We were also having a lot of problems with the large volumes of SMF data we produce (though it was CICS data we us) and associated contention issues. What we did was increase the size of the MANx datasets so they could never fill them in 30 mins and have our DUMPXY's run every 30 mins to clear them out. The DUMPXYs write to system specific GDGs, creating a +1 every half an hour. At end of day, we run a job that mods all the GDGs to a single cartridge as one large system specific file and then another job read in the cartridges datsets for all our LPAR's and writes them back to DASD, splitting them into CICS, DB2, etc GDGs for further processing...
The benefits are... 1) We don;t have to process those large volumes of CICS records to look at RMF records etc 2) The GDG's created every half an hour mean that these can be read on the day they were created without impacting SMF collection (by referencing them using the absolute generation). Very handy for problem determination. 3) Also, these half an hour GDGs mean it's very easy to resolve any invalid records caused by space problems etc (not that any one should get space issues in this day and age! but...). Our daily file used to be some where in the region of 15,000 cyls and running a sort to pick out one dodgy record was a long winded operation to say the least! 4) Using REXX to find out the absolute GDG names and using that info as input to our nightly dump job means we can backup yesterdays data without impacting the collection process... Cheers, Duncan ---------------------------------------------------------------------- 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