Overview
Explore a conference talk from USENIX ATC '23 that delves into improving secondary indexing in LSM-based storage systems using persistent memory. Learn about PERSEID, an innovative secondary indexing mechanism designed to enhance performance in LSM-based systems. Discover how this approach leverages the characteristics of persistent memory to achieve significant speed improvements over existing techniques. Gain insights into the specifically designed index structure, hybrid PM-DRAM validation approach, and optimizations for primary table searching. Understand how these advancements lead to performance gains of 3-7 times over existing PM-based indexes and approximately two orders of magnitude improvement compared to current LSM-based secondary indexing techniques.
Syllabus
USENIX ATC '23 - Revisiting Secondary Indexing in LSM-based Storage Systems with Persistent Memory
Taught by
USENIX