Hello,
Short time ago I've took over sql server after somebody; machine behaves strange - its performance is poor in term of disk access. Server is on vmware, in the meantime was moved from slower to faster storage array - but problem still persists.
It looks like:
- big file transfers, backup restore etc works fine (up to 300mb/s), while
- doing multiple small oeprations (ie. deleting ~20millions rows from one table) takes ~1,3 hour
Disk queue requests (shown on vm, not array side) are up to 20 in queue then, what suggests io subsystem has issues; on the other hand total io is ~20-25 mbs in same time; this storage may do better.
I believe I've checked all areas where I could expect problems. At the momemt I would say I do not have doubts problem lays on storage / io side. Since few days I suspect problem may be in ntfs setup - seems that disks where databases are located were formatted with standard (4kb) cluster size, while ms recommend to have 64kb cluster size; my question is:
- did anyone notices real benefits after changing cluster size on ntfs? or
- this difference (4 vs 64) may not have serious impact on sql performance?
If only I would have hope it may help, I'll take some efforts to change 4->64, but I am not convinced it may have so huge impact on sql performance - can anyone agree/disagree, based not on books but own experience? I would appreciate a lot.
regards
Tomasz