Oracle FAQ | Your Portal to the Oracle Knowledge Grid |
Home -> Community -> Usenet -> c.d.o.server -> Re: Performance problem
"Tony" <trindata_at_bellsouth.net> wrote in message
news:Ildlf.19132$wi2.16460_at_bignews1.bellsouth.net...
> After devoting a great deal of effort to tune a particular batch process
> some physical changes were made to our environment. The result is that
our
> batch job now takes twice the amount of time to execute (7 hours instead
of
> 3.5!)..
>
>
>
> The changes made were as follows:
>
> 1.. 1 GB Nics were replaced by 100 MB cards
> 2.. 15k rpm smaller drives were replaced by larger 10k rpm drives
> 3.. Disk layouts were changed to make use of the larger drives
> Other than that, the application has not been changed. The operating
system
> and Oracle release have not been upgraded either. I am trying to make a
> case for returning the environment to its previous state (Nics and drive
> speed) while collecting information using in-house tools such as
Statspack,
> TKPROF etc. In the absence of detailed analysis of the previous runs (3.5
> hours) what can I use to highlight the negative impact of the changes made
> to our environment?
>
>
You need a benchmark that was on the old system and on the new system. I
would think the batch process you cite is that benchmark. Why do people
think replacing drives with slower ones will speed things up? The NIC cards
may not make any different. You could measure the IO and max IO rate on the
drives in stats pack but what do you have it to compare it to?
Jim
Received on Tue Dec 06 2005 - 07:20:45 CST