The following are some additional features
of SAS that you can control to improve system performance and make
efficient use of your computer's resources. For additional information
about optimizing SAS performance, see the chapter on optimizing system
performance in
SAS Language Reference: Concepts.
-
Create SAS data sets instead of accessing flat
ASCII files. SAS can access a SAS data set more efficiently than it can read flat files.
Also, you should convert
existing data sets that you use frequently to SAS 9.4 format.
-
In your SAS code, use IF-THEN-ELSE
conditional structures instead of multiple IF-THEN structures. When
one condition in the IF-THEN-ELSE structure is met, control returns
to the top of the structure (skipping the ELSE clause, which might
contain subsequent IF-THEN structures). With multiple IF-THEN structures,
each condition must be checked.
-
When using arrays, make them _TEMPORARY_ if possible. This action requires less
memory and less time for memory allocation.
-
Use programming structures that
reduce file I/O, the most time-intensive aspect of SAS processing.
Some ideas for reducing file I/O are:
-
Use the WHERE statement in a procedure
to reduce extra data processing.
-
Use indexed data sets to speed
access to the desired observations.
-
Use the SQL procedure to subset
and group your data.
-
Experiment with the value of the CATCACHE
system option, which specifies the number of SAS catalogs to keep open at one time. By default,
no catalogs are cached in memory (and CATCACHE is set to 0). Caching catalogs is an
advantage if one SAS application
uses catalogs that are subsequently needed by another SAS application. The second
SAS application can access the cached catalog more efficiently.
Note: Storing catalogs in memory
can consume considerable resources. Use this technique only if memory
issues are not a concern.
-
Store your data sets in a compressed format (using the COMPRESS data set option).
This action can improve the performance of your SAS application, though it might require
more
CPU time to decompress observations as SAS needs them. The COMPRESS data set option is described
in the data set options section of
SAS Data Set Options: Reference.
-
If you specify the Scatter-read/Gather-write system option, SGIO, SAS bypasses intermediate
buffer transfers when reading or writing data. SAS
reads ahead the number of pages specified by the BUFNO system option and place the
data in memory before it is needed. When the data is needed it is already in memory,
and is in effect
a direct memory access. Different values for the BUFNO system option should be tried
for each SAS job to find the maximum performance benefit. For more information see
Achieving Better I/O Throughput Using SGIO in the Microsoft
Windows EnvironmentScatter–read
/ gather–write is active only for SAS
I/O opened in INPUT,
OUTPUT mode, and UPDATE mode if the access pattern is sequential.
If any SAS
I/O files are opened in UPDATE or RANDOM mode, SGIO is inactive for that process. Compressed
and encrypted files can also be read ahead using scatter-read/gather-write. For more
information about the SGIO system option, see
SGIO System Option: Windows .