Low End-to-End Latency atop a Speculative Shared Log with Fix-Ante Ordering
summary
Proposes SpecLog, a speculative shared log that predetermines global cuts via fix-ante ordering, implements it as Belfast, and overlaps application compute with ordering so records become consumable immediately after durability. Belfast enforces per-shard quotas, pads with no-op records or delays excess appends, and handles view changes; whole-shard failures cause rare mispredictions. Evaluation shows Belfast delivers records up to 3.5× earlier and yields 1.6× lower end-to-end latency; prototypes include high-frequency trading, fraud and intrusion detection.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 3:27Motivation for Speculative Shared Logs
- 7:19High delivery latency in logs
- 9:23Predicting record positions
- 10:27Fix-ante ordering
- 12:16Speculation success or failure
- 13:41Record path with SpecLog
- 14:23Design of Belfast
- 15:44End-to-end latency benefits
- 16:28Belfast for real-world apps
- 17:11Conclusion
related talks
19mScalog: Seamless Reconfiguration and Total Order in a Scalable Shared Log
01h 12mScaling distributed ledgers: Achieving arbitrary throughput and sub-second latency
19mSpeculative Recovery: Cheap, Highly Available Fault Tolerance with Disaggregated...
21mLogDevice At Scale
23mBodega: Localized Linearizable Reads at Anywhere Anytime via Roster Leases
19mBuilding a Performant Configuration-Store Layer
36mStreamlio: Towards a Unified End-to-end Solution
33mLibraBFTv2: Optimistically-linear BFT Consensus with Concrete Latency Bounds