>Why worry? >RMF III and RMF post processor reports will both give you the percentile breakdown of the activity (given a percentile goal). Just adjust the duration of >750 SU's until A) 75% of transactions end in 1/2 second or less and then use that value or B) if 75% of trans are ending wthin the approved duration(750 >or other), adjust the goal to match the observed response time E.G. 75% completion in 0.3 sec (or other).
The only problem with that approach is that it presumes that the behavior is already acceptable and you're simply trying to calibrate the goal to the actual achievement. The point in examining some actual values is to provide additional insight into the actual behavior of the workload being managed. I still maintain that 75% is far too liberal a goal for first period TSO. Given the natural constraint of having a duration coded, there shouldn't be as much "spread" in the load as alluded to in the original post. "Its current definition is 50% complete in 1 second and I've got *one* 4hr sample of 65% at 0.5 sec and 27% at >4 sec." In my opinion, these are two distinct workloads. A 800% differential should never occur for comparable units of work (within the same goal and performance period). This indicates (to me) that the duration is allowing longer running transactions to be retained in first period and skew the results which is why a percentile of 75% is necessary. I believe the duration should be reduced to tighten up the first period controls and that the percentile should also be tightened up. This would force the longer running transactions into second period where they can be managed with comparably running units of work rather than usurping resources that should be dedicated to trivial transactions. There's nothing wrong with having extra performance periods to accommodate such situations, since there is nothing special about having only three performance periods for TSO. The point in multiple performance periods is to allow different types of work to migrate to their respective levels of consumption so that they can be managed similarly. I don't know where the idea came from that somehow a certain percentage HAD to complete in any period to be considered "good". Anyway ... My two cents Adam ---------------------------------------------------------------------- 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
