Date: Thu, 12 Nov 2009 15:39:13 -0600 From: "Sam Fourman Jr." <sfourman@gmail.com> To: Scott Ullrich <sullrich@gmail.com> Cc: Rick Macklem <rmacklem@uoguelph.ca>, Dan Nelson <dnelson@allantgroup.com>, Artem Belevich <fbsdlist@src.cx>, freebsd-current@freebsd.org, Ivan Voras <ivoras@freebsd.org> Subject: Re: Help ZFS FreeBSD 8.0 RC2 Write performance issue Message-ID: <11167f520911121339y4161a1a2sae8b6196c3b7a1fb@mail.gmail.com> In-Reply-To: <d5992baf0911121314k7f535f2cra2d3e8367774d8ca@mail.gmail.com> References: <11167f520911111050j36dd94far667c81e6f5c18e69@mail.gmail.com> <20091111204903.GI89052@dan.emsphone.com> <11167f520911111326v13bb442bt36e853afbecdf834@mail.gmail.com> <9bbcef730911111352t12188bdajbca71bcf35a5beb5@mail.gmail.com> <Pine.GSO.4.63.0911121147310.9126@muncher.cs.uoguelph.ca> <11167f520911121044l74744c30u5a4d9ca008ab863c@mail.gmail.com> <d5992baf0911121048o76c1543el2b4d9de84ffb107@mail.gmail.com> <11167f520911121101o403751ddmb544dfaf1c61bf1e@mail.gmail.com> <ed91d4a80911121219sa0f3fa3s934791725028a966@mail.gmail.com> <d5992baf0911121314k7f535f2cra2d3e8367774d8ca@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Thu, Nov 12, 2009 at 3:14 PM, Scott Ullrich <sullrich@gmail.com> wrote: > On Thu, Nov 12, 2009 at 3:19 PM, Artem Belevich <fbsdlist@src.cx> wrote: >> >> Log seems to be somewhat weak point in ZFS. If you lose your log >> device, you will lose your pool. Plus, there's no way to remove log >> device from the pool. So, once you attach some device as a log, you'd >> better be sure that device does not disappear, because it will take >> the rest of the pool with it. From that point, real ram-disk (i.e. >> /dev/mdN) is definitely a recipe for disaster. External ramdisk with a >> battery backup may be an option, but even in mirrored configuration >> seems rather risky to me. >> >> http://jmlittle.blogspot.com/2008/05/problem-with-slogs-how-i-lost.html >> >> I'd say that SSD are probably the best fit for slog role. > > Indeed. =A0 I mirrored 2 SSDs on an Areca in case I loose one of them. > Partitioned the SSD into a log device and the rest being cache (see > the ZFS best practices guide for details). > > Needless to say my performance matches that of normal writes and reads > when using NFS now. if you wouldn't mind posting, what kind of NFS performance can you achieve? my application is office Fileserver (the clients are FreeBSD 8 diskless PXE boot) so idk if it is possible to achieve gigabit NFS performance. iperf is able to get ~970mbit from a client to the NFS server. I realize that disk i/o is the limiting factor so I bought 6x SATA2 disks and a Areca card Sam Fourman Jr.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?11167f520911121339y4161a1a2sae8b6196c3b7a1fb>