Knowing John's background, I would suspect he may have intended to state
"For RECFM=U, 32760 MAY indeed BE optimum", or perhaps intended to
restrict the remark just to load libraries. I believe all the arguments
in the archives for using block size 32760 with RECFM=U on 3390 were
specifically directed to datasets that are used for Load Libraries.
While this is probably the most common usage for RECFM=U, there is
nothing to disallow other usage of RECFM=U; and those other
applications, unlike program management, might not be smart enough to
utilize shorter blocks to effectively utilize a DASD track.
I agree that it has been established that BLKSIZE 32760 makes sense for
a RECFM=U load library on a 3390. But, a RECFM=U data set that is not a
load library would not necessarily meet the additional requirement of
containing a combination of long and short blocks, which as John states
is a necessary condition for a block size greater than half-track to
have any chance of decent track utilization on a 3390.
Joel C. Ewing
On 08/17/2011 12:54 PM, John Eells wrote:
That really depends on the RECFM and, for variable blocked RECFMs, the
distribution of block sizes. For RECFM=U, 32760 is indeed optimum
(search the archives for why), and for some (not necessarily common)
distributions of variable blocks it might well use space better than
half track blocking. Anything that writes a combination of long and
short blocks can yield surprising results. (We found that z/OS fonts,
for example, use the least space at a counterintuitive block size.)
For plain old sequential FB data, though, it's quite right that 32K is a
poor choice of block size.
Bill Fairchild wrote:
3390 full track size is ca. 56K, and 3380 is ca. 47K. But you were
right - 32K is still not optimal for 3390.
<snip>
--
Joel C. Ewing, Bentonville, AR [email protected]
----------------------------------------------------------------------
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