Understanding vault cache memory

When a Meridian vault is stored in SQL Server, two levels of caching are in effect instead of just one when Hypertrieve is used; the Meridian object cache and the SQL Server buffer. For information on configuring the cache size, see Configuring the MaximumCacheSize setting and Configuring the RelativeCacheSize setting.

Similarly, we recommend 15 percent of the database size for the size of the SQL Server buffer as a rule of thumb. For information on configuring the SQL Server buffer size, see the SQL Server documentation. When physical memory is scarce, it is best to give priority to the Meridian object cache. This will ensure that queries remain responsive, although the update throughput will be diminished if the SQL Server buffer is smaller than recommended.

Managing SQL Server memory consumption may need special attention. SQL Server is quite aggressive in requesting memory from Windows, and can push the Meridian object cache out of physical memory (if they are both running on the same computer), which will degrade performance and can cause “Out of Memory” errors. To prevent this, limit the maximum amount of memory SQL Server can use. You can set this limit in the SQL Server Enterprise Manager on the Memory property page of the SQL Server integrated with Meridian. When calculating this limit, remember that it is for the whole server, including all vaults and other applications using the server.