Skip to content

flightdata.md

Latest commit

 

History

History
90 lines (65 loc) · 4.02 KB

flightdata.md

File metadata and controls

90 lines (65 loc) · 4.02 KB
Raw

Flight Data Benchmark

CLI name: flightdata - use benchbox run --benchmark flightdata

Query templates: every Flight Data query

The Flight Data benchmark uses US Bureau of Transportation Statistics (BTS) On-Time Performance data to exercise real-world aviation analytics - delays, routes, carriers, and temporal patterns - across 20 OLAP queries.

Overview

Property Value
Total Queries 20
Categories 5 (on-time, delay, routes, temporal, carriers)
Data Source BTS TranStats (downloader included)
Schema Flight records with carrier, airport, and delay fields
Scale SF=0.01 (1 month sample) → SF≥10 (full corpus)

Query Categories

Category Queries Focus
ontime 5 On-time performance rates and trends
delay 4 Delay causes and attribution
routes 4 Route performance and connectivity
temporal 4 Day-of-week, month, and hour-of-day patterns
carriers 3 Carrier comparison and ranking

Scale Factors

Scale Approx. Flights Approx. Size Notes
0.01 ~600 K ~10 MB 1-month BTS / dev sample
0.1 ~2.4 M ~100 MB ~4 months of BTS data
1.0 ~24 M ~2.8 GB ~41 months (~3.4 years), uncompressed CSV
≥10.0 Full ~10 GB+ Approaches full historical corpus

Usage

benchbox run --platform duckdb --benchmark flightdata --scale 1.0

benchbox run --platform duckdb --benchmark flightdata --scale 0.01

benchbox run --platform duckdb --benchmark flightdata --scale 1.0 \
  --benchmark-option end_year=2024

benchbox run --platform duckdb --benchmark flightdata --scale 1.0 \
  --benchmark-option seed=42

The first command uses the default scale (SF=1.0, about 24M flights, about 2.8 GB of uncompressed CSV). The second is a quick development sample. The third pins the most recent year included; the default is LAST_AVAILABLE_YEAR. The fourth makes runs reproducible through a seed.

Benchmark Options

Option Default Description
end_year LAST_AVAILABLE_YEAR Most recent year to include (works back)
seed None Random seed for reproducibility
force_regenerate False Force data regeneration

Data Generation

BenchBox downloads BTS On-Time Performance extracts via the bundled downloader. The initial pull can be large for SF ≥ 1 - use the generate phase in isolation to pre-stage data:

benchbox datagen --benchmark flightdata --scale 1.0

Platform Support

Any SQL platform supported by BenchBox can run Flight Data. It is a useful real-world complement to TPC-H and NYC Taxi for temporal / categorical workloads and is supported on DataFrame platforms (Polars, Pandas, etc.) via the bundled DataFrame query definitions.

See Also