msgssz

[Donny Jekels]

Does anyone know what the correct size for msgssz should be for a nPar on SD64 with 18 oracle database instances.

currently it is set to 16, and we are experiencing IPC degregation?

[A. Clay Stephenson · Accepted solution]

Typically something about 128 (prevents going to the well for more msg space too often) is a good compromise although typically it's sema4's that are more prone to cause bottlenecks.

Since you mentioned Oracle, it is possible that you "borrowed" some older kernel settings and set timeslice very low -- possibly 1 -- that would cause IPCS problems out the wazoo.

[twang]

msgssz: in generalm, msg segment size should be word size multiple. I think it is difficult to say how mauch size it should be, you must simulate or run the application to measure the average fullness of a queue or its average length. This process may require a lot of trial and error: you may need to estimate values for your tunables before running the application, and then adjust them after running under performance analysis.
For a large system, analyze the effects of parameter settings on the size of the operating system kernel. If they are unacceptable, reduce the number of application processes or you may need to distribute the application across more machines.

[James R. Ferguson]

Hi Donny:

As a guide to choosing an appropriate value, follow the discusssion (and its links) here:

http://www.docs.hp.com/hpux/onlinedocs/TKP-90202/TKP-90202.html

Regards!

...JRF...

[Steven E. Protter]

You need to pick a figure based on ipc use.

If you are using apache, that uses ipcs.

If you are using oracle, that uses LOTS of ipcs. I see you are using LOTS of oracle.

I think you need to collect some data so you can set an intelligent figure.

I can provide you figures off of an oracle box with three instances thats running very nicely right now.

See attachment.

SEP