On Tue, 2009-06-23 at 15:21 +0200, Bohmer, Andre ten wrote:
> ]# vmstat 1
> procs -----------memory---------- ---swap-- -----io---- --system--
> -----cpu------
>  r  b   swpd   free   buff  cache   si   so    bi    bo   in   cs us sy id
> wa st
>  2 10 110012  90012   8012 2934768    1    0  2797   237    3    2 10  1 80
> 9  0
>  2 10 110012 125728   8024 2897996    0    0 96560    44 7027 7927 22  3 33
> 42  0
>  1 10 110012 101376   8072 2922136    4    0 73452    36 6834 7377 22  3 46
> 29  0
>  2  9 110012 110076   8040 2913072    0    0 118236    96 7716 8594 22  3 43
> 31  0
>  1 10 110012 104080   8112 2919176    0    0 84912  7880 7096 7920 19  3 37
> 41  0
>  1 10 110012  92572   8076 2931128    0    0 96220    48 7225 8276 24  3 32
> 42  0
>  2  9 110012 112192   7996 2911412    0    0 99088   148 7444 8531 21  3 44
> 32  0
>  4  7 110012  97496   8040 2926508    0    0 86388    60 7475 8212 22  3 44
> 31  0

That output certainly shows a system that is doing a lot of read I/O,
but I wouldn't describe that as an I/O "problem".  This output shows
that your system is reading nearly 100,000 blocks/sec which, probably
50MB/sec, which, if it's Oracle single block sequential reads, and based
on the number of spindles your running against, may or may not be all
that bad as, assuming spinning platters (i.e. not SSD) your sequential
block reads are typically limited by the number of IOPS your array will
produce, not by it's throughput.

> Server is a HP Prolaint Blade BL460C with 2 FC-HBA's, dual quad core en 16GB
> memory.
> 
> Any input is welcome about how to improve IO performance or how to determine
> why IO is the bottleneck.

Well, I think we would need more data.  What is the wait profile of the
Oracle query you are running?  What is the size of the data set?  Is it
expected to fit within the Oracle block cache?  How big is the Oracle
SGA/block buffer cache?  How many spindles and what type of disk make up
the storage?  Can you give us some output of 'iostat -x' while a query
is running?

Later,
Tom


_______________________________________________
rhelv5-list mailing list
[email protected]
https://www.redhat.com/mailman/listinfo/rhelv5-list

Reply via email to