Stripeless Data Placement for Erasure-Coded In-Memory Storage
summary
Stripeless scheme replaces stripe coordination by having each primary replicate to K backups and each backup XOR-encodes the K replicas into parity, using a symmetric balanced incomplete block design (SBIBD (lambda=1)) so any two primaries share at most one backup and recovery requires adversary control of >half the nodes. Implementation runs as a background encoder in a key‑value store (Nostor implemented in ~16k Rust lines) on a 16‑node CloudLab cluster and shows 1.61x–2.60x higher throughput on small objects while paying extra P+1 replica bandwidth and rigid cluster-size constraints.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 1:46Erasure coding basics
- 3:51Limitations of existing approaches
- 4:36Stripeless design principles
- 6:22Backup node selection
- 7:24Correctness and recovery
- 8:27Implementation benefits
- 9:08Limitations and constraints
- 9:51Performance results
- 10:34Conclusion
related talks
01h 14mErasure Coding at Scale
11mExploiting Combined Locality for Wide-Stripe Erasure Coding in Distributed Storage
15mLESS is More for I/O-Efficient Repairs in Erasure-Coded Storage
24mSurviving a Disk Apocalypse with Single Overlap Declustered Parity
14mNimble: Rollback Protection for Confidential Cloud Services
20mFine Grained Replicated State Machines for a Cluster Storage System
50mImmutable Values in the Big Wide World
15mSelectiveEC: Selective Reconstruction in Erasure-coded Storage Systems