On Fri, 2009-07-03 at 08:50 +0100, Nick Lunt wrote:
> Hi  folks
> 
>  
> 
> anyone running Oracle with iSCSI ? Were getting 8MB/s max throughput
> with Oracle 10gR2.

8MB/s throughput doing what exactly?  If this is the speed you're seeing
when doing full tables scans, or backups, or other things like that,
well, then you have a problem.  If this is the performance you're seeing
when doing single block reads (i.e. db_file_sequential_read) then based
on the block size of your database you're most likely IOP limited.  How
many spindles?  What's you Oracle block size, read latency, and total
IOP capacity.

We run multiple Oracle databases on Equallogic iSCSI SAN, fairly small
(250GB) but reasonably transactional (Oracle eBusiness 11i) and pretty
small block size (8K).  Our production Equallogic array is 15K RPM SAS
drives and performance is about 20% faster than what we used to see with
our EMC Clariion fiber channel solution (10K RPM drives, but more of
them).

If you're doing single block reads (db_file_sequential_read, i.e. index
scan by ROWID) you will be drastically limited by number of spindles and
the read latency of your array.  For example, if your block size is 8K,
and average read latency is 1ms (which is pretty fast for spinning
media), then you can read at most 1000 blocks per second or about
8MB/sec.  Of course that's assuming only a single query, multiple
query's will likely perform better since requests can be merged.

Later,
Tom



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

Reply via email to