>Can someone please explain why adding the total column isn't yielding a >multiple of 100?
Welcome to Capacity Planning 101. And, I don't mean that disparagingly. This phenomenon has been observed for aeons. There is hardware measurement and there is software measurement. First, the simple one: Hardware measurement: This is extremely accurate. The hardware clock measures wait time and subtracts it from elapsed time to get a consumption figure. Now, the tricky one: Software measurement: The instrumentation does not collect all consumption information, because of overhead concerns. In general, it costs about 8,000 instructions to measure an event. That means that it's not worth the cost to track everything. Why would you measure the CPU for a 500 instruction event? Where is the cut-off? It changes from release to release. And, it varies by job type. Non-swappable jobs have more reported. I/O-intensive have less. It's also gotten better. Used to be just TCB & SRB. Now, we get initiator time, pagein, HIPERSPACE (not as relevent in 64-bit, & more). But, we will never get 100%. This discrepency is called a capture ratio. In other words, the ratio between reported and consumed. This requires a linear regression model of the type: aWKL1+bWKL1+...nWKLn=Hardware clock. It has to be done with many RMF intervals, at a high utilisation. Lumping which jobs into which WKLx is a black art, in itself. I've done it with a pencil, paper, calculator, & an RMF report. With SAS & MXG (or MICS), or EXCEL. It is time-consuming. BTW, when I started a System-Wide Capture Ratio of 70% was considered good. Things have improved to where 80% is good. Your figure is around 83% (249/300). If it's any comfort, that's pretty good. CMG International still has seminars on how to calculate Capture Ratios, after all these years. I guess because newer people discover the diScrepency. Joe Major, now retired from IBM Canada [many years], first wrote about it in November 1981. And, he gave courses one Capacity Planning (I attended the second he ever did), with a major section on Capture Ratios. - Too busy driving to stop for gas! ---------------------------------------------------------------------- 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
