On Fri, Aug 6, 2010 at 6:44 AM, P-O Yliniemi <p...@bsd-guide.net> wrote:
> Hello! > > I have built a OpenSolaris / ZFS based storage system for one of our > customers. The configuration is about this: > > Motherboard/CPU: SuperMicro X7SBE / Xeon (something, sorry - can't remember > and do not have my specification nearby) > RAM: 8GB ECC (X7SBE won't take more) > Drives for storage: 16*1.5TB Seagate ST31500341AS, connected to two > AOC-SAT2-MV8 controllers > Drives for operating system: 2*80GB Intel X25-M (mirror) > > ZFS configuration: Two vdevs, raid-z of 7+1 disks per set, striped together > (gives a zpool with about 21TB storage space) > > Disk performance: around 700-800MB/s, tested and timed with 'mkfile' and > 'time' (a 40GB file is created in just about a minute) > I have a spare X25-M drive of 40GB to use for cache or log (or both), but > since the disk array is a lot faster than the SSD-disk, I can not see the > advantage in using it as a cache device. > > Is there any advantages for using a separate log or cache device in this > case ? > > Regards, > PeO > > _______________________________________________ > zfs-discuss mailing list > zfs-discuss@opensolaris.org > http://mail.opensolaris.org/mailman/listinfo/zfs-discuss > I can tell you for sure that there can be a really nice advantage for sequential writes. to see this yourself, do the following: create a filesystem, share it out NFS create a really big tar.gz file and put it in the filessytem log in from a network client via nfs and extract the tar.ball using something like: time tar xzfv some.tar.gz do this a few times to get an average, then add the SSD as a log device. I have the exact same motherboard with a very similar setup, and i noticed a 400% nfs performance boost by doing this. try it yourself =)
_______________________________________________ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
