The Record Selection and Validation program accepts SMF input through only one DD statement. Therefore, it is subject to operating system rules for the concatenation of data sets. This means that all of the SMF data to be processed by CRURRSV must reside on the same device type and all input data sets should have the same block size (to avoid problems with having to put the data set with the largest block size first in the concatenation).

This is likely to be a consideration when dumping the most current SMF data for a recovery situation. Because of the urgency of the catalog recovery, it would be most desirable to use disk data sets and thereby avoid tape mounts.

If your installation maintains some current SMF dump data sets on disk for a period longer than the interval between catalog backups, then the current SMF data sets can be dumped to disk data sets and CRURRSV can run with only disk input.

If any of the SMF data since the last catalog backup is on tape, then you can either dump the current SMF data sets to tape also and run CRURRSV with only tape input, or you can run CRURRSV twice — once with tape input and again with disk input. The "two-run" approach has the distinct disadvantage that the two sets of resulting reports must be analyzed together to see whether all needed SMF data has been included.

Regardless of the technique used to manage the SMF dump data sets, you should be able to identify, for each system, the data set that contains records just prior to the backup date and time, that is, the recovery start date and time, and all data sets covering the subsequent interval through catalog closing or last update. No problem results from the inclusion of extraneous SMF data sets, but the run time for Record Selection and Validation will be extended by the time needed to pass the additional records.