Performance Considerations › System Generation Parameters › Allocating Primary and Secondary Storage Pools

Allocating Primary and Secondary Storage Pools

Record Buffer Block

The runtime system allocates and initializes record buffers for use by executing dialogs. When an application is initiated, CA ADS allocates a Record Buffer Block (RBB) from the DC/UCF storage pool to hold the subschema, map, and work records accessed by the dialogs in the application thread. The RBB must be large enough to accommodate the largest of these records.

There is one primary RBB for each application. CA ADS allocates a secondary RBB when the RBB becomes full during execution of the application or does not have enough remaining space to hold a record. Additional secondary RBBs can be allocated by the CA ADS runtime system as necessary.

The data communications administrator (DCA) can specify the size of the primary and secondary RBBs with the PRIMARY POOL and SECONDARY POOL parameters of the ADSO statement. When allocating the primary and secondary storage pools, the DCA needs to consider the size and number of the records used by the application as well as the header records maintained by the buffers.

The primary RBB should be large enough to satisfy the records associated with the most-frequently used dialogs. The secondary RBB should be large enough to accommodate the largest record.

Alternatively, the runtime system can be directed to calculate the size of the RBBs for an application or dialog and to use the calculated size when acquiring storage space for the RBBs. Use the calculated size in systems where there is a high number of records and storage space is a concern.

The following diagram shows the structure of the Record Buffer Block. Each record buffer contains a 24-byte header to keep track of available space. For each record in the pool, CA ADS maintains a record header (RBE) that requires 44 bytes of storage for database records and 56 bytes of storage for logical records. There is also a header for each element of a logical record. All records and headers are aligned on doubleword boundaries. Each buffer pool must be large enough to accommodate the largest subschema, map, work, database, or logical record used by a dialog in the application.