Re: Upgrading an Oracle server to dual processors

David,

        Last year I conducted benchmarks using a representative work load against 1,2,3, and 4 processors in a COMPAQ Proliant running Oracle 7.2.x. My benchmarks showed significant performance gains from adding processors. For the particular application set that I benchmarked increasing the number of processors resulted in the greatest performance gain. Given the cost of processors and your current CPU load -- go for it! (If I were you I wouldn't stop at two processors, but I would go directly to four processors.)

        In addition you should consider adding memory since 192MB isn't a whole lot when you are supporting 45 users. Going to 512MB and increasing the size of the SGA will also make a significant difference in performance. Even if you don't increase the size of the SGA check the initialization parameters and tweak where necessary. (Get yourself a copy of O'Reilly & Associates Oracle Performance Tuning by Peter Corrigan and Mark Gurry.)

        Incidentally, as a rule of thumb anytime a resource on a system shows more than 75% utilization performance can be improved by upgrading that resource.

Jerry

David Lord wrote:
>
> I am contemplating an upgrade to my Oracle server to try to improve its
> performance.
>
> We currently have a Compaq Proliant 4500 with one P120 processor, 192Mb RAM
> and a 4Gb Fast-Wide SCSI RAID 5 disk array. It is running NetWare 4.11 and
> Oracle Workgroup Server 7.3.2.3.0. We have around 45 users connecting to
> it regularly and the CPU utilisation is averaging 80-90% with peaks of 100%
> for long periods during busy times. The SGA is 53Mb and most cache hit
> ratios are good (the main buffer cache is hitting 80% of the time). Most
> use is via a client-server application using Delphi 1 on the Windows95
> desktops. Our network is unsegmented and is sometimes very congested.
>
> I believe that the major bottleneck is the CPU and I think that upgrading
> to a dual Pentium Pro 200 configuration will help, but I don't have any
> solid evidence that this will improve its perfomance.
>
> If anyone has any views or experiences either way, I would be grateful to
> hear about it.
>
> Thanks in advance,
>
> David Lord.
Received on Wed Oct 08 1997 - 00:00:00 CDT