On Mon, 19 May 2008, Andy Lubel wrote: > Or just do ZFS (or even SVM) setup like Bob and I did. Its actually > pretty nice because the traffic will split to both controllers > giving you theoretically more throughput so long as MPxIO is > functioning properly. Only (minor) downside is parity is being > transmitted from the host to the disks rather than living on the > controller entirely.
The bottleneck is in the StorageTek 2540 storage array itself rather than the connections to it. Note that for mirroring, ZFS needs to send more data (2X) over the fiber channel when writing than if the storage array was doing the RAID. Regardless, there is only a very tiny reduction of sequential write performance due to using ZFS with a LUN per disk rather than RAID-1 in the storage array. The sequential read performance is improved considerably since ZFS can intelligently load-share its reads across the mirrors without depending on the RAID array to do that. Giving more of the responsibility to ZFS allows performance to improve since ZFS is more aware of the task to be performed than the drive array is. Bob ====================================== Bob Friesenhahn [EMAIL PROTECTED], http://www.simplesystems.org/users/bfriesen/ GraphicsMagick Maintainer, http://www.GraphicsMagick.org/ _______________________________________________ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
