EMC San - SQL Server Performance vs Direct Attach

  • We're buying a new server and EMC wants to make me a "deal we can't refuse".

    We are getting a 32G Compaq DL740, 4 processors (initially), W2003EE and SQL2KEE.

    We were going to get a pair of 6402 controllers and 28 Ultra320 disks in two shelves, maxing the cache (256M each).

    EMC is offering us a good deal on a CX600 with 40 FC 15k drives, 4x2Ghz processors and 8GB of cache. It's quite a bit more money, but in terms of street costs looks very attractive. We could then add as many HBA's as we want to connect (up to 8) for load balancing.

    They say it will be faster. Quantifying it is not something we seem able to do.

    Anyone have wisdom to share?

  • Depends on how you lay out the SAN. More spindles usually means more performance, but if you carve out a slice of disks for SQL and carve out another slice (Across some of the same spindles) for a print server, performance may or may not be better. Mixed response here. Initially everything seems smoking, later some perf issues, but hard to tie back to the SAN as the culprit. If you need the space, SANs let you get much bigger than with direct attach.

    Steve Jones

    sjones@sqlservercentral.com

    http://www.sqlservercentral.com/columnists/sjones

    The Best of SQL Server Central.com 2002 - http://www.sqlservercentral.com/bestof/

    http://www.dkranch.net

  • Despite being a SAN we would likely attach no other servers to the CX600, just SQL Server on one DL740 system.

  • No one with other wisdom to share?

    we've decided to give this a try. We're getting 45 drives in the CX600 (a bit of negotiation) and about the 2nd week in January we ought to find out if we are making a huge mistake.

    Further advice always welcomed, especially people who have used the CX line from EMC on SQL. Otherwise will share some experiences afterwards.

Viewing 4 posts - 1 through 3 (of 3 total)

You must be logged in to reply to this topic. Login to reply