DispersedLedger: High-Throughput Byzantine Consensus on Variable Bandwidth Networks
Talkat 19th USENIX Symposium on Networked Systems Design and Implementation (NSDI '22) · 2022-04-04..04-06·
summary
Uses erasure coding and a new verifiable information dispersal to separate data availability from block delivery so nodes agree on block hashes without downloading full blocks, then fetch chunks asynchronously. Replaces an O(n^2) verification step with O(n) per-node cost in the new avidM primitive, implemented in ~8k LOC and tested on 16-node AWS and other clouds. Prototype reports 2x throughput and 74% latency reduction versus HoneyBadger, so it works under variable bandwidth.
AI-written summary · about
chapters (6)
- 0:00Introduction to consensus
- 2:06Identifying the bottleneck
- 4:24Decoupling block delivery
- 6:47The AvidM protocol
- 10:29Evaluation and results
- 14:28Conclusion and summary
related talks
22mOptimizing network topologies...
01h 12mScaling distributed ledgers: Achieving arbitrary throughput and sub-second latency
56mConsensus Through Herding
29mBuilding Probabilistic BFT Protocols Through Subsampled Voting
42mBeyond Blockchain: Convergent Consensus
25mEnhancing Bitcoin Security and Performance with Strong Consistency via...
13mBuilding the Fastest Decentralized Internet Protocol for Web3
30mStrengthening Reliable Decentralized Systems at Interchain