IPCS Subsystem Queue

[Wayne Moyer]

Performance reporting shows IPCS Subsystem Queue consistently in the range of 8 to 10. This is on a HP9000 RP7420 with 8 CPU’s. Does this seem high for this server or should this be close to the number of CPU’s? All other run queues are in the range of 1 to 2.

[A. Clay Stephenson · Accepted solution]

This is truly an Apples to Oranges comparison and run queues have nothing to do with IPCS message queues or semaphores. The length of IPCS message queues generally has much more to do with how an application is designed that any other criterion ---- including overall performance of a system. For example, you may have designed an application that uses IPCS messages to tranmit data between processes. It can assign numerical message types that correspond to priority so that it might be completely normal for a herd of lower priority messages to queue up while higher priority messages are processed.

Without an understanding of the underlying software, it just isn't possible to read much into the length of an IPCS queue. You can use tusc to help you understand what is going on.

[Steven E. Protter]

Shalom,

Answer: It depends (shamelessly stolen from A. Clay).

What is the server used for? Database, OLTP, Data mine.

What is the software in use? Where the tuning guidelines for it followed?

SEP

[Wayne Moyer]

Thanks for the replies. This server is running Oracle and a 3rd party software application. All the vendorsâ recommendations have been followed. I noticed this metric and wanted to know if this could be normal. The vendor has not responded to this yet but I will follow up with them.