Hi folks!! Thanks a lot for your answers. I'm learning a lot with you.
At DB2-list (http://www.idugdb2-l.org/archives/db2-l.html) I found following discussion: ********************************************************************** I read the value 100 in a Chuck Hoover's article: 'If the sum of these page (read+write page-ins - my note) counters divided by the time period of the data recording cycle is greater than 100 per second, and expanded storage is not over taxed by non- DB2 work...' He said that in this case pools are overallocated and it may be useful to create hiperpools. Besides in a CITL (1999) about BP tuning document is suggested: 'As a rule of tomb (ROT), a paging rate below 500/sec per engine (CP) is acceptable' (this is a SYSTEM paging rate however)' It was missing from my mail that 100 page ins/sec is for SYSTEM paging rate, not for BPs nor DB2 subsystem, as we was discussing about how to check some MVS parameters. On the other hand I read in your article (IDUG 2000) ' Buffer Pools & DASD Subsystem - Analysis & Tuning' the phrase 'If this (paging rate per sec for all BPs - my note) is consistentlygreater than 10' which is relative to ALL BPs (not to DB2 subsystem) and of course is correct. ********************************************************************** The old 500 pages/sec times the number of engines has been out there for decades.. About a dozen yrs ago I was in an IBM MVS tuning clas, and the instructor added... except when you are running DB2 - then it should be less than 250 pages/sec per engine. These nembers were an absolute max, beyond which the system was really dead.... However, my observations over the years have shown than when the total system paging rate gets higher than 200-300, all work in the system is being impacted by paging - to various degrees. Then you have to determine and evaluate the paging rate impacts, versus the saving from I/O delay and the cpu saving by not doing an I/O. ********************************************************************** I absolute agree. I observed the same, even if I saw problems with a value of paging rate slightly greater than 100 (say 150) but there were other problems in that system. When the rate is > 200 then problems begin.... So I think that 100 pageins/sec as SYSTEM paging rate can be used as a ROT in the evaluation you suggest in your reply. A value of 10 pageins/sec for all BPs is the max you can admit for BP tuning. ********************************************************************** The end ********************************************************************** Last week, I assisted to MVS Performance course and instructor says that interval of 300-500 pag/sec was the recommendation in 1991. Also he told us that is possible that this values could be true today but we think that it'll be too high. Maybe, 50 pag/sec could be a good value? If MVS is Multiple Virtual Storage... paging should be used, not? ---------------------------------------------------------------------- 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
