Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> c.d.o.misc -> Re: Benchmarking

Re: Benchmarking

From: Mark Rosenbaum <mjr_at_netcom.com>
Date: 1997/06/21
Message-ID: <mjrEC4tH3.H5s@netcom.com>#1/1

In article <5o5as3$9r2$2_at_pornstorm.eit.com>, Adrian J. Blakey <ajb_at_uniblab.alameda.net> wrote:
>I am looking for the benefit of some collective wisdom about
>benchmarking Oracle 7 for data warehousing.
>
>Specifically:
>
>What was the characteristic of the test application you ran?

Both EIS/DSS and database marketing

>What were the metrics?

Load times, query times and purge times.

>What did you tune?

Hardware config, OS kernal, Oracle init.ora.

Hardware is processors, memory, I/O channels and disk. OS kernal varies with each platform
Oracle init.ora varies with release

>What hardware/OS/platform considerations did you apply?

Today you would like to give serious thought to 64 bit OSes. Also Symmetric Multi-Processors are a must.

>I am about to do something like this and also need some paid help.

Adrian,

I have done a fair amount of benchmarking in general and Data Warehousing in specific. Here are some general recommendations:

Try to make the benchmark reflect the app as much as possible

Benchmark the exact platform and release that you will be using.

Test full functionality at full volume.

Test complete cycle (load, query, purge) and not just query only.

I spent years around vendors and integrators benchmarketing systems and have seen more than once techniques that increased the speed of the benchmark did not increase the speed of the final app.

Things change from every release of everything. Benchmarking slightly different releases can lead to very unpleasant surprises.

I have been caught on things that worked fine at low volumns but had bugs at high volumns.

Most people test query times and most project run into problems with loads and purges.

In data warehousing the frontend tools can have a MAJOR impact on queries. One of the best ways to speed up queries is to use summary or aggregate tables. To take advantage of these you need an aggregate aware frontend.

I hope this helps. If you would like to discuss this further feel free to contact me directly.

Mark Rosenbaum			Otey-Rosenbaum & Frazier, Inc.
mjr_at_netcom.com			Consultants in High Performance and
(303) 727-7956			Scalable Computing and Applications
POB 1397			ftp://ftp.netcom.com/pub/mj/mjr/resume/
Boulder CO 80306 Received on Sat Jun 21 1997 - 00:00:00 CDT

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US