The hardest task in implementing the Auto-Restore function is in determining how much time is allowed to elapse between the time a data set is last referenced and when it becomes a candidate for archival. If this time window is too large, few archives take place and a severe shortage of DASD space probably occurs. On the other hand, if the time window is too small, too many archives take place. If this happens, the cost of archiving and restoring the data sets probably exceed any benefits realized by the temporary DASD space savings.
Obviously there are too many variables to determine an absolute rule in calculating the best time to archive a data set based on last use date. The best method appears to be by trial and error. First, pick a period of time that seems reasonable based on current knowledge of the user environment. Then set up a monitoring period and keep totals on the number of auto- restore requests being processed on a per-shift basis. If this number stays sufficiently low during the trial evaluation, and no shortage of DASD space is evidenced during this period, then you probably have struck upon a good balance between archiving and restoring. Your evaluation period should run through your center's month-end processing cycle to verify that this load does not produce an unreasonable number of restore requests. If it does, your archival scheme probably needs to be re-evaluated.
One issue that must be considered on the archival side is the increased catalog usage due to the need to recatalog data set to the CA Disk pseudo‑volume so they can be automatically restored.
One facility that can help in the maintenance of your catalogs is the specification of sysparm UNCATPSU with a value of Y. This causes both MERGE and IXMAINT to issue an uncatalog (if the data set is cataloged and does not have a F1-DSCB on the source volume) when the last DSNINDEX record for the data set is deleted from the archive index. We highly recommended that this sysparm be specified with a Y if you are using the Auto-Restore function.
The established CA Disk customer faces the problem that all data sets that were archived under an old system of scratch, uncatalog are not accessible by the Auto-Restore function. To assist these customers, a procedure has been implemented to catalog data sets that are found in the archive index but that are no longer cataloged and do not have a F1-DSCB on the source DASD volume. This procedure, named IXCATLG, is supplied with CA Disk and is documented in the IXCATLG Utility section.
|
Copyright © 2015 CA Technologies.
All rights reserved.
|
|