interesting this topic should come up - i just did an exercise to reblock the TMC, and to add about 400,000 volume records + 150,000 DSNB records on Sunday (in Production).
with BLKSIZE=340, the old copy was 9600 trks with 93% used. with BLKSIZE=8840, i allocated 9000 trks --> used 61% amazing! only hiccup was, the Common Tape Subsystem address space choked (80A-10) when started at REGION=3M. when asked, the regional CA support simply stated to bump it up to 4M, but no calculations offered. - ravi. On Mon, 20 Aug 2007 18:41:40 -0500, Russell Witt <[EMAIL PROTECTED]> wrote: >Tim, > >The main reason to block the TMC is space utilization. If you are currently >allocated as 340 X 340 and you change to the recommended 340 X 8840 you will >end up using only about 1/3 the amount of dasd. > >The reason that 8840 was chosen was that it gives well over 90% utilization >on the DASD track; yet doesn't require a huge amount of ECSA to be reserved. >Also, for direct I/O (read a single record, write a single record) we have >to read and write a block at each time. So, to read a 340-byte record >requires a read of either 8840 bytes of 27880 bytes. Granted, the majority >of I/O time is spent on connecting to the device versus actually >transferring the data; but still do you really want to xfer 27880 bytes >every time you want to read or write a single record. > >We do have some clients runing with 27880 and they report no problems. So, >the choice is yours. But for space utilization, 8840 is "close" to optimum >but only requires a total of 26,520 bytes of ECSA versus 83,640 bytes. > >Strictly your choice, but the only reason for blocking the TMC is improved >space utilization and 8840 will give you a three-fold improvement. > >Russell Witt >CA-1 Level-2 Support Manager ---------------------------------------------------------------------- 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
