Re: Storage cache impact
- From: Martin Berger <martin.a.berger@xxxxxxxxx>
- To: Laimutis.Nedzinskas@xxxxxx
- Date: Tue, 1 Feb 2011 20:47:56 +0100
Are you talking about exadata storage index,
or just any cache in a general SAN?
If you are talking about general SAN and their cache:
I never found a storage-admin who accepted to flush or even worse restart it's
box just to test a 'cold' cache.
But you can just easily do some estimations:
1st get the real average physical IO time of the disks from your SAs (you
probably will not get these numbers at the first attempt. Try to go for some
beers with them!)
then get the AWR or statspack report of the test-run you did. There replace all
the average IO-times (scattered, sequential, direct, ...) with the probably
higher numbers you got from your SA.
Do the reverse calculations. This should give you a 'worst time' scenario!
this is by far not 100% perfect. But it was the best I got.
Martin
Am 01.02.2011 um 15:43 schrieb Laimutis.Nedzinskas@xxxxxx:
>
> Hi
>
> I have a question which may be regarded not a practical one but still...
>
> When assesing SQL it is nice to know the worst time it will take to
> execute.
>
> This is the reason one can flush buffer cache of a database.
>
> But what about storage cache?
>
> Anyone tried to estimate DB performance with cold and hot SAN-storage
> cache ?
>
> How does storage cache help database ? For example, currently my total DB
> cache (accross all db's servers using the same SAN) is way larger than SAN
> cache.
> My understanding is that in such a situation SAN cache helps mainly SAN
> itself ( like caching RAID5 writes, etc ) but it can not effectively cache
> database data.
>
> Any commenst are wellcomed.
>
> Thank you in advance, Laimis N
>
>
>
>
> ---------------------------------------------------------------------------------
>
> Please consider the environment before printing this e-mail
>
> --
> //www.freelists.org/webpage/oracle-l
>
>
--
//www.freelists.org/webpage/oracle-l
Other related posts:
