Date: Sat, 29 Oct 2005 00:17:54 +0200 From: =?ISO-8859-1?Q?S=F8ren_Schmidt?= <sos@FreeBSD.ORG> To: Mikhail Teterin <mi+mx@aldan.algebra.com> Cc: Peter Jeremy <PeterJeremy@optushome.com.au>, hackers@FreeBSD.ORG, hardware@FreeBSD.ORG Subject: Re: Very slow writing to SATA disk Message-ID: <D64EDD7D-644E-4DCD-8B99-450403234A5D@FreeBSD.ORG> In-Reply-To: <200510281803.15734.mi%2Bmx@aldan.algebra.com> References: <200510280518.j9S5I3oQ025430@blue.virtual-estates.net> <200510281745.43138.mi%2Bmx@aldan.algebra.com> <2835B7B9-03AA-4EB2-996B-8CEC485071C6@FreeBSD.ORG> <200510281803.15734.mi%2Bmx@aldan.algebra.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On 29/10/2005, at 0:03, Mikhail Teterin wrote: >> Ask the maintainer to get it [ataidle -mi] fixed, but be warned =20 >> experience >> says it might hose your data... > The maintainer did not break it. An incompatible change to the API =20 > did :) You > are, probably, in the best position to show us, how the new API =20 > should be > used. Look in smartmontools I provided patches for that, its not rocket =20 science you know... > >> Now, you say read speed is OK, but write speed isnt, is that on the >> raw disk device or though the filesystem ? > > Everething is through the filesystem -- as stated in my original e-=20 > mail in > this thread. There is no other activity, when a single cp reads the =20= > huge file > from a SCSI disk (da1) to the IDE (ad8). > > According to `systat -vm', da1 is barely breaking a sweat, while =20 > ad8 is at > 99-101% throughput at 7Mb/second. > > Soft-updates are on. The filesystem is almost empty. The box has a =20 > single > dual-core Opteron-275 with 2Gb of RAM. > > The SATA controller is your "favorite": > > atapci1: <SiI 3114 SATA150 controller> port > 0xac00-0xac07,0xa480-0xa483,0xa400-0xa407,0xa080-0xa083,0xa000-0xa00f =20= > mem 0 > xbe6fbc00-0xbe6fbfff irq 25 at device 5.0 on pci3 OK, then we dont know where the slowdown is yet... You need to find out what the transfer rates are for the RAW disk, ie =20= by doing a dd from /dev/zero to the disk with a a resonable blocksize =20= say 1M to minimize overhead. Also read speed from disk to /dev/null =20 blocksize 1M would be helpfull. S=F8ren Schmidt sos@FreeBSD.org
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?D64EDD7D-644E-4DCD-8B99-450403234A5D>