Re: analyzing db for performance bottlenecks

In article <8qbmds$q9j$1_at_garnet.tc.umn.edu>,   hause011_at_garnet.tc.umn.edu (Steven Hauser) wrote:
> You need to give us a few more details and take a system wide
> view of performance. Your testing might not even be relevant.
>
> What sort of Oracle server hardware is it, a one disk box with a
 single
> CPU and 64MB RAM? Or a EMC 10 Terabyte EMC disk array with an
> E100000 128 CPU Sun?
>

It is an E4500, with 8G of RAM and 8 400MHz CPUs. The disk array was not EMC, but SUN's. I don't remember the exact configuration, but it was huge.

> 4*1800/minute is 120/second SQL processed. That might be all you
> get for small old hardware.
>
> How much data? The typical one row per table development environment
 or
> is that 10 TB EMC disk array full up?

This is a typical development env database. It is not really one row/table, more like 100 rows/table.

>
> Is your SQL even using indexes? Using partitioning?
We have defined indexes, but not explicitly specifying which indexes to use in our SQL. The oracle expert did some partitioning, i need to find out the details of that. But, on analysis, disk use seemed ok.
>
> What sort of network? 10mb nics on a shared segment with 127 other
> machines running lots of chatty protocols?
It was dedicated connection between java server machines and db machine (100 Mb)
>
> Default init.ora for instance configuration? With the default 'demo'
> instance?
>

Init.ora was modified to give lot of shared pool and other things like block size, sort area etc. were also modified.

> And I hope you are not starting and stopping a connection to the
 database
> each time you do a request.

No. The connections are pooled.

Manju.

Sent via Deja.com http://www.deja.com/
Before you buy. Received on Thu Sep 21 2000 - 00:00:00 CDT