A couple of my clients have accomplished this by taking the system to 2GB of memory (an absolute minimum for V6 or V7 and really small for V5) and setting up the remaining memory the same as it would be in production. Then let 5 or 10 folks go at it. Slowly increase the amount of available memory until your at full production. Each test with more memory should give you some indication what the system would be like when loaded up. This is not particularly scientific, but it will do for most auditors. Clearly there is a wide range of things that could be different but if the company does not want to spend any money, it might be your best bet. Use a subsystem with a private memory pool to sequester the memory. You could also use dynamic LPAR to remove processor capability as well. Start out at 0.1 processor and move up until your at the point where the system will run.

IBM does maintain a center for just this type of testing. (there used to be one in Rochester I don't know if it is still there or not) It will not be inexpensive to do a test there.

Jim Oberholtzer
Chief Technical Architect
Agile Technology Architects

On 10/4/2012 10:38 PM, James LeLeux wrote:

I need to perform some kind of stress test on our new Power System before
going live. This is an audit requirement, yet somewhat vague.

I did some searching (ML& Google) and it was typically web related.

I'm dealing with an traditional DB2/RPG app (MAPICS) with green-screen
& GUI client, looking to simulate heavy I/O, perhaps high user count.

I'm open to do-it-yourself and very low-cost (open source) solutions, and
appreciate any guidance you can offer.

Thanks in advance...james

This thread ...


Return to Archive home page | Return to MIDRANGE.COM home page