Use multithreaded processing when the PWX reader thread of a bulk data movement session uses 100% of a CPU on a multi-CPU server on the PowerCenter Integration Service machine. In this situation, multithreaded processing can help improve session throughput by spreading bulk data processing across multiple threads. However, if the single CPU is not fully consumed, additional threads do not improve throughput.

If the network processing between the source system and PowerCenter Integration Service machine is slow, specify 1 for the

Worker Threads

connection attribute to help improve throughput. When you use 1 or more worker threads, PowerExchange overlaps network processing with bulk data processing on the PowerCenter Integration Service machine.

For optimal performance, the value for the

Worker Threads

attribute cannot exceed the number of installed or available processors on the PowerCenter Integration Service machine.

You cannot use multithreaded processing for sequential-file and flat-file data maps that specify

VS

for the

Variable

property in the SEQ Access Method dialog box.

To use multithreaded processing on Linux, UNIX, or Windows for z/OS sequential files that have variable length records, ensure that the file transfer mechanism does not remove the record descriptors. Each record must be prefixed with the record length.

You cannot use multiple worker threads to write data to targets. Instead, consider using writer partitioning if the data contains only Inserts.

If you use reader or writer partitioning for a bulk data movement session, the

Worker Threads

setting is ignored.

Not all PWX NRDB Batch connection types support worker threads. If the

Worker Threads

connection attribute for one of these connections is set to a nonzero value, processing continues without threads.

Message PWX-31505 reports connection performance settings, including the number or worker threads. If worker threads are not used, message PWX-31505 reports a zero value for worker threads.