Jerry, MSU settings (LPAR or group) are enforced by software. LPAR weights (LPAR level only) and absolute capping (at the LPAR or group level) are enforced by the hardware/firmware itself.
Weights apply only to active LPARs and control the LPAR's portion of shared processors. In your example, if all 4 LPARs are active and have weights equal to their initial values (30, 10, 10, 30), then LPAR #1 can use 30/80 or about 37.5% of the shared processors. If LPAR #1 doesn't have capping explicitly turned on, then if LPARs #2-4 aren't using their portion of resources, then LPAR #1 could get more. Conversely, if you wind up with only LPAR #1 active, then LPAR #1 can use 30/30 of the processors or 100% of the system. If you don't use software management (like WLM), then you can leave the defined capacity (which is == MSUs) set to 0. If you want LPARs #1 and #4 to not be capped, then make sure the image profile is set up to not cap (and you can check/change via "Change LPAR Controls" on the HMC). If you want LPAR #2 to be capped, then you can turn on capping in the image profile. But, what type of capping do you want? Are you trying to ensure you don't exceed some license agreement for processing power in LPAR #2? If so, then you should instead use "absolute capping". You could either set an absolute cap limit on that specific LPAR or if appropriate, you could set an absolute cap limit on an LPAR group if you want a group of LPARs to be limited in total. For example, maybe you are licensed to only run at most 2 processors worth of power in a given LPAR and no more than say 5 processors worth of power for the entire machine. In that case, you could set each LPAR with this restriction up be capped at 2 processors and then put these LPARs into a group with an absolute cap of 5. On the other hand, if you just want to ensure that LPAR #2 doesn't use too much power (perhaps because it is used for test purposes), then simply capping in the image might be enough because you presumably wouldn't care if the test LPAR used all of the processing power if it happened that it was the only LPAR that was active. The "Processor Resource/Systems Manager Planning Guide" (e.g. SB10-7175) has a lot more details on how LPAR weights and capping and the like work. Still, I know that it is complicated, particularly as there are a number of ways that an LPAR can wind up being capped, either explicitly or implicitly. ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
