Dear Readers,

I get a lot of emails from SAS Tech Report readers, and in December, I received a couple of emails that I want to share with you because they give you more information and make your work easier.

The first email was from Mike Zdeb, from the University at Albany School of Public Health, Rensselaer, New York. He noted that the method used in a SAS Sample in the Tips & Techniques section (Counting the Number of Missing and Non-Missing Values for Each Variable in a Data Set) "seems overly complicated."

Zdeb wrote a paper for NESUG 2011 where he conquered this problem. He believes that his code is a lot easier to follow.  He says there are “none of those multiple ampersands: there are seven occurrences in the SAS support code. And the output is pretty nice too – the SAS code in the paper uses PROC FREQ just once and gets all of the counts in one pass." [The SAS support code makes multiple passes through PROC FREQ (once per variable).] 

The second email was also very helpful. It came from Mike O'Neil, Manager of the Data Warehouse for the Ministry of Social Development, Wellington, New Zealand. O'Neil wrote regarding Andrea Wainwright-Zimmerman's SAS Global Forum paper, While You Were Sleeping, SAS® Was Hard at Work .

“There are a number of techniques presented in this paper [that] are perfectly appropriate for a small organisation using SAS or a single SAS user managing their own workflow. For a large IT installation where there are dependencies on completion of non-SAS processing or other events such as file arrival, and where there is a considerable stream of SAS processing that has its own set of dependencies, the techniques described do not scale. 

The ability to re-start processing part way through a processing stream of many tasks, or place some job streams on hold because of a failure elsewhere, is an important part of production control.

We are a large Unix-based IT shop where more than 60 percent of the entire IO on the disk arrays is generated by data warehouse processing or user processes accessing the warehouse data, and we use Control-M to schedule more than 500 processing tasks, most of which run daily. Our biggest headache is user-processing scheduled by them using the Unix ‘at’ command. We encourage users to use Control-M as soon as they know the job will run regularly.

The techniques described in this paper can be improved to make them more 'fit for IT.' SAS users and Administrators at sites need to be aware that these techniques do not scale."

Thank you Mike and Mike, this is great information. If you have information like this, please send it to me to share with our SAS colleagues.