For a 80-160 dimension OLAP query over three or four tables, each with 50-100 million rows, the query takes an hour on average on SQL Server. It is doing full table scans 60% of the time, lots of paging, i/o queues are 5-6, there are a lot of group by and order by clauses which do a lot of sorting, mostly in disk and some in memory. Would doubling the memory and striping across twice as many disks solve the problem or is it that SQL Server is too overworked in this case? The CPU utilization is 100% with three users. There are seven or eight users, but they cannot get into the system. Now we are using a 2 CPU Xeon 2.4 GHz machine with 4 GB of RAM. The database size is 300 GB. The average table size is 50 GB, and the max table size is 80 GB.

    Requires Free Membership to View

    Login

I would recommend taking a look at the data design before throwing more hardware at the problem. 80-160 dimensions in a query seems a bit excessive. Could it be you have an example of a one-size fits all data mart? I suggest getting your data architect/business analyst involved to gain an in-depth appreciation for how that query result set is being used, then look at the data architecture for areas where the design can be improved.

For More Information


This was first published in January 2004

Join the conversationComment

Share
Comments

    Results

    Contribute to the conversation

    All fields are required. Comments will appear at the bottom of the article.
    Back to top