JetStress (WR1.2): Where storage hardware cache size is used

  • Articolo

JetStressUI 2004 (WR1.2: 6.5.7529.0) uses 'storage hardware cache size' for two places:

  • One is to check the test database size is large enough and to "please make sure that the total database size is at least double the size of the storage hardward cache size".
  • The other is to determine the initial number of threads for the tuning phase. JetStressUI 2004 (6.5.7529.0) initializes this number based on the target IOPS (I/O per seconds) and the cache size.
  • NOTE: JetStressUI 2004 (WR 2.1: 6.5.7726.0) initializes this number to be the number of CPU(s).

This posting is provided "AS IS" with no warranties, and confers no rights.

Comments

  • Anonymous
    August 28, 2006
    We have just upgraded our SAN and are preparing to migrate to Exchagne 2003.  However, there is a question on what to use for the cache size.  The SAN has 96GB of cache and we'll be running JetStress on 24 servers in tandem.  Should the cache on all 24 be set to 96GB or divided accross the 24 and set at 4GB?  Also, for #of disk drives, we are using 6 spindles per SG assigned to one logical drive.  Lastly, would you recommend AutoTuning in this environment?

    Thanks
  • Anonymous
    August 29, 2006
    You should divide it across 24 servers. The total database file size across 24 servers must be 198 GB to have the cache saturated and then to get realistic performance characteristics for the disk sub-system.

    The automatic tuning is not well suited for this type of complext deployment scenario. Under a stressful load situation, one server contines ruducing load generation, and the other continues increasing load generation --- which means very unbalanced result. From one test run to another, the result would be very unpredictable.

    I would recommend you supressing the tuning at all and entering your own number of threads and the other test parameters.