HotRAP - Hot Record Retention and Promotion for LSM-trees with Tiered Storage

Explore a conference presentation introducing HotRAP, an innovative key-value store enhancement for Log-Structured Merge-trees (LSM-trees) operating on tiered storage architectures. Learn how researchers from Tsinghua University and Shanghai Qi Zhi Institute address the fundamental challenge of balancing efficient read and write operations in tiered storage systems, where traditional approaches force a trade-off between write performance and read access to hot data. Discover how HotRAP builds upon RocksDB to implement a tiering approach that individually promotes hot records from slow to fast storage while maintaining them in fast storage during periods of high access frequency. Examine the system's fine-grained hotness tracking mechanism using a specialized on-disk LSM-tree data structure and dual pathways that minimize delays in moving hot records to fast storage. Understand the performance improvements demonstrated through comprehensive experiments, showing up to 1.6× better performance compared to state-of-the-art LSM-trees under read-write-balanced YCSB workloads with common access skew patterns, and up to 1.5× improvement under Twitter production workloads, making this solution particularly valuable for cost-efficient storage systems that combine fast and slow storage mediums.