Exploring Data Deduplication in LSM Tree-based Key-Value Stores (Work-in-Progress Report)
summary
Explores adding value-based deduplication to LSM trees, showing inline duplication (fingerprinting on writes with a chunk information table) causes severe write-performance drop and metadata inconsistency from in-place updates; builds DeltaKV, an offline dedupe applied at flush that delays fingerprinting to background threads, splits metadata into a value information table and a B+ tree for reads, and writes only unique values to level-0 SSTables; preliminary evaluation reports DeltaKV matches RocksTV performance and outperforms mamdado.
AI-written summary · about
related talks
18mPreview: KV-Stores and Databases
13mDifferentiated Key-Value Storage Management for Balanced I/O Performance
17mLight-Dedup: A Light-weight Inline Deduplication Framework for Non-Volatile...
21mMatrixKV: Reducing Write Stalls and Write Amplification in LSM-tree Based KV Stores
15mXLL: Cross-Layer Logging for Data Deduplication in Consensus-Based Storage
01h 16mDeep Dive into On-Disk Data Structures
12mThe Dilemma between Deduplication and Locality: Can Both be Achieved?
18mDon't Maintain Twice, It's Alright: Merged Metadata Management in Deduplication File...
