PhysicalDisk\Average Disk sec/Read and PhysicalDisk\Average Disk sec/Write

  • Hi,

    http://technet.microsoft.com/en-us/library/5bcdd349-dcc6-43eb-9dc3-54175f7061ad.aspx

    Is the below recommendation for Average Disk sec/Read & Average Disk sec/Write still valid? or the values should be less than that for modern SAN?

    PhysicalDisk\Average Disk sec/Read

    Indicates the average time (in seconds) to read data from the disk.

    The average value should be below 20 ms.

    Spikes (maximum values) should not be higher than 50 ms.

    PhysicalDisk\Average Disk sec/Write

    Indicates the average time (in seconds) to write data to the disk.

    The average value should be below 20 ms.

    Spikes (maximum values) should not be higher than 50 ms.

    We have EMC SAN.

  • It's all rather subjective and as usual just depends.

    There are a number of problems concerning spikes and averages and the granularity you choose to plot your data will affect your perception.

    However to put this in context you really want the shortest time possible for everything and there are various ways to configure your storage ( sometimes ) and sometimes your server. I use a SAN for one system and typically service times are under 5ms. I usually collect stats at 5 sec granularity, to me a time of 20ms for a read or a write is poor - however caching in all its various guises and the buffer cache is such an example, will affect performance.

    I usually assume that most of the data I require will be in cache so reads will come from memory - I tune specificially in an attempt to make sure I don't have data I don't need in cache and where practical I use compression to effectively get more data usuage. I keep an eye on other memory pools/usuage to make sure I get maximum use of memory - e.g. I try to avoid physical reads altogether. However tempdb is an exception whereby it does physical io. Writes are tuned in an attempt to hit the maximum iops in the shortest time subject to the system capability - so my best system shows 1 ms latency for writes and approx 10 ms for reads ( this make of san is pretty rubbish for reads ) however my read iops are very low so the figure is probably spurious - log drive shows under 3ms for reads ( log shipping ) and under 1 ms for writes. fyi I use only raid 10 15k disks on the san and have 8gb fc connections.

    [font="Comic Sans MS"]The GrumpyOldDBA[/font]
    www.grumpyolddba.co.uk
    http://sqlblogcasts.com/blogs/grumpyolddba/

Viewing 2 posts - 1 through 1 (of 1 total)

You must be logged in to reply to this topic. Login to reply