On 15 May 2009 17:24:25 -0700, in bit.listserv.ibm-main you wrote: >Frank, > >I suggest you download the Redbook, VSAM De-Mystified. It will answer your >questions concerning VSAM performance. >As my instructor told the class years ago - Don't take the defaults! >Specify CI sizes that give the best space utilization for your record size. >Choose a large enough Index CI Size to avoid 'dead areas' in your data >component. >Specify a robust value for your Bufferspace parameter, in most cases the >larger the better. (I'm sure someone will volunteer a 'war story' where this >wasn't the case but generally a large buffer space improves throughput.) >For batch processing look into BLSR.
While I did a lot with BLSR, from what I read in the manual there are other parameters available for managing VSAM access that give at least as good results. Check the JCL manual. The constructs may only apply to SMS managed data sets. >For CICS use LSRpools, again be generous. The more data stored in memory the >fewer physical I/O. > >HTH, >Dave O'Brien >NIH Contractor >________________________________________ >From: IBM Mainframe Discussion List [[email protected]] On Behalf Of Frank >Swarbrick [[email protected]] >Sent: Friday, May 15, 2009 6:20 PM >To: [email protected] >Subject: Re: VSE I/O Performance VS MVS [was BLOCK CONTAINS] > >On Fri, 15 May 2009 14:09:33 -0400, Thompson, Steve ><[email protected]> wrote: > >>Because of past conversions, I think this needs to be said: >> >>1) VSE/ESA got to use XA I/O just like MVS. This means, to the VSE shop, >>that some slick stuff that got offloaded to the I/O Subsystem (shall we >>say parts of VM's and MVS' I/O Supervisor code) became available w/o any >>JCL or application coding changes. Things like dual (or multi) pathing >>with dynamic pathing. >> >>What does this have to do with anything? Well, the typical throughput >>performance gains seen in the past when going from VSE to MVS don't >>happen because what was giving those (for the most part) has already >>been realized. >> >>2) VSAM is implemented in VSE differently than in MVS. So, the way >>sharing and buffer management is done changes and WILL cause performance >>issues when you get to MVS. >> >>3) CICS is impacted by these changes, and you may see less throughput. >>Although, with the ability to have more storage than z/VSE allows, you >>may over come it. But be sure to have sufficient page volumes. > >Are you saying the MVS VSAM is "less efficient" then VSE VSAM? Hmmm! One >might start to wonder why we are migrating at all! :-) > >Thanks for the info. I will pass it on to our systems programmers (who >hopefully already know what you are talking about anyway, but it couldn't >hurt). > >Frank > >---------------------------------------------------------------------- >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 > >---------------------------------------------------------------------- >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 ---------------------------------------------------------------------- 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
