SAS/CONNECT User's Guide

Parallel Agent Processing

Given that agents can spawn other agents, independent tasks can be partitioned into subordinate agents. These subordinate agents can be launched by a controlling agent to execute on various nodes within the network. This way each agent can be processed at the same time in parallel with one another.

Using a completion notification message queue, the controlling agent could determine when each of the subordinate agents have finished executing. The controlling agent could also contain logic that synchronizes the work that is performed by the subordinate agents.

Client/server applications can use parallel processing to minimize overall elapsed time of a task's execution. By breaking the application into discrete encapsulated tasks, each independent task can be processed concurrently to reduce the execution time rather than running each task sequentially.

Chapter Contents
Previous
Next
Top of Page