Date: Wed, 20 Oct 2004 11:23:40 +0800 From: Xin LI <delphij@frontfree.net> To: Scott Long <scottl@freebsd.org> Cc: freebsd-current@freebsd.org Subject: Re: FreeBSD 5.3b7and poor ata performance Message-ID: <20041020032340.GC3797@frontfree.net> In-Reply-To: <4175D269.208@freebsd.org> References: <4173D66F.6010200@DeepCore.dk> <4173F2E9.7010407@ng.fadesa.es> <417406E3.9010706@DeepCore.dk> <4174FD04.8040000@ng.fadesa.es> <20041019104525.ikgw8kcw8sw480os@www.sweetdreamsracing.biz> <4129.192.168.0.188.1098211592.squirrel@192.168.0.188> <20041019200908.GA655@frontfree.net> <4259.192.168.0.188.1098217092.squirrel@192.168.0.188> <20041020024908.GA3797@frontfree.net> <4175D269.208@freebsd.org>
index | next in thread | previous in thread | raw e-mail
[-- Attachment #1 --] Hi, Scott, On Tue, Oct 19, 2004 at 08:50:17PM -0600, Scott Long wrote: > >You may want to try other block sizes, like 65536, 262144, 524288, 1048576 > >or so. The peak performance block size depends heavily on hardware... > > > >Cheers, > > This won't really matter. physio will chop the blocks up into 128k > segments, and GEOM will cut them again into 64k segments. Other than > a minor amount of coelscing in these stages, it won't make a difference. Oh... Maybe the last experiment was done on a RELENG_4 box :-( Here is my result on: mpt0: <LSILogic 1030 Ultra4 Adapter> da0: <FUJITSU MAS3735NC 5B08> Fixed Direct Access SCSI-3 device da0: 320.000MB/s transfers (160.000MHz, offset 127, 16bit), Tagged Queueing Enabled da0: 70007MB (143374650 512 byte sectors: 255H 63S/T 8924C) FreeBSD 5.3-BETA7 (compiled three days before) cvs# dd if=/dev/da0 of=/dev/null count=2000 bs=262144 2000+0 records in 2000+0 records out 524288000 bytes transferred in 6.654783 secs (78783636 bytes/sec) cvs# dd if=/dev/da0 of=/dev/null count=4000 bs=131072 4000+0 records in 4000+0 records out 524288000 bytes transferred in 6.646773 secs (78878579 bytes/sec) cvs# dd if=/dev/da0 of=/dev/null count=8000 bs=65536 8000+0 records in 8000+0 records out 524288000 bytes transferred in 6.643376 secs (78918913 bytes/sec) cvs# dd if=/dev/da0 of=/dev/null count=16000 bs=32768 16000+0 records in 16000+0 records out 524288000 bytes transferred in 6.639478 secs (78965244 bytes/sec) cvs# dd if=/dev/da0 of=/dev/null count=32000 bs=16384 32000+0 records in 32000+0 records out 524288000 bytes transferred in 6.637433 secs (78989573 bytes/sec) cvs# dd if=/dev/da0 of=/dev/null count=64000 bs=8192 64000+0 records in 64000+0 records out 524288000 bytes transferred in 11.262354 secs (46552257 bytes/sec) cvs# dd if=/dev/da0 of=/dev/null count=128000 bs=4096 128000+0 records in 128000+0 records out 524288000 bytes transferred in 20.617009 secs (25429877 bytes/sec) cvs# dd if=/dev/da0 of=/dev/null count=256000 bs=2048 256000+0 records in 256000+0 records out 524288000 bytes transferred in 39.554797 secs (13254726 bytes/sec) cvs# dd if=/dev/da0 of=/dev/null count=512000 bs=1024 512000+0 records in 512000+0 records out 524288000 bytes transferred in 77.219607 secs (6789571 bytes/sec) cvs# dd if=/dev/da0 of=/dev/null count=1024000 bs=512 1024000+0 records in 1024000+0 records out 524288000 bytes transferred in 154.969467 secs (3383170 bytes/sec) However, with a systat -vm, it seems that the first case do less transactions, so I guess I/O tps on da0 is the transaction with GEOM? Cheers, -- Xin LI <delphij frontfree net> http://www.delphij.net/ See complete headers for GPG key and other information. [-- Attachment #2 --] -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.6 (FreeBSD) iD8DBQFBddo8/cVsHxFZiIoRAnSbAJ9vEWrXoizlJhE6rJNmGAd5MpEZGgCcCeSh A+cta2mChYjerV6RwhRY6Bg= =tkzz -----END PGP SIGNATURE-----help
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20041020032340.GC3797>