You can use Storage Class Memory(SCM) with a CF structure to give more capacity in the structure. Think of the days when you were young and systems had a limited amount of real storage, but a lot of virtual pagable storage.
See Why use SCM? in the MQ knowledge center for more information.
MP16: Capacity Planning & Tuning Guide - IBM has a good section on SCM - in particular the use of "Non-sequential gets from deep shared queue" and the "example use cases for IBM MQ with SCM"..
What should I look at when using System Class memory with Coupling facilities?
SCM technology can offload data from a CF structure and so increase the size of the structure. If a message is needed from the structure, and it is not in the CF storage "a page fault is generated".
When the processing is sequential such as get next, the data has usually been prefetched from SCM into the CF.
If the access is random, such as get by msgsid, then there is a chance that the data is not in the structure and has to be fetched from SCM, think of it as a page fault.
The RMF report DELAYED FAULTS %ALL Percentage of delayed faults in relation to all requests for the structure.
The RMF data below is for a one minute period with random message access in the queue ( so a deliberately "bad" behaviour")
The read and write rates are similar in value and we see 26.2% of requests delayed - or 73.8% not delayed - but still reading 80GB ( 870718 MB) in a 60 second interval..
Look at the right hand side. DELAYED FAULTS CNT 45029) and %ALL (26.2)
During the time interval 26.2% of the requests were delayed. This is calculated as number_of_delayed_page_faults / ( synchronous + asynchronous CF requests)
There were 45029 page faults. From another part of the report there were 171866 synchronous + asynchronous CF requests, this gives 26.2%
A large percentage does not indicate a problem, but to reduce it you could do
- Reduce the number of messages on the queue
- Avoid random gets from deep queues
- Add more storage to the Structure.
With thanks to Tony Sharkey who captured the data, and explained things to me.