Abstract
Emerging persistent memory technologies such as Intel’s 3D XPoint™ will provide an opportunity to optimize database applications for a new tier of high capacity memory that is lower cost than DRAM. This presentation describes how to use persistent memory to augment DRAM in the Redis in-memory key-value database. We describe different ways to modify Redis to use persistent memory either as a large, volatile tier, or as a fast, persistent media. We then explore how to achieve better performance based on the characteristics of the media. Finally, we highlight two interesting challenges encountered during this work, and describe their solutions.
Learning Objectives:
1. Understand the benefit from emerging persistence memory such as Intel’s 3D XPoint™
2. Describe different ways on how to migrate database application such as Redis to use persistent memory to replace DRAM
3. Highlight the difference between volatile and persistent usage of persistent memory
4. Performance optimization based on persistence memory characteristics