Date: Tue, 27 Mar 2007 10:07:25 -0700 From: Chuck Swiger <cswiger@mac.com> To: "Marc G. Fournier" <freebsd@hub.org> Cc: freebsd-questions@freebsd.org Subject: Re: mdconfig device no faster then direct disk ... Message-ID: <BFCEF5BC-D9CD-4BBE-B90D-51C08A16D2A1@mac.com> In-Reply-To: <A9FC3123E859D7647C969090@ganymede.hub.org> References: <A9FC3123E859D7647C969090@ganymede.hub.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Mar 26, 2007, at 6:55 PM, Marc G. Fournier wrote: > 45 processes: 1 running, 44 sleeping > CPU states: 0.0% user, 0.0% nice, 0.4% system, 0.4% interrupt, > 99.2% idle > Mem: 35M Active, 285M Inact, 271M Wired, 44K Cache, 111M Buf, 402M > Free > Swap: 2007M Total, 2007M Free > > I just did: > > mdconfig -a -t malloc -s 200m -o reserve > newfs /dev/md0 > > Now, my understanding, this builds a file system 'in core', vs on > the disk ... > with memory being faster then disk, I would have assumed that read/ > write > performance would have been better, but, using iozone, I'm not > finding enough > of a difference in performance to understand why I'd want to use a > memory file > system: In order to do useful disk benchmarks, you've got to perform I/O on large enough files that they don't fit into RAM. If you've got 400- odd MB completely unused according to top, you'd really like to use at least 1-2 GB worth of file data. Of course, trying to do I/O tests on a RAM disk means that you want the data to fit into RAM without swapping, which then means that trying to do identical testing between disk and RAMdisk doesn't really work too well. -- -Chuck
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?BFCEF5BC-D9CD-4BBE-B90D-51C08A16D2A1>