We've typically seen performance increases when moving from local storage to
the filer.
What kind of pipe are you using? If you're using 100 Mbps Ethernet, you'll
see a performance hit since that's at best 8 MB/sec network throughput,
which is trivial compared to local attach. If you are using gigabit
Ethernet you should see excellent performance.
Also how many disks are you using? Just like with local storage, more
spindles = better performance.
Are you using a private (or at least dedicated) network connection for the
SQL Server to filer connection? If you're sending SQL queries to the box on
the same interface that is talking to the filer that could cause significant
performance degradation.
Finally are you using OLTP or OLAP databases with the filer? (i.e. random or
sequential reads?) NetApp folks might disagree but I think WAFL is a bad
thing to mix with OLAP if you're using snapshots. It's great, absolutely
phenomenal for OLTP, but sequential reads are not a filer's strong point
(that's why it's called write ANYWHERE file layout, not "Write In Order and
Minimize Fragmentation File Layout").
(Please no flames on how WAFL minimizes fragmentation--if you're using
snapshots and you change any data in a database, the new data will be
written elsewhere than the original disk location, causing the head to jump
around like mad during what should be sequential reads. A copy-on-write
snapshot would be MUCH better in this scenario.)
Tell us more about your configuration so we can try to figure out why you're
seeing problems. Also try pinging your sales rep, NetApp has people
dedicated to making SQL Server work well with filers and I'm sure you could
get some pointers from them.
MD
-----Original Message-----
From: Dale Wolf [mailto:wolfd@agcs.com]
Sent: Tuesday, June 05, 2001 7:32 PM
To: toasters(a)mathworks.com
Subject: SQL Database on Filer ??
Has anyone placed a SQL database on a filer and seen a significant
performance increase of 10% or more or a performance decrease of 10% or
more?
Our initial test of a source change management application using a SQL
database on our filer showed a significant decrease in performance to the
tune of 2-3 times slower than a local storage array. We understand ours
wasn't an apples to apples comparison but we weren't expecting the
difference to be this great.
Has anyone else experienced a similar story where they have seen
negative/positive results they would like to share?
Our filer is a model F840.