Ricc, It sounds like you may be able to find out a bit more about the IO profile for this dataset by looking at some of the SMF records - type14 and 15 for the access method, and 42 subtype 6 for the breakout of directory and member access.
Then there was the old trick of copying these heavily use members into a PDS in VIO and then referencing the VIO dataset. I'm not sure if that would be helpful here, but it would give each ASID their own in-storage copy of the frequently used members. Ron > -----Original Message----- > From: IBM Mainframe Discussion List [mailto:[email protected]] On Behalf Of > Ricc Harding > Sent: Friday, January 14, 2011 9:39 AM > To: [email protected] > Subject: Re: [IBM-MAIN] DLF - DLFCLASS Using DLF for PDS data? > > Thanks Ron; > > I know DLF does not support a PDS directly, but I did not know if the > application was doing a dynalloc for the pds(mbr) and reading with QSAM > (which the DOC says DLF does support?) The client is only running 3 > standalone LPARS without any shared data so the RETAIN appeared to be an OK > option for that reason and since it is almost never updated. > > Thank you for giving me some keywords to search the archives on. I have > converted the dataset to a PDSE as a first step. I do not know enough about > the application to know if it uses BPAM or QSAM so will have to do more data > gathering to say if it is BLDL activity which could take advantage of the > LLA FROZEN option you mentioned. > > As a side note, and looking at all the things that VLF and DLF do, and do > NOT do, it seems that a very cool vendor product would be one that takes a > dataset and virtualizes it completely, and supports it without any change to > an application (like rewriting it using COFxxx macros :( ). > > I know from working at several software development companies for many > years, that there are already many products that hook both OPEN/CLOSE/EOV > and all of the access method interfaces GET/PUT/READ/WRITE etc so it seems > to be an almost trivial step once that has been done, to move a small data > set into memory and have no physical I/O for it (of course that would be a > read only dataset). More robust support could develop as the need manifest. > (steps down off soapbox). > > Ricc > > ---------------------------------------------------------------------- > 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
