How to Determine Database Performance Limits

From: Ethan Post <Blah_at_Blah.com>
Date: Wed, 03 Oct 2001 00:02:40 GMT
Message-ID: <AAsu7.19228$Xk4.1286159@news1.rdc1.sdca.home.com>

One of my tasks at hand is ensuring our 12 CPU AIX box can process 150,000 lines of sales orders per hour via J.D. Edwards. We have performed extensive testing and determine we will have to move the application logic to other machines and leave the DB where it is. My job now is to ensure the current box running 8.1.6 is capable of the task at hand. I am currently focusing on redo because I think that could be a potential bottleneck. Based on the numbers we already have during testing and extrapolating them we will need to be capable of the following:

120-160 MB of redo generated per minute
300-400 MB of traffic per minute sent from client (middle tier) via SQL*Net
600-800 MB of traffic per minute sent to client (middle tier)
1.6-2 million SQL*Net roundtrips per minute.

45-60 thousand commits per minute.
1.4-1.8 million user calls per minute.

What will this achieve? An amazing 41 records per second! This gives you some idea of the swell coding going on in a modern day ERP system.

Anyway, my question is how the heck do I determine if the box can actually do this? What is the best way to determine if the current box will handle the load. I can tell you simulating a load this heavy is hard to do!

Your advice?

Thanks,
Ethan Post Received on Tue Oct 02 2001 - 19:02:40 CDT