SAS EG Connection Profile vs Rsubmit

[gphunt]

In SAS Windows Enterprise Guide, is there any difference in performance between running a SAS program on Linux via a Connection Profile vs running it via an rsubmit from within SAS EG vs logging directly into Linux?  (Ver 7.15 HF2)

[SASKiwi]

EG adds a bit of overhead for displaying results (logs and output), but apart from that server performance should be the same.

If your EG network link is slow then there may be delays in displaying results.

[gphunt]

Thanks, SASKiwi,

This is what I hoped and suspected, but I needed an opinion of someone with more knowledge/experience than myself.

Best,

Gary

[nhvdwalt]

Fully agree with @SASKiwi. At the end of the day it's a bunch of Base SAS code being run on a server.

Just out of curiosity, what sparked the question ? All things being equal, the question is not really 'Do I use rsubmit or EG?' These are two totally different interfaces that get selected for different reasons.

[gphunt]

Thanks,nhvdwalt.  I agree, they are two different interfaces used for different reasons.  However, we have one well respected person on out team who vehemently insists that jobs submited via EG gets much worse performance when connecting to Linux than rsubmits (without consideration for internet traffic).

I wonder what happens when the connection drops between EG and the Linux host running the job?

Best,

Gary

[nhvdwalt · Accepted solution]

Working on the assumption that your EG and rsubmit runs code on the same SAS server....

I would start by comparing the options in effect when running through the two interfaces. The easiest would be to compare the output from proc options; run; between the two sessions. It's entirely possible that the Workspace server and the Connect Spawner are configured with different setting, causing a difference in performance. For example...

- They may have different memory settings

- They may have different sorting options

- They may be using different disks for WORK which are not configure the same

- They may be using different disks for UTIL which are not configure the same

- Etc, etc

In addition to the above, I would run the same user code on both and compare the results with -FULLSTIMER switched on. You're dealing with perception here, so you need to break everything down into comparable stats.

Bottom line is this:

Yes, your user may have a different experience between the two, but it's most likely caused by back-end configuration.

Please let us know the outcome if you are able to test the above. 

[gphunt]

Thanks again,nhvdwalt!

These are great ideas.  I'll look into them and let you know what I find.

Best,

Gary