Speeding Up SAS Viya Jobs with Compute Tasks

4 Likes

If you have migrated simple stored processes from SAS 9.4, you may have noticed that SAS Viya jobs don’t always execute as quickly as they did in SAS Enterprise Guide. This has to do with the underlying architecture used by SAS Viya, where a new job request causes a compute pod dedicated solely to that job to be spun up. While this has a lot of benefits for longer running jobs in terms of resilience and stability, there may be times when you just need results fast. In this article, I’ll discuss how to take advantage of compute tasks, a new feature introduced in 2025.9, to accelerate the execution of short duration jobs in SAS Viya.

What are Compute Tasks?

Compute Tasks are designed to be a lightweight alternative to Job Execution Service. When using Compute Tasks, you can submit a program and receive its output in a single HTTP request. This reduces the unnecessary performance impact created when using the Job Execution Service.

Requirements for Using Compute Tasks

Compute Tasks require a reusable compute server. Generally, a compute server terminates when the session that was being used by the client or application is ended. A reusable compute server does not end, and can be reused by a later session. You find more information about how to configure a reusable compute server here.

Compute Tasks are also not currently supported in SAS Studio. To execute jobs using the Compute Tasks, you will need to use the SAS Job Execution web application or via the URL.

Configuring a Job to Run Using Compute Tasks

Once you have a reusable compute context set up, configuring a job to run as a compute task is very straightforward. There is a new job definition parameter called _executionTasks that you can use to specify whether you want to execute the job as a compute task. Set _executionTasks=True to run the job as a compute task.

Select any image to see a larger version.
Mobile users: To view the images, select the "Full" version at the bottom of the page.

Putting It All Together

Let’s look at an example. I have created two simple job definitions that use the following code:

proc print data=sashelp.class;
run;

One uses the SAS Studio compute context and executes as a normal job using the Job Execution Service. The other uses a reusable compute context and has the _executionTasks parameter set to true.

The standard job takes about 7.5 seconds to complete. The actual code executes very quickly, but there is some overhead that comes with starting up a new compute server.

However, the job using Compute Tasks executes extremely fast: less than 500 milliseconds.

Considerations

While Compute Tasks can dramatically increase the speed of short duration jobs, there are a few things to keep in mind:

It is best practice to execute long-running jobs or large files using the Job Execution service rather than using Compute Tasks.
Compute Tasks time out after a maximum of 30 seconds.
Compute Tasks are not listed on the Jobs page.
Output files cannot be downloaded.
Compute Tasks cannot be scheduled – they occur synchronously.

Conclusion

Compute Tasks offer a powerful new way to accelerate simple job execution in SAS Viya. By reducing the startup overhead associated with the Job Execution Service, they provide near-instant responsiveness for short programs that don’t require long-running sessions, scheduled execution, or file downloads. While they aren’t a replacement for every job use case, they fill an important gap for workloads where speed matters more than robustness or complex runtime features. If you’ve migrated stored processes from SAS 9.4 or have lightweight code that needs to run quickly and repeatedly, Compute Tasks give you a fast, efficient, and easy-to-implement option within SAS Viya.