ElasticBF: Elastic Bloom Filter with Hotness Awareness for Boosting Read Performance in Large Key-Value Stores
Talkat USENIX ATC '19 · on Thursday, 2019-07-11· at 16:55· USENIX ATC Track II: Grand Ballroom VII–IX·
summary
Describes ElasticBF, a Bloom-filter management scheme for LSM-tree key-value stores (implemented in LevelDB, RocksDB, PebblesDB) that tracks and exploits access skewness within SSTables and segments to dynamically allocate heterogeneous (fine-grained) Bloom filters based on data hotness, reducing false-positive driven I/Os. Presents integration details and performance results showing 2.3–2.6x read throughput improvements while preserving write and range-query performance.
AI-written summary · about
chapters (12)
- 0:00Intro
- 0:29Background
- 0:53LSM-tree-based KV Stores
- 2:06Limitation of Bloom Filters
- 2:40Main Idea
- 3:34ElasticBF Design
- 7:05Performance Evaluation
- 7:46Micro-benchmarks
- 9:06YCSB Benchmarks
- 9:26Comparison with Monkey
- 10:18Impact of System Configurations
- 10:43Conclusion
related talks
13mClosing the B+-tree vs. LSM-tree Write Amplification Gap on Modern Storage Hardware...
51mProbabilistic Data Structures with Redis Bloom
39mAn Introduction to B-trees and LSM-trees for DEVs and DevOps
24mPerformance tuning in BlueStore & RocksDB
12mSpanDB: A Fast, Cost-Effective LSM-tree Based KV Store on Hybrid Storage
29mReasoning about performance (in the context of search)
24mSplinterDB: Closing the Bandwidth Gap for NVMe Key-Value Stores
30mCaching for Speed: Practical Performance Optimization in OpenSearch & Lucene