You should select the data sets that benefit from sequential data
striping based on your knowledge of how batch jobs use these data
sets. Use the following criteria to select your candidates for sequential
data striping:
- Select large, physical sequential data sets
The VTOC must
be accessed for each volume that contains a part of the striped data
set. There is a slight performance penalty because of this additional
processing that is not offset for small data sets.
- Select data sets processed using BSAM or QSAM only (no EXCP)
- Select
data sets for jobs that are I/O-bound
Read the RMF™™ Device Activity Report statistics
to find how long the jobs were executing. Volumes with high device
CONNECT times are candidates for sequential data striping. Use caution
in interpreting these statistics. If the high CONNECT time results
from I/O contention because of other data sets on the volume, sequential
data striping cannot improve performance. Therefore, it is beneficial
to have a separate storage group for striping. It is also beneficial
to have volumes in this storage group from as many different storage
controls and serially connected channels as possible. A single storage
control supports up to four paths. Therefore, you can have up to
four volumes per data set per storage control.
- Select data sets used by applications that can coexist with a
system-determined block size
Blocks for striped data sets contain
additional control information. Applications and JCL might require
changes to block size and buffer size specifications to ensure that
striped data sets are efficiently stored on DASD.
Tip: You can use the DFSMS Optimizer Function to help you
select data sets that can benefit from striping.