Recommendation - Drilling Down in Very Large Data Sources
However, when using a Microsoft SQL Server 2005 Analysis Services (SSAS) cube, the memory consumed by the SSAS client is not always released in a way that other queries can effectively re-use. The server continues to run with insufficient memory, causing new queries to either run very slowly or fail completely. You may encounter an error, and, to continue, the system administrator must manually stop and restart the IBM® Cognos® Analytics service.
To avoid these problems, consider the size of hierarchy levels and sets when creating reports and avoid combining them in ways that create large queries. To determine the size of a set, create and run a very small report that includes only a count of the members within that set. You can also use filtering techniques to focus your data.