-------------------------<snip>-------------------------
The PDS directory records are a fixed size, LRECL=255, I believe. My
reason to allocate a PDS filling a track, or multiples, with the
directory comes from thinking half-track blocking is efficient for reads
and if I use one half, or full track directory then I am doing the least
amount of I/O as possible. Reality may be different. The optimum number
of PDS entries to allocate depends on the type of PDS, loadlib or other.
For other, are statistics kept or not. How many of us ignore taking the
time to plan the need and just allocate with a larger number than we
need? I chose to allocate larger based on number of tracks. I no longer
use a PDS for non-loadlib datasets.
--------------------------<unsnip>-----------------------
Ken, the physical BLKSIZE of the directory is 255, plus a 8-byte
physical key. You can set a PDS BLKSIZE to any value you like, but it
will have NO effect on the directory characteristics. On a 3390, that
means that you will always get 45 directory blocks to a track, or 44
blocks plus the EOF mark.
-------------------------<snip>--------------------------
I believe in my testing I found 44 directory blocks will fit on one
track with the end-of-file block following it on the same track. There
will be room for the starting text of one member on the same track. 45
directory blocks will fit on one track with the end-of-file block
spilling over to the second track. For this reason I had been using
multiples of 44 directory blocks when I used a PDS. When you use ISPF to
access a member list, you must read all of the necessary directory
entries to populate the list. 44 means reading two half track blocks,
while 45 means reading three to reach the end-of-file block. That was
from m belief the directory entries fit in the half-track blksize I used
to allocate the PDS.
-----------------------<unsnip>------------------------
You still do the same number, 45 reads (assuming no CCW chaining) for
each track of directory space. See above.
------------------------<snip>--------------------------
Nowadays I would still consider using 44 as my smallest, and if I need
more I would use multiples of 45 plus 44 for the last.
------------------------<unsnip>------------------------
I do the same, partly because I'm just an old-fashioned cuss.
-------------------------<snip>---------------------------
Conclusion - vendors should allocate a directory to fit what they
deliver plus maybe 10% for future growth.
-------------------------<unsnip>-----------------------
Make that 20% and round to a full track and I'll go along with you.
There may be a few poor slobs stuck on physical 3390's yet and in that
case, there are RPS considerations.
----------------------------------------------------------------------
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