Legacy HDS Forums

vxstat stats vs raw hard disk IO/s

Discussion created by Legacy HDS Forums on May 7, 2008
Latest reply on Dec 19, 2008 by Legacy HDS Forums

Originally posted by: przemolb



Hi all,

I am facing a problem which comes from question "How do you define IO ?" (IOPS).
We have hds 9570 with 12 HDDs (+2 spare drives) and two RAID-5 RGs (5D+1D). There are several LUNs on each RG. It is connected to sparc server with Solaris 10 and Veritas 4.1. We have several Oracle databases (OLTP) working on the server/array. When I run vxstat I get:
[code]
drozd:/root# vxstat -g drozd-sierra2-dg -i 60
                           OPERATIONS            BLOCKS             AVG TIME(ms)
TYP NAME                 READ      WRITE       READ      WRITE   READ  WRITE

[...]

Wed May 07 15:07:13 2008
vol drozd-sierra2-vol     466239        548   7450629       7714     0.3     0.6

Wed May 07 15:08:13 2008
vol drozd-sierra2-vol     319741        693   5108947      10307     0.3     0.9

Wed May 07 15:09:13 2008
vol drozd-sierra2-vol     331786        404   5301400       4036     0.3     0.4
[/code]
An average HDD can do about 200 IOPS. So if I count 200*12 I get 2400 IOPS. So how do we get 7770/s READs operatios (466239/60) ? Is the controller able to optimize the workload to achieve the result or do I mix different terms ?

Outcomes