Tuning the size of staging data sets

Important

This section assumes you are using log streams that use coupling facility structures. For related information about DASD-only log streams, see Tuning for DASD-only logging.

MVS™ keeps a second copy of data written to the coupling facility in a data space, for use when rebuilding a coupling facility in the event of an error. This is satisfactory as long as the coupling facility is failure-independent (in a separate CPC and non-volatile) from MVS.

Where the coupling facility is in the same CPC, or uses volatile storage, the MVS system logger supports staging data sets for copies of log stream data that would otherwise be vulnerable to failures that impact both the coupling facility and the MVS images.

Elements (groups of log records) are written to staging data sets in blocks of 4K bytes (not in 256-byte or 512-byte units as for log stream data sets).

Recommendations

Use the following formulae to help you tune the size of your staging data sets:

staging data set size= (NR * AVGBUFSIZE rounded up to next unit of 4096)

where NR is the number of records to fill the coupling facility structure. This can be calculated as follows:

NR = coupling facility structure size / (AVGBUFSIZE rounded up to next element)

Ensure that the coupling facility structure and staging data set can hold the same number of records. Staging data sets are subject to the same offloading thresholds as log streams are. It is sensible, therefore, to ensure as far as possible that offloading activity will be at the same frequency.

You are recommended to overestimate, rather than underestimate, staging data set size. To calculate staging data set size to accommodate the maximum number of records (where there is one record per element), use the following formulae:

maximum staging data set size = 8 * coupling facility structure size

where element size is 512 bytes, and

maximum staging data set size = 16 * coupling facility structure size

where element size is 256 bytes.

Investigate using DASD FastWrite facilities with a view to storing data in the DASD cache, as opposed to writing it directly to the staging data set. This also enables a faster retrieval of data should it be required. Be aware, however, that if you fill the cache, data is also then written out to the staging data set whenever data is written to the cache.

Related tasks
Logging and journaling: performance considerations
Monitoring the logger environment
Performance implications of average blocksize
Performance implications of the number of log streams in the coupling facility structure
Setting LOWOFFLOAD and HIGHOFFLOAD parameters on log stream definition
Setting the activity keypoint frequency (AKPFREQ)
Specifying the log defer interval (LGDFINT)
Tuning for DASD-only logging
[[ Contents Previous Page | Next Page Index ]]