>- Software MSUs are in really Marketing MSU's, in fact any other name >than MSU would make the situation more clear. >- Hardware MSUs are really hardware configuration denpendent, which >means that any variation in hardware will change the SU conversion >factor. Vary a CPU online and you have a different amount of MP overhead >and therefor a different SU factor, which is indeed per processor as you >discovered above.
I heartily dislike both MSUs and MIPS in any shape, form or size, as it does not translate into something I can comprehend. Having said that, I used yesterday's data from one of our boxes where the current grafic from the type70PR records shows 100% cpu usage (actually, 99,8%) on the box. Adding up all the cpu consumed during each of those intervals gets me a number pretty close to 3600s. That answers the big question I had - how many cpu seconds can I achieve in a 10-minute interval on 6 processors? Everybody who said it's equal to wall clock time times cps was right, and I was just to dense to understand. But this means (to my way of thinking) that 1 cpu second on a slowed down processor model is not equal to 1 cpu second on a not-slowed down processor model. In future, I'll be wary when I hear that a job used so-and-so much cpu seconds. Now I also understand (I think) why service units were invented. So, in a grafic showing the cpu seconds consumed, my 'capacity line' will be no.of.cps*10min for a ten-minute interval. And if CoD is used again, management can *see* where we cross that line because we had varied more processors online and how many cpu seconds were consumed beyond our 'normal' capacity. Thanks for your patience, Barbara ---------------------------------------------------------------------- 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
