Please start any new threads on our new
site at https://forums.sqlteam.com. We've got lots of great SQL Server
experts to answer whatever question you can come up with.
| Author |
Topic |
|
idleflow
Starting Member
4 Posts |
 Posted - 2009-11-11 : 06:31:46
|
| Machine specs:sqkl server 2005 enterprise 64bit. MDF file is sitting at 145GB32GB of ram with 8 hyper threaded intel xeon E5345 2.33ghz processors.db files are sitting on a raid 5 with 3 drivesI've noticed over the past few months that we've been experiencing some pretty bad write latency. I've gone through a series of steps such as removing unused indices, adding some extra indices and including identity columns (as clustered indices) on our larger tables (the original designers of the database used unique identifiers for PK's). I have also moved our larger ldf file onto a different drive as all the db files were sitting on the same drive. TempDB and the main database mdf are on the same drive.I understand that a raid 5 setup isn't optimal and this is something we're going to address very soon. One possibility is that the cache is being exhausted by another drive on the san, possibly a sata drive that experiences large amounts of traffic. How could I prove this to the network engineers?So far I have planned to seperate the tempDB filegroup into 4 files and move them onto a different drive as i understand this is the norm.What else can I check/do to improve/resolve the performance of file writes? |
|
|
|
|
|
Write latency