Can we make shared logs faster without sacrificing consistency? Introducing LazyLog — a new log abstraction that rethinks how streaming systems handle ordering to achieve dramatically lower write latency.
Traditional shared logs enforce a strict, global order at write time, which guarantees consistency but slows ingestion. LazyLog flips this model: it defers the costly ordering step until data is read. The result? Faster writes, lower latency, and the same strong guarantees when order actually matters.
In this talk, you’ll learn:
- Why traditional log-based systems create latency bottlenecks
- How LazyLog delays ordering to accelerate data ingestion
- Real-world examples of two systems built with the LazyLog abstraction
- How LazyLog maintains consistency while boosting throughput
Developed at the University of Illinois, LazyLog was presented at SOSP, the premier conference for systems research — where it won the Best Paper Award.
If you’re building low-latency data pipelines or high-throughput distributed systems, this talk reveals a breakthrough approach that could reshape your streaming architecture.
