Set the node cache size based on site size (318976)

MORE INFORMATION

The following SQL query calculates the number of items in the Content Management Server (CMS) database that must be accounted for when you configure the size of the node cache. To run this query, point your SQL Query Analyzer at your CMS database, and then paste and run the following SQL script in the SQL Query Analyzer window.

Npte Microsoft strongly recommends that you set your node cache to at least the number that is returned by this script. If constant authoring and updates occur on your site, increase the node cache size accordingly.

SELECT count(*) FROM node
WHERE (datediff(hour, archivedwhen, getdate()) < 1
  OR archivedwhen IS null)
  AND (deletedwhen is NULL)
  AND datediff(hour, expirydate, getdate()) < 1
  AND (datediff(hour, effectivedate, getdate()) > 1
    OR effectivedate IS null)
  AND (datediff(hour, approvedwhen, getdate()) > 1
    OR (type >= 65535 and approvedwhen is null))

				

If the returned value is higher than 35,000, you may not be able to load the whole site in memory. The available memory varies greatly. Factors, such as level of isolation for Microsoft Internet Information Services (IIS) and the other processes that IIS is running, can vary greatly, and all these factors influence your available memory. This dictates how high you can set your node cache. Therefore, it is a good idea to avoid increasing your node cache to anything higher than 35,000.

If the node cache is more than 35,000, Microsoft recommends that you run stress tests against your non-production servers to tune the node cache number. The stress test must be performed across the whole site and not only on several pages. If you perform a partial stress test, it does not create the appropriate environment for node cache tuning. If the return value from the previous script is between 2,000 and 35,000, the site will perform best on the number that is returned by the script.

If memory consumption is too high soon after you change the node cache, you must reduce the node cache number. If memory consumption is too high, more than 1 GB of virtual memory is being used on the instance that is running MCMS or performance is degraded. This problem is not likely to occur, but because there is no easy way to calculate the available memory, the node cache number may use all the available memory and therefore cause memory issues.