A new in-memory cache implementation for the SAS OLAP Server is
available with the first 9.1.3 hot fix.

The default cache size is 16MB. It is recommended not to change this
value. The NWAY aggregation is never cached. The in-memory cache only
loads complete aggregations. The first time a stored aggregation is
being accessed by a query, the cache tries to load it.

Only aggregations with a row count below the load threshold are being
loaded. The load threshold value is 200 times the cache size in MB. For
example, for the default cache size setting of 16MB, aggregations with a
rowcount < 3200 (200*16) are being loaded into the cache if there is
enough memory left in the cache. The length of the aggregation records
is not taken into account.

Rows in the in-memory cache are not indexed, while aggregations on disk
typically are. The larger an aggregation, the slower the access to it in
memory. The current threshold of 3200 rows is a good maximum for
efficient un-indexed in-memory tables.

Cube aggregations loaded into the cache stay there until the OLAP server
is stopped or until the cube is refreshed using the OLAP Server Monitor:
Refresh Cubes utility.

The in-memory cache is shared for all cubes served by the OLAP Server.
The cache memory is not reserved.  If regular OLAP Server processing
runs out of memory, it may use memory from the cache memory pool if
there is any left.

Please note that the details of how the OLAP Server in-memory cache
functionality works may vary from release to release.


A fix for SAS 9.1.3 (9.1 TS1M3) for this issue is available at:

http://www.sas.com/techsup/download/hotfix/e9_sbcs_prod_list.html#013344

For customers running SAS with Asian Language Support (DBCS), this
fix should be downloaded from:

http://www.sas.com/techsup/download/hotfix/e9_dbcs_prod_list.html#013344

Operating System and Release Information

A fix for SAS 9.1.3 (9.1 TS1M3) for this issue is available at:

A fix for SAS 9.1.3 (9.1 TS1M3) with Asian Language Support (DBCS) for this issue is available at :