The OS Average Disk Queue Length Per Disk (Count) alert provides the average number of both read and write requests queued for each disk on the SQL Server computer.
Reduce queue length
If this queue often remains above six per physical disk for more than 10 seconds at a time then your disk subsystem is overloaded. In this case consider the following:
- Check OS Paging to make sure that paging from/to the swap file is not causing these IOs.
- Replacing disks with faster disks.
- Add more disks to you RAID array.
- Switch your RAID array from RAID 5 to RAID 10 solution as each write IO results in two writes using RAID 10 vs. 4 for RAID 5.
- Move other applications to another computer.
- So long as the RAID controller has some form of battery backup then switch its cache mode from Write-through to Write-back as this increases the system ability to handle write IOs by an order of magnitude.
- If the computer is running multiple instances of SQL Server, then consider placing each instance on a separate physical computer.
Create an alert response bundle with the OS Average Disk Queue Length Per Disk (Count) alert and related alerts. For additional information, see Create alert response bundles.