SQL Server Performance

Storage related performance issue, yet very low avg disk queue length

Discussion in 'ALL SQL SERVER QUESTIONS' started by Glippy, Feb 5, 2012.

  1. Glippy New Member

    Hello,

    At a customer we are experiencing severe performance issues. The performance issues seem to be storage related, since it matters a whole bunch on which drives we place the database. When the database is placed on SAN 1, certain query's take about 4 minutes to complete, while when the exact database is placed on SAN 2 the same query's take less than a minute to complete.

    But what I just cannot comprehend is the low average disk queue length, while monitoring the performance on both SAN's. On the fast SAN the disk queue length is '1' for about a minute, while on the slow SAN the disk queue length is '1' as well but for about 4 minutes. I would think that the disk queue length, especially on the slow SAN, would be a lot higher if the performance bottleneck is really storage related.

    It doesn't make sence to me that the average (read) disk queue length does not exceed 1. Why on earth does SQL Server (2005) not through a whole bunch of those I/O requests in the queue? Am I missing something here?

    Thanks for thinking with me.

    Glippy
  2. davidfarr Member

    As I understand it; the number of i/o requests sent by SQL server largely depends on the the query execution plan calculated by the engine. For the same query on the same database, it might have happened that the query sent 4 i/o requests to both SANs. 3 of those requests were serviced immediately (therefore not queued) and 1 was queued, resulting in the same average queue value of '1' on both SANs.
    An average disk queue length of 1 or 2 is actually good queue performance, only values above 2 are considered bad.

Share This Page