EG vs Base SAS

[tommy81]

Hi ,

While comparing exection times on both EG and BASE SAS , iits been noticed that EG performs pathetically.. execution real time is many times that of BAS SAS.

Please could someone help me tweak or suggest methods to improve performance

[jklaverstijn]

This is a bold statement indeed. Before anything else could you please substantiate this? Examples, benchmarks, comparisons, code, ...? Give people something to respond to.

Jan

[DBailey]

I've noticed a few things:
1) The EG query wizard produces proc sql. Sometimes that seems slower than a data step for some kinds queries but some kinds of joins are best served through proc sql.

2)Since the query wizard uses proc sql, it doesn't require data sets to be sorted prior to running. data step requires sorting. If you sort the datasets, I would suspect similar run times. Message was edited by: DBailey

[ChrisHemedinger]

EG also generates richer ODS output (like SAS Report or HTML) than Base SAS does by default. You can tweak that in Tools->Options->Results.

If your Real Time numbers in the SAS log are higher from EG than in Base SAS, it's worth a track with tech support to see what might be different.

Chris

[Doc_Duke]

One thing that is documented to decrease performance on EGuide is if you just open datasets into EGuide, the data ends up getting read twice. This can really add if it is large.

Use a LIBNAME for your datasets to get more comparable performance.

See this video from Chris on the difference
http://www.youtube.com/watch?v=OSTa1EUpKT8

Doc Muhlbaier
Duke

[jklaverstijn]

I see lots of guessing. I suggest the OP provides us with some real-life cases so we can offer help. Otherwise we're in the dark. All comments are true in themselves but do not explain "pathetic" performance. EG is my daily tool and the increased productivity outweighs any performance penalty one might have in comparison to DMS / plain code. Nothing pathetic about EG.

But again, please stop providing answers when the question is not clear. For lucky shots there is plenty occasion at SGF11.

- Jan.

[DBailey]

huh?

Not sure why it matters to you why I choose to suggest possible issues comparing EG to base sas.

I think most of us here try to be helpful and civil.

[jklaverstijn]

Sorry for appearing rude. Maybe I unintentionally sounded less civil (could be because I'm Dutch) but I was certainly trying to be helpful. It was my assesment that the OP could help us help him by providing more information. The adjective "pathetic" suggests there must be some powerful evidence yet to be disclosed.

Consider me the doctor that will not just try to prescribe some medicine before doing some q&a and poking in the sore spots. In a very civil fashion.

Regards, Jan.

[DBailey]

In the spirit of international detente...

Ik verontschuldig me voor het lezen van te veel in je reactie. Ik ben het eens dat een aantal voorbeelden kunnen ons helpen om het opsporen van de OP's problemen. Alles is goed. Message was edited by: DBailey

[tommy81]

Hi All,

Thankyou for the comments.

Infact I`m in no way trying to undermine the power of EG as a tool.

Its the time taken to execute a query , thats the point of contention .

Also please note that the execution is happening from a remote location.

[nrose]

Hi,

If you are accessing data from a remote server, and processing a query locally using SAS EG then SAS EG has to copy all the data to the local session before it can process it. Perhaps this is the problem.

Nick

[jklaverstijn]

🙂 actually very readable Dutch. I'm impressed.

[SASKiwi]

How about exporting the code of a sample EG process which you have already benchmarked in EG, then run the exported code in base SAS? If it performs the same as in EG, then it is not the EG environment that is the issue but the code itself.

A comparison of techniques with your different base SAS code may provide the answers.

[TomKari]

I don't know if these relate to this issue, but here are two scenarios that I have found result in very slow processing for EG users, don't really relate to SAS or EG, and are very easy to fall into for new users.

1. Moving large amounts of data between machines. This can be because EG submits code to a server that needs to access data on another server, or because EG runs the SAS session locally, and the local computer pulls data from a server. The SAS session is actually doing next to nothing, just ticking along waiting for the data communication.

To diagnose this, if you can use performance monitor look for very high network card volume. Another option is to move the data to the machine that the session will run on, and then run EG against it.

2. Running an EG descriptive task (e.g. Summary Statistics, Summary Tables, Table Analysis) with either a very large (tens of thousands to millions) number of combinations of classification variables, or using a variable with very high cardinality as a classification variable. SAS will actually run this very quickly, but if you're requesting results in one of the viewable formats, rendering the formatted document will take hours, and may never return. HTML is (I think) the worst.

To diagnose this, set the descriptive task to save the statistics to a dataset and turn off all of the viewable result formats. EG may run very quickly.

[DBailey]

translate.google.com is my friend