--On April 19, 2006 12:47:26 PM -0500 John Wade <jwade@xxxxxxxxxx> wrote:
We run cyrus on RedHat on a CX600. For our administrative users, we put the whole thing on a nine disk fibre channel RAIDgroup, split into 6 luns, 5 for mail spool partitions and 1 for all the other metadata. Total mail spool size is 300GB. We average about 300 concurrent users. The CX600 is heavily used by other applications. I/O performance has never been an issue as far as I can tell. The large write caches in the CX600 solve the write problem and the host memory caching seems to solve the read problem. Your mileage may vary. We couldn't do a direct cyrus SAN performance comparision because when we moved to this box, we migrated to the SAN and new server hardware at the same time, but tests on other systems where we just moved storage on to the array saw a huge performance boost vs. locally attached SCSI RAID 5 storage with no write cache enabled.
Our mail cluster lives and dies by our CX200's write caches. When they get turned off (as can happen during a failover rebuild on the SAN, or if you lose a vault drive) the mail cluster starts choking. Even though we have used array's with write caching before the algorithms in the CX are definitely better. ----
Cyrus Home Page: http://asg.web.cmu.edu/cyrus Cyrus Wiki/FAQ: http://cyruswiki.andrew.cmu.edu List Archives/Info: http://asg.web.cmu.edu/cyrus/mailing-list.html
