Date: Wed, 8 Nov 2000 17:55:25 -0800 (PST) From: Tom <tom@uniserve.com> To: Brian Behlendorf <brian@collab.net> Cc: stable@freebsd.org Subject: Re: disk I/O Message-ID: <Pine.BSF.4.05.10011081745020.11646-100000@shell.uniserve.ca> In-Reply-To: <Pine.BSF.4.21.0011081625011.1844-100000@yez.hyperreal.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, 8 Nov 2000, Brian Behlendorf wrote: > Ideas? This is actually during a more lightly loaded time period. It > just "feels" like I should be able to push more than 80-100 tps through > da0 (lots of small writes and fsyncs, admittedly) and more than 150tps on > da1 (which is almost all reads). Anything else I should be looking at? A 100tps average is about 10ms per transfer. A good hard drive takes about 8ms average per seek. Assuming every transfer requires a seek, you are not doing too badly. However, tagged commands allow multiple pending IOs per disk. If tags are disabled, or not functioning you will not get this benefit. See camcontrol. Also, mounting your disks with noatime (mount -u -o noatime /blah) requires no downtime and immediately improvement for a lot of things. Basically, it elimiates an IO to update the atime on files and directories. Also, softupdates will be very helpful, but will probably require a reboot. > Brian Tom Uniserve To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-stable" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSF.4.05.10011081745020.11646-100000>