How We Built SQL in Hazelcast
summary
Description of how Vladimir and his team built Hazelcast Mustang, a new distributed SQL engine added to Hazelcast IMDG. Covers concrete technical decisions around the query optimizer, the network protocol, and parallel query execution.
AI-written summary · about
chapters (21)
- 0:00SQL in Hazelcast intro
- 2:41Why add SQL to maps
- 5:47SQL engine timeline
- 8:18Optimizer challenges with Calcite
- 12:30Semantic analysis and schema
- 16:22Relational operator tree
- 17:47Query optimization basics
- 23:43Physical optimization rules
- 27:16Index selection and distribution
- 32:17Virtual table schema
- 33:22Runtime execution model
- 36:11Volcano model and execution
- 40:43Worker pool and parallelism
- 42:30Distributed query graph
- 47:47Query cancellation and protocol
- 51:00Race conditions in protocol
- 54:29Flow control and backpressure
- 56:56Fault tolerance and recovery
- 1:01:50Protocol design trade-offs
- 1:03:57Development journey lessons
- 1:06:06Q&A
related talks
59mDistributed applications using Hazelcast
05mStartup Showcase: Hazelcast
55mHazelcast - Hitchhiker’s Guide. Grzegorz Piwowarek. Java Fest 2020 Online
32mMastering stream-to-stream JOINs
30mDistributed Computing with Hazelcast
44mNonrelational SQL
44mHAWQ Architecture
45mEasy Scaling with Open Source Data Structures