There are several considerations in how the CRUISE cross-reference files are allocated and used. In addition to the need to improving the performance which has been addressed in other RFEs, there are other enhancements that would provide more flexibility and usability.
1. Cross-Reference (XREF) File Initialization
The initialization of the cross-reference files at the start of a restore is a single threaded process. The main driver ECB initializes one ordinal at a time using ‘delete' all full file. This process takes longer and longer as the no. of ordinals is increased to handle larger databases. The initialization should be improved by creating 1 ECB for each ordinal. With multiple ECBS working in parallel through multi-threaded ECBs, the initialization time can be reduced considerably.
2. Stand-alone entry to initialize the Cross-Reference Files
Another beneficial feature would be to provide a stand-alone entry to initialize the cross-reference files. The test system can be readied by initializing the cross-reference files even while the capture is still running on the capture system. This will save time for a large restore. For example:
ZFCRU INIT IALIZE or ZFCRU INIT XREF. Since there is only 1 set of cross-reference files no parameters are necessary. The possibility exists to add the specific file as a parameter, e.g., TOP, FROM or OLDNEW. One advantage of providing the file specific initialization is that some CRUISE functions only use the TOP cross-reference file (e.g. TOP).
3. ZFCRU DISPLAY should show progress during top cross-reference file processing.
During the processing of the FROM file cross-reference files, the display is updated to show total count and the ECB portion of the display shows the no. of ECBs active. However, during the TOP file cross-reference, there is no update of any of the total counts. The ECB display is also zero, but that makes sense since only the driver is processing. When there is a large no. of fixed records to restore, the display is static and there is no obvious indication that progress is made or how much processing remains. The ZFCRU display should update the prime or total count to show many records have been updated. It would be nice if there was also a percentage of the total top cross-reference that had been processed. Another helpful feature would be a distinction in the display (and possibly in the log as well) of whether the restore was running the FROM file-address exchange or the TOP file address exchange. Currently there is no distinction.
4. Cross-Reference (XREF) File Allocation
There are 3 cross-reference files used by CRUISE. The allocation of these files makes a very significant difference in the restore time. The default allocations are adequate for the small restore jobs but performance begins to deteriorate rapidly as the no. of records gets larger and larger. Better documentation and guidelines are needed for the proper sizing and allocation of the cross-reference files. The difference and be very dramatic especially when using a VPARS
IBM does not intend to implement this in the foreseeable future.
Due to processing by IBM, this request was reassigned to have the following updated attributes:
Brand - Servers and Systems Software
Product family - z Systems Software
Product - z/TPF
For recording keeping, the previous attributes were:
Brand - WebSphere
Product family - Transaction Processing
Product - z/TPF