Slow average service times on SAN

Joe,
How did you measure response time of your SAN?


"Jochen Kaiser" wrote in message
...
Hello,

I'd like to verify whether the following situation can be considered
'normal'.

Our databases (Solaris 8, DB2, UFS, 300GB OLTP and a 150GB DWH) are
housed on a SAN array. (HDS9570V equipped with 146GB disks, McData
switching in between, single hop, two FCs from host-side)

I see (sar) very slow service times from the array, especially on those
partitions housing our database-files. The response times (including
queuing time) average at 50-100ms and peak at up to 500ms when we're
loading data to the data warehouse or the OLTP.
These scenarios result in lots of random reads on 4k and 16k tablespaces
and a fair share of random writes in the same space-range. I'd expect a
ratio of 15:1 concerning reads vs. writes.

My Datacenter is telling me that they consider this situation normal,
although they admit that the service times 'could be better'.
We're sharing the array with a number of other applications, all of
which fall into the OLTP category, but I'm unable to measure any
performance increase during the night (when the OLTPs should be silent).

I'd expect that service times are much better, 20ms max.

Is this a realistic assumption?
What would you try to get as an agreed SL?
Are service times like the ones above normal in your SAN?


Many thanks in advance for you help,

Jochen