Mitigating Resource Usage Dependency in Sorting-based KV Stores on Hybrid...
summary
DecouKV decouples sorting into index merge tasks and data append and data flush tasks by separating indexes (stored in a mergeable skip list) from values (append-only files), adds elastic level capacity and IMQ/DFQ queue-based scheduling, and auto-tunes queue parameters based on Optane/MVMe + SATA hybrid experiments. Compared to RocksDB, MatrixKV, PrismDB, SplitDB and ADOC, DecouKV raises CPU utilization by 25.4–32.3%, boosts write-intensive throughput 2.3–4.9× (read-intensive 1.2–2.3×) and cuts tail latency 74.3–91.4%.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 1:45Resource competition in LSM-tree KV stores
- 3:51Coupling issues in sorting KV stores
- 5:59Existing strategies for hybrid storage
- 7:27Operation decoupling in DO-KV
- 10:22Auto-tuning and resource monitoring
- 11:25Evaluation and performance
- 14:42Q&A
related talks
18mSplitKV: Splitting IO Paths for Different Sized Key-Value Items with Advanced
13mDifferentiated Key-Value Storage Management for Balanced I/O Performance
21mMatrixKV: Reducing Write Stalls and Write Amplification in LSM-tree Based KV Stores
12mSpanDB: A Fast, Cost-Effective LSM-tree Based KV Store on Hybrid Storage
15mKVIMR: Key-Value Store Aware Data Management Middleware for Interlaced Magnetic...
18mPreview: KV-Stores and Databases
19mBuilding KVStore for ML Workloads
16mAegonKV: A High Bandwidth, Low Tail Latency, and Low Storage Cost KV-Separated LSM Store