Tom,
Thanks for your feedback.
Tom "Mad Dog" Yergeau wrote:
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.
Filer and server both use gigabit Ethernet and are plugged into the same switch. We are checking to see if there is any kind of problem with auto negotiating. We had a similar problem a few years ago with some EMC equipment where it would negotiate at 10 Mbps on 100 Mbps Ethernet interfaces.
Also how many disks are you using? Just like with local storage, more spindles = better performance.
Separate volume with six partitioned spindles. Half for an low utilized Oracle database and the other half dedicated to SQL database. We are contacting NetApp to find out if our disk configuration could be part of the problem.
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.
SQL queries are being sent across the same gigabit interface used to talk to the filer. The switch shows roughly 20-30% utilization.
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").
Vendor says application is primarirly random reads.
(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.
We are contacting our local reps too. They say we should get equal or better performance.
MD
-----Original Message----- From: Dale Wolf [mailto:wolfd@agcs.com] Sent: Tuesday, June 05, 2001 7:32 PM To: toasters@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.