Using Intel Optane Persistent Memory for Database Read Caching

Everybody wants faster access to their data, but DRAM is expensive and limited in size.  Persistent memory promises larger capacity at a cheaper price point, but with a higher latency than DRAM.

This talk investigates how a mixture of DRAM and persistent memory can benefit databases, specifically for read caching.

Intel Optane persistent memory comes in two flavours (Memory Mode and App Direct Mode). This talk looks at the experience of using both these modes for an In-Memory SQL database called Oracle TimesTen. Persistent memory is great, but it is not magic and you still need to make hard design choices.

We look at things like NUMA effects and how the ratio of DRAM to persistent memory affects latency and system capacity.

We also look at the cost/benefits of using large DRAM machines versus using clusters of persistent memory machines for read caching.