Event Streams at Groupon with Storm, Mesos & Griddle
summary
Groupon runs Storm on Mesos to consolidate dozens of topologies and process 100–200k events/sec by mapping Mesos CPU/memory offers into Storm worker slots, isolating topologies with cgroup-backed Mesos executors and a submitter app. A synced Nimbus backup removes a SPOF, IMetrics consumers feed unified visibility, and a compiled Griddle DSL replaces groups of bolts to cut latency (roughly ~10 ms per migrated bolt) while preserving throughput and operational efficiency.
AI-written summary · about
chapters (5)
- 0:00Start
- 12:42Storm Mesos Framework
- 13:14Resource Model: Storm
- 18:17Groupon Storm as a Service
- 20:47Pros vs. Native Storm
related talks
38mHighload — An Unexpected Journey
42mCloud based low cost low maintenance scalable data platform
42mUsing Mesos to Drive Devops Adoption at Scale
21mSelf-Service Chaos Engineering: Fitting Gremlin into Grubhub's DevOps Culture
18mStormdriver One Year Later
15mSharding Clouddriver with Stormdriver
21mLightning Talks
59mLightning Talks