The BLOCKSET mode is a performance feature for customers with large and active audit data sets. In this mode, records in the audit data set are allocated by blocks instead of records. Each allocated block is then assigned to an LPAR for its exclusive use. The assignment greatly reduces the number of RESERVEs and updates of the audit control record 4. It is not necessary to read a block before adding a record. The block is written each time that a record is added but the buffer is retained by the LPAR. It is not necessary to reread the block. After the LPAR has written all the records in the block, the LPAR will allocate another block with a RESERVE and update of control record 4. When an LPAR allocates a block, the next available record pointer in control record 4 is incremented by the number of records in a block.
The new control record 5 is the first record in a BLOCKSET audit data set. Control record 4 is now the second record. The remainder of the first block is padded with dummy audit exception records. Control record 5 contains a table of up to 32 LPARs, each LPAR with the record number of the block that is allocated to it. Control record 5 also contains a timestamp of when a block was last allocated.
Each time a block is allocated, a check is made for the audit threshold and full conditions. When the physical end of the audit data set is reached, it will wrap around to the second block in the data set. Only control records occupy the first block. When calculating percent used, all the records in a block are consider used. The blocks for the control records are not counted.
|
Copyright © 2013 CA Technologies.
All rights reserved.
|
|