Shane, Thanks for your thoughts. IBM ETR respones was as follows and several other folks concurred. Luke
I see what you mean. I checked with our hardware support about how PR/SM functions in the case of if SYSA is not using its full share of the box. If the other two LPARs are throttled back to their capped weight, then SYSA gets 83% of the box, and since SYSA has 5 logical engines and the box has 5 physical engines, that means PR/SM dispatches each of those 5 logical engines 83% of the time. However if SYSA is not using its full 83% share, then not all of its 5 logical engines will have work to do 100% of the time. Some of SYSA's logical engines will be in a no-work-wait, while one engine will always be busy with that CICS work. The LPAR is still guaranteed 83% of the box based on weights. If some of its engines are idle, then that one logical engine that has work to do will be allowed to be dispatched to a physical engine more than 83% of the time. Therefore uncapping SYSB will not cause CICS to be throttled back to less than 83% of an engine. Limitting each logical CP to a certain amount happens when the box is maxed out (so LPARs are being throttled back to their WEIGHTs) or whan an LPAR is capped. We suggest that you also reduce the # of logical CPs for SYSB. If you expect SYSB to sometimes use up to 25% of the box, you could give it 3 logical engines (for example) and with that it could still use more than its current weight. Of course I don't know what kind of single-threaded work you might have on SYSB that might cause you to prefer to have more CPs for SYSB. Reducing the number of logical CPs would reduces PR/SM overhead and leaves more physical CPU resource for real work. ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
