Database Health Monitor Reports on slow IO on TempDB.
If the write stall or read stall for TempDB is greater than 30ms you will see this warning.
The numbers shown in the Quick Scan report are the average stalls per read and stalls per write since your SQL Server instance was restarted.
What type of IO response should you see on TempDB?
It depends mostly on your I/O subsystem, but typically anything over 20ms to 30ms should be considered excessive.
If you are seeing numbers greater that 50ms this is a very bad sign.
Things you can do to improve TempDB I/O speed:
- Move TempDB onto a dedicated drive of its own.
- Move TempDB to faster disks (SSD).
- Reduce the usage of TempDB.
- Consider Trace Flag 1118.
- Contact Steve at Stedman Solutions, LLC for help.
Stedman Solutions, the provider of the Database Health Monitor Application offers consulting solutions, and can help with getting your backups and restores, or any other database administration issues you may have.