The first question to ask when you want to determine how much CBM memory you need is whether you want the data set to be able to fit in memory. I/O performance is very important to achieve maximum query performance and using main memory to store the data (in addition to disk for high availability and recovery reasons) can avoid a potential data storage layer bottleneck.

By default Vector keeps data compressed in CBM memory. Maximum performance can be achieved when data is stored uncompressed in memory--as described in Database Configuration but this will increase the memory size requirement.

If your data set is significantly greater than your planned CBM memory size, then consider the size of the frequently accessed data, which should ideally fit into CBM memory. Vector performs smart buffer management to maximize the throughput across multiple operations when data retrieval from disk is required.

Unless configured otherwise, per database Vector will configure 25% of your available memory to CBM memory by default.