Distributed Join Algorithms in CrateDB
summary
Describes how CrateDB replaced nested-loop and block-nested-loop equi-joins with a distributed block hash join to handle joins across a cluster. Explains adapting single-node hash/block-hash joins to block-based, memory-aware, distributed execution across CrateDB nodes and presents benchmarks (up to 50M-row tables) showing a ~23,000x speedup over the original nested-loop implementation.
AI-written summary · about
related talks
38mAccelerating Query Performance in CrateDB
23mWhen ms matter: Maximizing query performance in CrateDB
55mSort vs Hash A Duality
41mParallel and distributed big joins in H2O
07mJoin Algorithms Meet New Constraints: Scalability, Latency and Privacy
36mCrateDB: A Search Engine or a Database? Both! #bbuzz
39mStorage-Partitioned Join for Apache Spark
41mDistributed Java DBs Under the Hood: Main Components and Interactions Between Them