topic another question about large dataset and computer memory in SAS Procedures

another question about large dataset and computer memory

littlestone — Thu, 16 Feb 2012 15:56:37 GMT

Hello, All

I understand that, in data step, SAS process data record by record; so memory should not be a problem when dealing with large data set.

Now suppose I have a dataset "test" which is 100G, and available memory is 16G; if I want to do some statistics about the data, such as a logistic regression:

PROC LOGISTIC DATA=test;

How does SAS deal with such Procs? still record by record? will memory shortage be a problem in such case?

Thank you in advance for educating me.

another question about large dataset and computer memory

art297 — Thu, 16 Feb 2012 16:18:22 GMT

Memory should only be a problem if the procedure or datastep is using a method that requires such memory, like the hash method.

In other cases, it should only affect performance. However, at least with some of the cluster algorithms, that degredation of performance is "almost" equivalent to non-functionality (e.g., a 2 second task still running after 10 hours).

However, with 16GB, I doubt if you will often confront such issues.

another question about large dataset and computer memory

littlestone — Thu, 16 Feb 2012 16:31:22 GMT

Thank you.

So, you suggest that memory should not be a problem, even if the memory is lower than 16G (say 2 G); only the calculation speed might be lower. Right?

another question about large dataset and computer memory

art297 — Thu, 16 Feb 2012 16:41:32 GMT

Yes, with those few exceptions, at least from my own experience.

another question about large dataset and computer memory

littlestone — Thu, 16 Feb 2012 16:48:50 GMT

I guess in " PROC LOGISTIC DATA=test;" SAS does NOT process data record by record, or does it?

another question about large dataset and computer memory

art297 — Thu, 16 Feb 2012 16:54:38 GMT

I think that the correct answer depends upon the options selected. Take a look at:

http://support.sas.com/documentation/cdl/en/statug/63347/HTML/default/viewer.htm#statug_logistic_sect054.htm

another question about large dataset and computer memory

littlestone — Thu, 16 Feb 2012 17:10:24 GMT

Thank you very much. I will look into the article you recommended.