People often ask, if they need twice the memory when using Database In-Memory.

After the last meetup when this question came up again, the idea was born to setup an AskTom Database In-Memory (EMEA) Office Hours session with Andy Rivenes @TheInMemoryGuy to discuss this topic.

There is the unique Dual Format Architecture with the Row Store (Buffer Cache) and the In-Memory Column Store. Does analytic workload have any impact to the Buffer Cache, e.g. do we need significant additional blocks or is the work done in the In Memory Column Store.

To figure out the allocated space in Buffer Cache we query DBA_OBJECTS and V$BH. Following picture is showing the result when running single row queries (OLTP type – using Index STEP3_3) and analytic queries (using In Memory Column Store) on table LINEORDER.

number_of_blocks

This proves that there is absolutely no impact on Buffer Cache when running analytical queries against In Memory Column Store.

Do we even need less memory when using Database In-Memory? Get more details in our recording here at our YouTube channel: https://www.youtube.com/watch?v=cX9hK5QjQag

 

Original publish date: 12/13/2019