Author: Abid Hasan
Subject: Reply to: Allocate dataset withour BLKSIZE
Posted: Thu Sep 22, 2016 12:33 pm (GMT 5.5)
Hello,
But you do not have 'BLKSZ' specified in the snippet shown.
Nevertheless, REXX or otherwise - block size is for a dataset and it tells the system how/how-much data is to be arranged in storage blocks, so the same rules that apply to your normal JCL (where you'd specify a NEW/CATLG, and other definitions parameters, including/excluding BLKSIZE) apply here as well. If you specify a block-size for a DS, then you force the system to comply to certain 'storage rules', if not calculated correctly for a DS having large number of records (read > a mill. records), and say the DS is processed sequentially in batch, you might end up impacting the EXCPs used, CPU used, might end up increasing the clock-time as well; and vice-versa. General recommendation states, skip coding block-size and let system decide it for you - until and unless you're very sure of the number you want to punch in.
PS: Should you choose to calculate the number, do bear in mind the LRECL/Space allocation/expected-number-of-records for the DS as well; you can read more about block-size in the JCL Reference and DFSMS manuals; more information in 'Using Data Sets' manual.
_________________
Thanks.
Subject: Reply to: Allocate dataset withour BLKSIZE
Posted: Thu Sep 22, 2016 12:33 pm (GMT 5.5)
Hello,
But you do not have 'BLKSZ' specified in the snippet shown.
Nevertheless, REXX or otherwise - block size is for a dataset and it tells the system how/how-much data is to be arranged in storage blocks, so the same rules that apply to your normal JCL (where you'd specify a NEW/CATLG, and other definitions parameters, including/excluding BLKSIZE) apply here as well. If you specify a block-size for a DS, then you force the system to comply to certain 'storage rules', if not calculated correctly for a DS having large number of records (read > a mill. records), and say the DS is processed sequentially in batch, you might end up impacting the EXCPs used, CPU used, might end up increasing the clock-time as well; and vice-versa. General recommendation states, skip coding block-size and let system decide it for you - until and unless you're very sure of the number you want to punch in.
PS: Should you choose to calculate the number, do bear in mind the LRECL/Space allocation/expected-number-of-records for the DS as well; you can read more about block-size in the JCL Reference and DFSMS manuals; more information in 'Using Data Sets' manual.
_________________
Thanks.