How should you determine how much memory should be set aside for the buffer pool versus query execution?

If you use rather simple queries that do not require a lot of memory (for example, large scans that produce relatively few aggregates), you do not need a lot of query execution memory.

If you have complex queries (for example, joins between large tables, or aggregations that produce many results), you need more query memory.

If your disk system is fast or if your set of frequently accessed data is rather small, large disk buffer is not that crucial. Similarly, if you have a large disk-resident dataset and a slow disk, increasing the buffer pool size can help.

A rule of thumb is to start with query memory to about 50% and buffer pool to about 25% of your memory (the default), and adjust as needed.

For more information, see the Vector Deployment Guide.