On 08/11/2014 04:39 PM, Blaicher, Christopher Y. wrote: > I have not researched this, at all, so this is an educational question. > > From your response to Barry it seems you are saying the Binder will write a > 32K block and then write a short block on a track? If it is doing that, then > how is that any better than a Half-track blocksize? It is still two blocks > per track, or are you saying the binder wastes the remainder of the track? > > Chris Blaicher > Principal Software Engineer, Software Development > Syncsort Incorporated > 50 Tice Boulevard, Woodcliff Lake, NJ 07677 > P: 201-930-8260 | M: 512-627-3803 > E: [email protected] > > -----Original Message----- > From: IBM Mainframe Discussion List [mailto:[email protected]] On > Behalf Of Paul Gilmartin > Sent: Monday, August 11, 2014 5:30 PM > To: [email protected] > Subject: Re: Extents more than One for load modules library > > On Mon, 11 Aug 2014 16:23:31 -0500, Barry Merrill wrote: > >> Educate me why 32k with wasted space on a track is better than half >> track; I do defer to your knowledge and do not argue you are not right, but >> why? >> > The Binder exploits track balances. The "wasted space" you envision rarely > occurs. > > If other utilities were similarly well-designed, 32k would always be optimal. > If QSAM were as well-designed as Binder, 32k would more often be optimal. > > -- gil > > > ________________________________ > > Because of many embedded blocks with control information and varying CSECT sizes load modules always contain a mix of physical block sizes in combinations impossible for mere mortals to predict, and my guess is that some of those smaller blocks aren't split across tracks but may just be displaced to the next track if a potentially large text block earlier in the track is constrained by a smaller block size and splits adding another inter-block gap. A large TEXT block at the end of a track may be split across tracks to fit available remaining space, but not all TEXT blocks are large so such splits do not always occur. Empirical evidence seems to support that 32760 gives slightly fewer total blocks for a module, which translates into slightly better track utilization and slightly better I/O performance. The only down side is that I/O buffers for the loadlib must be larger to reflect the larger max size, but memory is so plentiful these days that's no longer a significant issue.
-- Joel C. Ewing, Bentonville, AR [email protected] ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
