IBM has provided a feature
to improve VSAM performance, called Batch
LSR (Local Shared Resources). This provides
the ability to keep even the sequence set
of the VSAM data set in storage, and thus
further minimises IOs. It provides to batch
jobs the VSAM pooling features that have
been available for a long time in IMS and
CICS. It also provides the features of
deferred writes, and sequential read ahead.
Batch LSR needs to be switched on by including
a line in
SYS1.PARMLIB(IEFSSNxx)
To implement Batch LSR for your VSAM job,
you need a JCL change:
instead of
//DATASET DD DISP=SHR,DSN=MY.DATA.SET
switch to
//DATASET DD SUBSYS=(BLSR,'DDNAME=NEWDD','BUFND=dd','BUFNI=ii')
//NEWDD DD DISP=SHR,DSN=MY.DATA.SET
Where ii is the number of Index buffers
and DD is the number of data buffers required.
Ten index buffers and 5 data buffers is
a good start. Hiperspace buffers can also
be specified if necessary in special circumstances
(BUFHNI=n, BUFHND=n).
There are also a number of proprietary packages
that will switch on LSR for you.
Next z/OS Tip |
