I can't put my hands on detail numbers, but a while back we were running a catalog scan on each of seven sysplexes looking for something-or-other. We noticed that on one sysplex only, the job took 10 or 20 times (!) longer to run than on other systems. A little investigation implicated ENQ/DEQ. This was a parallel sysplex that was still using GRS ring because we had never bothered converting it to star. It had only two members ever, and the second one had not been IPLed in many moons. The GRSplex boundary was identical to the sysplex boundary.
So on a hunch, just for (in practice) a single system, we converted it to GRS star. The timing difference disappeared altogether. . . JO.Skip Robinson Southern California Edison Company SHARE MVS Program Co-Manager 626-302-7535 Office 323-715-0595 Mobile [EMAIL PROTECTED] Tom Rusnak <[EMAIL PROTECTED]> Sent by: IBM Mainframe Discussion List <[email protected]> 06/26/2006 09:59 PM Please respond to IBM Mainframe Discussion List <[email protected]> To [email protected] cc Subject Re: Catalog overhead in a GRS-plex This is to follow up on this thread with some results. After I had changed all my non-shared user catalogs to SHAREOPTIONS(3,3) I was expected so see more relief in ENQ/DEQ overhead but I wasn't getting the results. My catalog performance stats looked like this: -----CATALOG EVENT---- --COUNT-- ---AVERAGE--- Entries to Catalog 2,347K 36.096 MSEC BCS ENQ Shr Sys 4,024K 6.941 MSEC BCS ENQ Excl Sys 76,131 6.868 MSEC BCS DEQ 6,080K 6.638 MSEC I changed my GRSRNL definition so that the SYSIGGV2 for the non-shared usercatalogs were EXCLuded. These catalogs are on packs that are only online to one system so I felt comfortable with making this change. My catalog performance stats after one day are exactly as I had hoped: -----CATALOG EVENT---- --COUNT-- ---AVERAGE--- Entries to Catalog 107,531 9.064 MSEC BCS ENQ Shr 158,160 0.069 MSEC BCS ENQ Shr Sys 17,210 6.434 MSEC BCS ENQ Excl 5,221 0.067 MSEC BCS ENQ Excl Sys 110 0.317 MSEC BCS DEQ 251,153 0.837 MSEC On a side note, Mark had suggested converting the catalogs from ISC to VLF. I've done that and the hit ratio has been 95% on most catalogs as opposed to 40-60% before with ISC. Thanks Mark, and thanks to everyone who replied. Tom <snip> ---------------------------------------------------------------------- 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
