Skip to content

Latest commit

 

History

History
22 lines (15 loc) · 1.14 KB

File metadata and controls

22 lines (15 loc) · 1.14 KB

Benchmark results

Captured 2026-07-30T21:09:34Z on Intel(R) Xeon(R) Processor @ 2.10GHz, 2 logical CPUs, 7.8 GiB RAM, Python 3.11.15, DuckDB 1.5.5.

Method: Warm DuckDB file on the local filesystem, single process, no concurrency. Each scan figure is 12 repeated full scans of the warehouse; each attribution figure is 60 single day attribution calls, one per distinct date.

Scan latency by dataset size

Scale Fact rows Slices Baseline build (ms) Scan p50 (ms) Scan p95 (ms) Scan p99 (ms)
small 59,076 14 1479 133.1 153.3 158.6
medium 656,640 20 1892 273.1 370.6 415.7
large 1,313,280 23 1988 315.2 400.2 417.4

Attribution latency by slice count

Configuration Slices Baseline rows Baseline build (ms) Attribution p50 (ms) Attribution p95 (ms) Attribution p99 (ms)
depth1-22-slices 23 104,880 2323 19.96 38.05 64.18
depth2-212-slices 214 975,840 19306 33.25 53.32 63.59

Raw numbers: results.json in this directory. Regenerate with make bench.