Performance
Latency and scan efficiency across recent runs, and a fixed benchmark suite over the sample dataset.
Benchmark suite
benchmark data, not live telemetryA fixed set of query shapes run against the sample dataset, chosen to exercise each stage of the scan path: a full scan with nothing to prune, partition elimination, row-group skipping on min/max statistics, and column projection. Each case records wall-clock time alongside the compressed bytes the query actually paid for.
The byte columns are the point. For every query the engine reports which files it never opened, which row-groups it skipped and why, and which columns it left on disk — measured in compressed bytes, with a regression test asserting that a full scan reports exactly 100%. Most query engines will not tell you that, and neither will Athena.
| Query | Shape | Wall clock | Bytes read | Of dataset | Columns |
|---|---|---|---|---|---|
| Group by partition column | GROUP BY region | 68.5 ms | 3.1 MB | 51.9% | 2 of 7 |
| Filtered aggregate | WHERE + GROUP BY | 127.3 ms | 2.7 MB | 45.1% | 3 of 7 |
| Selective row-group scan | narrow WHERE on a sorted column | 30.6 ms | 271.0 KB | 4.5% | 1 of 7 |
| Partition-pruned count | WHERE on a partition column | 26.7 ms | 577.4 KB | 9.4% | 1 of 7 |
| High-cardinality group | GROUP BY pickup_hour | 109.6 ms | 1.8 MB | 30.4% | 2 of 7 |
Loaded from benchmark_results.json in the repository, regenerated against
the sample dataset on a single machine. Timings are reported as measured.
Latency over time
last 200 successful runsWall-clock per run, planning included. The spikes are cold runs: row-group metadata is cached after first read, so a repeated query against the same dataset skips re-reading every footer.
Scan efficiency
read vs avoidedColumn height is the compressed size of the whole dataset that run targeted. Both sides of the ratio are compressed on-disk bytes — comparing compressed reads against an uncompressed total would flatter every number here by about 1.5x.
Slowest runs
In history| Run | SQL | Duration | Rows scanned |
|---|---|---|---|
| #1 | SELECT * FROM trips LIMIT 20 | 102.01 ms | 300000 |
| #6 | SELECT * FROM trips LIMIT 20 | 101.34 ms | 300000 |
| #11 | SELECT * FROM trips LIMIT 20 | 89.94 ms | 300000 |
| #13 | SELECT COUNT(*) AS total FROM trips WHERE t… | 34.74 ms | 300000 |
| #3 | SELECT COUNT(*) AS total FROM trips WHERE t… | 31.97 ms | 300000 |
| #8 | SELECT COUNT(*) AS total FROM trips WHERE t… | 31.37 ms | 300000 |
| #4 | SELECT region, COUNT(*) AS n, AVG(distance)… | 30.89 ms | 300000 |
| #14 | SELECT region, COUNT(*) AS n, AVG(distance)… | 30.01 ms | 300000 |
Least efficient scans
In history| Run | SQL | Read | Of dataset |
|---|---|---|---|
| #1 | SELECT * FROM trips LIMIT 20 | 5.9 MB | 100.0% |
| #6 | SELECT * FROM trips LIMIT 20 | 5.9 MB | 100.0% |
| #11 | SELECT * FROM trips LIMIT 20 | 5.9 MB | 100.0% |
| #3 | SELECT COUNT(*) AS total FROM trips WHERE t… | 1.6 MB | 26.9% |
| #8 | SELECT COUNT(*) AS total FROM trips WHERE t… | 1.6 MB | 26.9% |
| #13 | SELECT COUNT(*) AS total FROM trips WHERE t… | 1.6 MB | 26.9% |
| #4 | SELECT region, COUNT(*) AS n, AVG(distance)… | 1.4 MB | 23.2% |
| #9 | SELECT region, COUNT(*) AS n, AVG(distance)… | 1.4 MB | 23.2% |
Most efficient scans
In history| Run | Dataset | SQL | Duration | Read | Of dataset | Pruned |
|---|---|---|---|---|---|---|
| #2 | trips | SELECT COUNT(*) AS total FROM trips WHERE region = 'EU' | 9.06 ms | 28.4 KB | 0.5% | 10f / 0rg |
| #7 | trips | SELECT COUNT(*) AS total FROM trips WHERE region = 'EU' | 8.12 ms | 28.4 KB | 0.5% | 10f / 0rg |
| #12 | trips | SELECT COUNT(*) AS total FROM trips WHERE region = 'EU' | 7.99 ms | 28.4 KB | 0.5% | 10f / 0rg |
| #5 | trips | SELECT region, COUNT(*) AS n, SUM(distance) AS total_di… | 17.57 ms | 1010.1 KB | 16.5% | 10f / 0rg |
| #10 | trips | SELECT region, COUNT(*) AS n, SUM(distance) AS total_di… | 13.7 ms | 1010.1 KB | 16.5% | 10f / 0rg |
| #15 | trips | SELECT region, COUNT(*) AS n, SUM(distance) AS total_di… | 17.76 ms | 1010.1 KB | 16.5% | 10f / 0rg |
| #4 | trips | SELECT region, COUNT(*) AS n, AVG(distance) AS avg_dist… | 30.89 ms | 1.4 MB | 23.2% | 0f / 0rg |
| #9 | trips | SELECT region, COUNT(*) AS n, AVG(distance) AS avg_dist… | 28.05 ms | 1.4 MB | 23.2% | 0f / 0rg |
| #14 | trips | SELECT region, COUNT(*) AS n, AVG(distance) AS avg_dist… | 30.01 ms | 1.4 MB | 23.2% | 0f / 0rg |
| #3 | trips | SELECT COUNT(*) AS total FROM trips WHERE trip_id > 2374 | 31.97 ms | 1.6 MB | 26.9% | 0f / 0rg |