Use the following steps as guidelines to help to improve the efficiency
of your XRC operations:
- Allocate journal data sets as single-extent data sets.
If secondary extents are used, the data set must be filled prior to
starting the XRC session if you want XRC to use the full amount of
space that you have allocated for journal use.
- Define an even number of journal data sets, preferably four to
eight for smaller XRC configurations (up to about 750 volumes). For
large XRC configurations, define from eight to 16 journal data sets.
- Define the journal data sets as striped SAM data sets.
When
you define the journal data sets as striped SAM data sets on different
volumes, you enable the system data mover to spread its I/O operations
across multiple volumes and storage controls. Spreading the work load
reduces contention for the same physical device. If you use nonstriped
journal data sets, each data set must reside on a single volume.
- Make journal data sets large enough to allow the system data mover
to offload data to secondary volumes during peak data load periods.
See Note.
- Do not put the control data set on any of the journal data set
volumes.
- Minimize non-XRC activity to the journal data set volumes.
- Do not allocate the journal data sets in the Extended
Address Space (EAS) of Extended Address Volumes (EAV) unless you are
certain that all SDM LPARs that may potentially require access to
the journals (for example, from all potential active and recovery
systems) are at z/OS® V1R11
or higher.
Note: To maximize performance and avoid stall
conditions, allocate each journal data set so that it can hold data
for every buffer that the session might use. This includes the number
of buffers specified in the TotalBuffers value, plus an additional
20% for "Emergency buffers". Also, when determining the number of
tracks required for a journal dataset, allow four journal tracks for
every three buffers to be stored. This translates to a requirement
of 1.6 journal tracks for each buffer. So, for example, an installation
that uses TotalBuffers(25000) should allocate each journal data set
with a total primary allocation of 25,000 * 1.6 = 40,000 tracks.