Uber’s Innovations in OpenSearch for a Cloud Native, Serverles...
summary
Describes concrete work shipped into OpenSearch 3.0: a move to a remote‑store, stateless design with reader‑writer separation, an experimental pull‑based ingestion from Kafka/Kinesis that supports replay and multi‑regional aggregation, and a faster gRPC + protobuf transport replacing REST for inter‑service calls. Shows production constraints: sub‑second freshness for driver locations updated every ~1–2s, millions of items and stores for Eats, and plans for priority‑aware ingestion, offline Spark builds, and support for clusters at hyper‑scale.
AI-written summary · about
chapters (13)
- 0:00Uber’s OpenSearch innovations
- 1:51OpenSearch architecture overview
- 3:36Cloud-native goals
- 8:59Search use cases at scale
- 13:38Old architecture challenges
- 19:00Delivery search optimizations
- 20:24Read-write separated architecture
- 22:31Project Sunrise collaboration
- 23:57OpenSearch 3.0 contributions
- 26:42Kafka ingestion improvements
- 27:47gRPC performance boost
- 28:54Future roadmap
- 29:59Closing
related talks
43mDeep Dive Into OpenSearch Pull-Based Ingestion at Uber
38mEvolution of Uber's Search Platform
42mAdvancing OpenSearch With gRPC and Protobuf at Uber: A High-Performance...
31mSharding 2.0: Unlocking Infinite Search Scalability in OpenSearch
35mHighly Available Search at Shopify
42mClosing Remarks
22mSharding 2.0: Unlocking Infinite Search Scalability in OpenSearch
57mMulti Tenant NoSQL and NewSQL Cloud data Design Patterns Deep dive