Introducing a Temporal Benchmark in Benchbase
summary
A Benchbase benchmark for temporal tables is implemented with two tables (employees, positions) and procedures for inserts, updates, deletes, joins and three FK-check SQL implementations: a 22-line exists rewrite, a 12-line lag/window version, and a 6-line rangeagg implementation. Three Postgres clusters run side-by-side; matplotlib, dstat/pcp and bpftrace analyze results. Rangeagg shows lowest 95th-percentile latency; exists only outruns others when artificially high FK failures expose short‑circuit plans. Benchbase commit/fsync behavior and reproducibility remain tricky.
AI-written summary · about
chapters (6)
- 0:00Introduction and motivation
- 3:16Why temporal needs its own benchmark
- 7:15Three temporal foreign key implementations
- 14:56Benchmarking foreign key checks
- 18:05Results and surprising findings
- 20:10Future work and next steps
related talks
31mDemanding the impossible rigorous database benchm
46mIntro to Postgres Hacking via Temporal Tables
01h 03mTemporal Database Demo
23mImplementing Temporal Features in PostgreSQL: SQL Standard and Beyond
29mPerformance Archaeology - 20 years of improvements
46mEnhancements to the bitemporal model support: integrity constraints
44mDemystifying Benchmarks: How to Use Them To Better Evaluate Databases
48mModern SQL features in PostgreSQL