SAS/CONNECT User's Guide

Considerations for Using Compute Services with RSUBMIT

Using compute services may introduce some effects that conflict with the goals of your distributed application. This section explains these side effects and offers alternatives to eliminate or minimize them.

Compute services require remote CPU cycles. Because the processing takes place on the remote machine, it requires CPU cycles from that machine. This could be a drawback if the goal is to offload work from the remote system.

You can minimize the remote CPU impact by using data transfer services to transfer a copy of the data to the local machine while maintaining the processing on the local CPU or by combining compute services and data transfer services. See Examples of Combining Compute Services and Data Transfer Services for more information.

Compute services also create network traffic in order to return the results of the processing to the local machine. If the processing needs to be repeated frequently, the best solution is to use data transfer services to transfer the necessary data to the local system and maintain the processing in the local environment.

Another effect of using compute services is its demand on remote production data systems as the data is subset or pre-processed. You can minimize this effect by careful implementation of your application to eliminate or to reduce multiple passes of the data.

Chapter Contents
Previous
Next
Top of Page