Configuration statements and connection attributes. Define certain DBMOVER statements and PWX Batch connection attributes.

Connection pooling. Network connection information is cached and reused, which decreases the connection time for subsequent connections. For example, opening a connection to a PowerExchange Listener on

z/OS

might take two elapsed seconds. With connection pooling, subsequent connections to the PowerExchange Listener take a fraction of a second.

Data maps caching. PowerExchange retrieves data maps defined for z/OS nonrelational data sources from

job-level

memory rather than by accessing the data maps file. By eliminating accesses to the data maps file, data maps caching improves performance.

Offload processing. For DB2 for z/OS tables and image copies, IMS unload data sets, VSAM data sets, and sequential files, use offload processing to distribute PowerExchange bulk data column-level processing and filtering to the PowerCenter Integration Service machine that runs the bulk data movement session. By distributing processing to another machine, you reduce PowerExchange bulk data movement overhead on the source system.

Multithreaded processing. If you use bulk data offload processing for DB2 for

z/OS

tables, VSAM data sets, or sequential files, you can also use multithreaded processing to attempt to increase throughput. Multithreaded processing uses multiple threads on the PowerCenter Integration Service machine to perform offloaded PowerExchange processing.

Pipeline partitioning. Use multiple partitions in a pipeline for a bulk data movement session to improve session performance. You can use reader partitioning, writer partitioning, or both. Partitioning enables the session to process data in the partitions concurrently, which is especially beneficial for resource-intensive, column-level processing that is offloaded to the PowerCenter Integration Service machine.