ELECT: Enabling Erasure Coding Tiering for LSM-tree-based Storage
summary
ELECT implements hybrid redundancy on top of Cassandra, encoding LSM last-level SSTables offline into erasure-coded coding groups while keeping hot keys replicated; experiments with YCSB on Alibaba Cloud plus object storage show 56.1% hot-tier storage savings and 39.1% overall savings, degraded-read recovery takes ~50% longer (network-bound) but cloud latency can cut worst-case reads from ~5x to ~1.2x, and the tunable alpha parameter yields actual savings between 9.2% and 86% while preserving normal-mode throughput.
AI-written summary · about
chapters (11)
- 0:00ELECT and storage tiering
- 2:04Storage tiering use cases
- 4:33Erasure coding vs replication
- 7:00ELECT design and contributions
- 8:46Redundancy transitioning
- 11:15Offline erasure coding
- 13:44Hotness-aware thresholds
- 16:29Experimental setup on Alibaba Cloud
- 17:33Storage savings and performance
- 19:40Recovery performance impact
- 20:45Conclusion and Q&A
related talks
01h 14mErasure Coding at Scale
24mOpenEC: Toward Unified and Configurable Erasure Coding Management...
13mClosing the B+-tree vs. LSM-tree Write Amplification Gap on Modern Storage Hardware...
16mPolarStore: High-Performance Data Compression for Large-Scale Cloud-Native Databases
18mPreview: KV-Stores and Databases
26mFast Erasure Coding for Data Storage: A Comprehensive Study of the Acceleration...
16mDEPART: Replica Decoupling for Distributed Key-Value Storage
16mHolistic and Automated Task Scheduling for Distributed LSM-tree-based Storage
