natsidan

updated

16 years ago

N

natsidan

1 Rookie

•

34 Posts

0

807

March 16th, 2011 14:00

Analyzer Output

When I view the Lun Read size KB it is a lot smaller than the disk Read size KB. Even when you combine all the Luns within the Raid group the Read size KB does not come close too what the Read size KB is on disk

e.g 4 Luns all Read size 2KB

yet the disk read size can be as high as 229 KB ?

any ideas

  • kelleg

    6 Operator

    •

    4537 Posts

    420

    0

    Posted March 22nd, 2011 15:00

    Response times at the LUN is what the user sees, not the disks. Also, be careful with response times - it's a calculated value based on queue length and IOPS. If IOPS are low (under about 100 IOPS), then response times can be very high - you can ignore these.

    glen

  • driskollt1

    131 Posts

    420

    0

    Posted March 17th, 2011 06:00

    The LUN stats are what the host is sending/receiving from the array.

    The disk/RG stats are what the disks are actually doing.  If the host sends some smaller sequential reads/writes - usually the array will combine them into larger reads/writes to the disk.  Also your array is probably prefetching reads.

    Basically your array is doing what it's supposed to be doing - Being more efficient at the disk level.

    The LUN numbers will almost never match with the drive/RAID group metrics.  LUN I/O stats don't account for RAID penalties, caching, etc.

  • natsidan

    1 Rookie

    •

    34 Posts

    420

    0

    Posted March 17th, 2011 09:00

    Ok thanks, that makes sense

    So does that make Lun reponse time it terms of host i/o a better counter than disk response time from a client point of view

    thanks