Date: Sun, 5 Apr 1998 22:03:36 -0500 (EST) From: "John S. Dyson" <toor@dyson.iquest.net> To: thomma@slip.net (Tamiji Homma) Cc: freebsd-current@FreeBSD.ORG Subject: Re: disk performance difference Message-ID: <199804060303.WAA07645@dyson.iquest.net> In-Reply-To: <E0yM1zh-0006HN-00@slip-3> from Tamiji Homma at "Apr 5, 98 07:50:44 pm"
next in thread | previous in thread | raw e-mail | index | archive | help
> Hi, > > I am not a -current but SNAP CD user. I noticed something interesting, > that 19971006-SNAP disk performance is much greater than 19980222-SNAP > specially on my SCSI disk according to what Bonnie's says. In fact, > SCSI disk performance is very poor... > Try building a -current kernel, and I wonder if you see such a regression now. There have been several improvements (and fixes) lately. (It is fairly easy to build a -current kernel, running on a recent snapshot. Sometimes, you'll have to rebuild libkvm and ps, but shouldn't be a major problem, once you go through the learning curve.) > > 19980222-SNAP: 3315.51 real 2313.34 user 599.32 sys > 19971006-SNAP: 3370.48 real 2328.91 user 633.53 sys > > The 'make buildworld' is done under following conditions: > /usr/src on wd0 > /usr/obj on ccd0 (/dev/sd0c /dev/sd1c) > NOPROFILE, NOCLEAN, NOTCL, -pipe, async,noatime > > I don't see much difference (actually faster on newer SNAP, which is > good:) between two SNAPs when I do make buildworld. > > Can anybody elaborate why? > I don't have any details, but there have been numerous problems introduced in the name of improvements. It is likely that you are running a snap with the ioopt feature enabled, and that option was not well thought out before including in the tree. (I know, I implemented it.) John To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199804060303.WAA07645>