Terry (and George) I'm not at all an expert - except that long, long ago I studied QSAM and BSAM very hard indeed on microfiche viewers[1]. I can imagine that PDSE logic sits between largely unchanged QSAM logic and whatever logic corresponds to PDSE support. Thinking about writing, once a QSAM block is full, it will "call" the PDSE logic. It may be that there is a probably slight performance advantage when the number of times the PDSE logic is "called" is minimised. Thus the larger the block size the better.
[1] I was putting together a package which simulated the English Electric System/4 (a 360 clone) DOS operating system I/O assembler interface within MVS. Eventually I had to produce my own version of the System/4 equivalent of QSAM and BSAM using EXCP as the interface to MVS. A further "incidentally": because System/4 appeared to support reading variable length records on tape backwards (according to seemingly available option combinations), I managed to do the same. It's possible that, in reality, it didn't - just as, although similar option combinations are seemingly available, MVS QSAM doesn't - or didn't at the time: the early '70's. Chris Mason ----- Original Message ----- From: "Terry Sambrooks" <[EMAIL PROTECTED]> Newsgroups: bit.listserv.ibm-main To: <[email protected]> Sent: Tuesday, 03 October, 2006 8:43 AM Subject: Re: Default System BLKSIZE for PDSE > Hi George, > > With regard too: > > "....I'm curious if its fine to take the default of 32720 even though its not using half track blocking like the PDS. Is it safe going with the default or am I wasting a bunch of space? Does it matter with PDSE's since they are radand written to differently? " > > Rather than not worry about it, it may be worth reflecting on the statement about performance that is mentioned in the manual. The manual indicates that whilst the BLKSIZE is not used to control the structure of the data itself, as data is record in 4K chunks, the BLKSIZE does come into play with performance as it is used to set buffer sizes. (At least that was my interpretation.) > > On this basis 32760, being closer to 32K than 27920, may well be better although I have not done any measurement to prove it. > > Kind regards - Terry > > Terry Sambrooks ---------------------------------------------------------------------- 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
