Reactive Caching for Composed Services
summary
Reactive caching maintains summaries and caches with bi-directional links to track dependencies and incrementally repair distributed read-only operations, guarantees eventual consistency under failures, and ships as an implementation that extends Orleans. Experiments on two cloud benchmarks show much better timeliness than polling (fewer missed deadlines), latency close to hand-written change notifications for cached results, a modest cold-query overhead from dependency bookkeeping, and throughput trade-offs under heavy write load that are mitigated by backpressure.
AI-written summary · about
chapters (10)
- 0:00Reactive caching research
- 7:45Composed services explained
- 10:56Keeping views fresh
- 13:04Algorithm and implementation
- 17:36Incremental repair vs reactive
- 19:01Fault tolerance handling
- 19:22Benchmarking performance
- 21:29Comparison with propagation
- 21:53Timeliness and deadlines
- 22:13Q&A
related talks
14mMuCache: A General Framework for Caching in Microservice Graphs
54mDistributed Caching in Real-Time Payment Systems
43mCaching half a billion user transactions
44mReacting to the future of application architecture
44mResilience in Microservices: Distributed Cache and More
01h 42mWhat is a Reactive Application?
47mReactors Road to Composable Distributed Computing
57mResilience by Design