^^^ low
> > >>
> > >> AMD64 X2 (2.5 GHz) with 4GB of RAM.
> > >
> > > Buildworld hardly uses the disk at all. It reads and writes a few hundred
> > > MB. Ideally the i/o should go at disk speeds of 50-200MB/S and thus take
> > > between 20 and 5 seconds. In practice, it will take a few more seconds.
> > > physically but perhaps even less virtually due to parallelism.
> > >
> > > Bruce
> >
> > Yes, on modern machines, buildworld is bound almost completely by disk
> > latency, and not at all by disk or controller bandwidth.
> >
> > Scott
> >
>
> Maybe I misunderstood something, but I thought the patch was supposed
> to improve queuing. Seems like all the seeks during a buildowrld
> would exercise that. All I can say is that the disk did _lots_ of
> seeking.
>
I suppose on not extreme multi-core systems (ncores <= 4) buildworld
is a good CPU test. For example:
time make -j3 TARGET_ARCH=i386 buildworld
[building]
--------------------------------------------------------------
>>> World build completed on Wed Feb 18 11:59:28 CET 2009
--------------------------------------------------------------
2163.449u 513.894s 24:17.49 183.6% 6113+2368k 26280+5968io 10602pf+0w
^^^^^^
This is on Core2Duo 3.0GHz with the single 500G WD SATA drive.
Pre-caching src directory (with 'tar -cvf /dev/null src', for example)
may speed up things by ~10%.
Alexey.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20090218111644.GB43516>