As with almost everything else, the answer is it depends. For "large" sequential datasets as you described, I agree.
But consider a card image PDS (JCL, source, etc) where most members average around 400 records. Half block on a 3390 is 349 records. The first member will occupy 2 blocks on the first track. The second member's first block will not fit on the track. The net result is that 300 records worth of space on track 1 is wasted. With a blocksize of 3120, 15 blocks fit on a track for a total of 585 records. Almost a 50% improvement. My preference is 6160 (reasonably efficient for both 3390 and 3380 - yes we have very old archived datasets which occasionally must be restored) which will allow 616 records per track. We previously had a similar discussion for load modules where 32760 is optimal for the Binder but not necessarily for IEBCOPY. :>: -----Original Message----- :>: From: IBM Mainframe Discussion List [mailto:IBM-MAIN@LISTSERV.UA.EDU] On :>: Behalf Of Eric Bielefeld :>: Sent: Tuesday, July 23, 2013 10:21 AM :>: To: IBM-MAIN@LISTSERV.UA.EDU :>: Subject: Re: BLKSIZE=3120 :>: :>: I believe that the net result of coding smaller blocksizes does result :>: in :>: being able to store less data. If you had 1,000 volumes all defined as :>: 3390-9s, and each volume had 100 datasets that filled the volume blocked :>: at :>: 512 bytes, you would store a fraction of the data if you blocked each of :>: those datasets at 1/2 track blocking. That is a function of the z/OS :>: archictecture. :>: :>: I don't know exactly how the data is stored on the tracks, but I believe :>: that the result of smaller blocksizes means that you will store a lot :>: less :>: data. ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
