The SAS IT Resource Management aggregation transformation makes use of PROC RANK to compute requested rank columns. PROC RANK is a memory-intensive procedure. The memory requirements for this procedure are documented in SAS Note 7543.

Problem Description

It is possible that a summarized aggregation could be specified in such a manner that the resulting output table has a large number of observations. In such a circumstance, the memory needed by PROC RANK may not be available.

Symtoms of this problem are error messages in the SAS log that are similar to the following (note that the amount of memory or the name of the dataset will differ):

   ERROR: Unable to allocate sufficient memory. At least 279664K 
          bytes were requested. You must either increase the amount 
          of memory available, or approach the problem differently.
   NOTE:  The SAS System stopped processing this step because of 
          insufficient memory.
   NOTE:  SAS set option OBS=0 and will continue to check statements. 
          This may cause NOTE: No observations in data set.
   WARNING: The data set WORK.RANK0_AaaaaaaG_AbbbbbbT may be 
          incomplete.  When this step was stopped there were 0
          observations and 0 variables.
   NOTE: PROCEDURE RANK used (Total process time):

This error will cause the aggregation transformation to exit and clean up any intermediate data sets. Following this error in the log, another error message will also be displayed, indicating that the macro that contains the aggregation code has ended with an error:

   ERROR: Compiled macro failed in transform 
          (AaaaaaaG.AbbbbbbJM). Trans_rc = 8.

This is followed by a final error message, indicating that the aggregation transformation has ended with an error:

   ERROR: Transform failed, trans_rc = 8.

Recovery

If you encounter this error you should restore your IT data mart to its previous state prior to running this aggregation job. You should use the methods tha are described in SAS Notes 32463 or 34284, depending on the level of hot fix that is implemented at your installation. After you successfully restore your IT data mart, implement one or more of the suggested workarounds that are documented below.

Suggested Workarounds

The function of any workaround is to restrict further the amount of data being processed in a single PROC RANK. Keeping this in mind, consider the following possible workarounds:

1. Run the aggregation job with OPTIONS MEMSIZE=MAX. This action might be sufficient to circumvent the problem.
2. Consider changing the class list that is used for summarizing the data, such that the resulting observations are not as granular. For example, if your class list had DAYDATE and HOUR, removing HOUR would result in a 24:1 reduction in output rows. This workaround is only viable if the resulting data is still appropriate for your desired analysis. If you make this change you must redeploy your aggregation job.
3. Consider using an aggregation input filter to restrict the number of rows of data and, presumably, the number of output class combinations. For example, if several machines contribute to an inordinate amount of data, you might want to use an input filter to omit data from these machines. It might be useful to work with that data as part of a separate aggregation. If you choose this option, you will need to remove the existing data for these machines from your existing aggregation. Please contact SAS Technical Support for some sample code that might assist you with this task.
4. Consider not requesting the troublesome rank. Typically, this problem occurs for ranks that are requested for all the data, rather than on a sub-group basis.

Operating System and Release Information