topic Re: SAS Viya Parallel processing in ML in SAS Data Science

SAS Viya Parallel processing in ML

C_Golen — Fri, 10 Feb 2023 07:42:06 GMT

Greetings,

In SAS Viya I learned that there is parallel processing with the help of workers and that I can check its activation by opening a cas session. But my question is at this point: let's say I am running a time series model (Auto-Forecasting). If my workers are active, does this process automatically run in parallel (In model Studio). Or do I have to write CAS code to develop in parallel? How do I understand and manage whether it works in parallel or not?

Thank you in advance for your related resources and detailed explanations!

Re: SAS Viya Parallel processing in ML

sbxkoenk — Fri, 10 Feb 2023 23:32:09 GMT

@C_Golen wrote:
But my question is at this point: let's say I am running a time series model (Auto-Forecasting). If my workers are active, does this process automatically run in parallel (In model Studio)?

Hello,

If you use Model Studio for a Visual Forecasting (VF-) project , then this project runs automatically on all available threads of all available nodes (distributed computing on in-memory data residing in a CASLIB).

To do parallel processing (for time series modelling) in SAS Studio , you need to use an action (from an action set) in PROC CAS.
Or use a CAS-enabled procedure :

PROC ARIMA , PROC SSM , PROC PANEL and PROC MODEL run on SPRE.
PROC CARIMA , PROC CSSM , PROC CPANEL and PROC TSMODEL (SAS Econometrics) run on CAS engine.

In SAS Studio you can easily control the number of workers and the number of threads per worker.
In Model Studio I am not aware of a manner to control the number of workers and the number of threads per worker.

In SAS Studio , you need to initiate a CAS session yourself.

In Model Studio (and Visual Analytics) it is taken care of (you do NOT have to start a CAS session yourself).

Good luck,
Koen

Re: SAS Viya Parallel processing in ML

C_Golen — Sat, 11 Feb 2023 15:34:55 GMT

Thank you for your detailed and explanatory answers! So far it is enough for me.