As I discovered a few months ago--and discussed here--you must not change the default of VIO=NO, i.e. do not specify VIO=YES. Even if your SMS rules redirect VIO to real DASD as ours did, DFSORT takes the *possibility* of actual VIO to mean that only traditional <=64K track data sets can be used for work files. If DFSORT cannot allocate LARGE data sets, the dynamic allocation algorithms will not function properly for very large sorts.
. . JO.Skip Robinson SCE Infrastructure Technology Services Electric Dragon Team Paddler SHARE MVS Program Co-Manager 626-302-7535 Office 323-715-0595 Mobile [email protected] From: Frank Swarbrick <[email protected]> To: [email protected] Date: 09/10/2012 10:17 AM Subject: Re: Disk data set space allocation, philosophy Sent by: IBM Mainframe Discussion List <[email protected]> My personal philosophy is that computers are better than humans, in general, dealing with these kinds of things. In any case, here are our DFSORT customizations in our ICEPRMxx member: JCL DYNALOC=(SYSDA,12) DYNAUTO=YES MSGDDN=SYSOUT PARMDDN=DFSPARM SMF=NO INV DYNALOC=(SYSDA,12) DYNAUTO=IGNWKDD ERET=ABEND MSGDDN=SORTMSG PARMDDN=SORTPARM SMF=SHORT Specifically, the ones of interest are DYNALOC and DYNAUTO. This allows one to not specify any SORTWKxx DDs. It dynamically allocates 12 of them. We don't override DYNSPC, so it defaults to 256 (megabytes). JCL is used for JCL invoked sorts; INV for program invoked sorts. If it were totally up to me I would have made the JCL options exactly the same as the INV options. We also have the following SMS data classes: DATACLAS AVG SPACE SPACE VOLUME ADDITIONAL DYNVOL NAME AVGREC VALUE PRIMARY SECONDARY COUNT VOLUME AMT COUNT --(2)--- -(8)-- -(9)- -(10)-- --(11)--- -(14)- ---(15)--- -(16)- JUMBO M ----- 3000 1500 --- --------- 9 LARGE M ----- 1000 500 --- --------- 9 MEDIUM M ----- 100 50 --- --------- 9 SMALL M ----- 10 5 --- --------- 9 TINY K ----- 1024 512 --- --------- 9 MINI K ----- 20 20 --- --------- 9 NULL - ----- 0 0 1 --------- -- (Notes: TINY and MINI probably should be reversed, but we had already defined and started using TINY before we decided to have an even smaller one. Also note that even though mini is defined as 20/20 it will always allocate the minimum of 1 track.) It's not perfect, as nothing is, but seems to work satisfactorily. Frank ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
