Beacon: Intelligent Latency-Aware and Load Shedding Service Routing
summary
Beacon is a service router that checks per-key last-update times against live replication delays and routes reads to either primary read-write or read-only replicas to guarantee freshness and shed load. Implementation uses a Beacon router + status server + replication latency domain table (the replication latency domain), moved from coarse sampling to per-domain continuous monitoring. In production Beacon handles 20.3 million calls in 24h, routing 17.8M to replicas and making routing decisions in ~2.6 ms.
AI-written summary · about
chapters (15)
- 0:00Start
- 1:20November 15, 2022 Taylor Swift
- 3:14Sites With Similar Challenges
- 4:16Architecture for Scaling and Resilience
- 4:44Database Replication for Load Distribution
- 5:08Infrastructure Support
- 6:19Intelligent Routing Guaranteeing QoS
- 14:22Beacon Original Architecture and Operational
- 15:02Database Replication for Resilience Customer Service
- 15:53Beacon 2.0
- 23:45Beacon as Infrastructure Service
- 24:17DB Latency in Production
- 25:33Beacon Routing of a Service During 24 Hours
- 27:01Beacon Routing of a Service at Market Open
- 28:16Meet Morgan Stanley
related talks
44mAvailability, Latency and Cost: Withstanding Outages
16mSkybridge: Bounded Staleness for Distributed Caches
48mReplication as a First-Class Citizen: The Art of Global Resilience
47mBEAM Architecture Handbook
30mScaling HDFS with Consistent Reads from Standby Replicas
51mReplicated service mesh: hardening systems against failure modes in load balancing
15mBeaver: Practical Partial Snapshots for Distributed Cloud Services
45mBuilding Available and Partition Tolerant...