Hydra's Heads – Vladimir Ozerov
summary
Company builds custom query engines and optimizers for distributed databases, reusing components like Apache Calcite and Flink but adapting rules, operators and execution strategies to each target system. Describes a real case where a change makes a query go from 1.5 → 5 minutes, traces that to subplan reuse and join ordering conflicts, and uses plan sampling and executing many alternatives to measure optimizer quality. Says ML is promising but brittle and slowly adopted.
AI-written summary · about
chapters (16)
- 0:00Welcome
- 1:02Vladimir Ozerov's database journey
- 2:27Starting a database career
- 4:59Hazelcast distributed systems
- 5:40Moving to CockroachDB
- 7:07Query optimization in distributed systems
- 10:39Custom query optimizer development
- 13:26Apache Calcite and open-source
- 18:27Academic research impact
- 21:17Query optimization challenges
- 24:28Optimizer quality assessment
- 26:32Machine learning in query optimization
- 29:23Database funding and trends
- 31:49Database team interview process
- 37:04Why new databases and Russia's role
- 39:52Database books recommendations
related talks
39mBuilding modern SQL query optimizers with Apache Calcite
01h 04mQuery Optimization and Acceleration at Dremio
33mHydra - An Open Source Processing Framework
41mNo stats, no problem: building feedback-driven optimizers for lakehouses
50mHow SQL Databases Come up with Algorithms that You Would Have Never Dreamed Of
16mVulcan: Automatic Query Planning for Live ML Analytics
01h 13mHow We Built SQL in Hazelcast
01h 01mOrca: A Modular Query Optimizer Architecture for VMware Greenplum