andygrove opened a new pull request, #2549: URL: https://github.com/apache/datafusion-ballista/pull/2549
## Summary Adds a **Trino 483** column to the TPC-H SF1000 results, from a run on the same cluster, the same data, and 32 workers of the same pod shape (8 vCPU, 64 GiB). The Ballista, Spark, and Comet numbers are unchanged. - **Trino 483:** **411.62 s** (all 22 queries passed, and the row counts were checked) - Spark 4.1.3 + Comet 1.1.0-rc2: 455.09 s - Ballista `d6d8bd91`: 571.55 s - Spark 4.1.3: 898.17 s This is **not an apples-to-apples comparison**, and the new **Trino configuration** section says why: - **Warm timings:** Trino's number is the mean of 3 iterations after an untimed warm-up run. The other columns include the cold first iteration. - **In-memory exchanges:** Trino streams data between stages and never writes shuffle files. Ballista, Spark, and Comet write every shuffle to disk. Spilling was off. - **Partition pruning:** Trino uses Hive tables with the partition columns declared, so it prunes partitions the way Spark and Comet do. Ballista doesn't. - **No statistics:** `ANALYZE` was not run. - **Untuned:** besides the listed settings, everything is at Trino's defaults, and no Trino expert has reviewed the config. The Versions, Environment, Results, and Reproducing sections now cover Trino too. ## Test plan - [x] `npx [email protected] --check docs/source/contributors-guide/benchmarking.md` - [ ] Preview the rendered docs page and check the tables and the `#trino-configuration` anchor -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
