Physical Disk Queue Length and SAN | SQL Server Performance Forums

SQL Server Performance Forum – Threads Archive

Physical Disk Queue Length and SAN

I’ve always been told that an average disk queue greater than 2 per disk needs to be investigated. How do you determine this when the disk in question is a chunk of SAN space? I have no idea how many physical spindles make up my drive space and our storage admin isn’t real interested in looking into it.
thats 2 per physical disk, so if your SAN array is made up of 10 disks, then upto around 20 is probably ok, but what you are really after is the disk latency, it just so happens that when disk queue exceeds 2, disk latency can start to climb rapidly, hence the rule of 2 what is your disk latency? Avg Disk Sec/Read & Avg Disk Sec/Write
get these for each disk separately, the total has no meaning
as monitor Avg Disk Read Queue Length, Read/sec, Read Bytes/sec and same for writes.
do this for each physical disk (the Perfmon counter Physical Disk that is) ie, follow my instructions in www.qdpma.com on what to monitor, it doesn’t matter if you want to solve the problem yourself, you still need the right information. also, rigid rules like this are stupid, you have to THINK about what your application is doing,
in a DSS database, let it run up to anything you want
on a OLTP db, spikes due to checkpoints are also ok so long as the duration of the checkpoint spike is reasonably short
]]>