Inside Twitter's Realtime Delivery of Tweets for Enterprise Customers
summary
PowerTrack reliably filters hundreds of millions of daily Tweets at thousands per second and delivers customizable subsets to enterprise customers with predicate indexing to avoid full AST evaluation and a Kafka-like eventbus per customer for backfill so fidelity approaches 100% When load spikes stacks scale by adding nodes and side A/B redundancy plus zookeeper leases coordinate cross‑datacenter stream connections and the streaming layer dedupes by tweet ID before HTTP streaming to clients
AI-written summary · about
chapters (5)
- 0:00Introduction and architecture
- 21:23Data filtering and scaling
- 23:44Configuring rule updates
- 27:21Mitigation and monitoring
- 32:28State management and Q&A
related talks
28mAn Architecture to Tweet Them All
34mBig and Small Data at @Twitter
28mThe Evolution of Twitter's Edge
45mRealtime Search at Twitter
01h 01mStream Data Processing for Fun and Profit
30mThe Distributed Database Behind Twitter
42mBuilding a near real time personalized feed using data streaming and Aerospike
06mPREVIEW: Real-time Stream Processing at Scale