Repository navigation
chore: Add PySpark-based benchmarks, starting with ETL example - #3065
Merged
Merged
Conversation
Member
Author
|
Current timings: edit: on later runs I actually saw much better timings, with Comet taking ~10% longer than Spark rather than being 2x slower 
JVM shuffle is slower than Spark shuffle when writing, but writes smaller shuffle file: 
|
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## main #3065 +/- ##
============================================
+ Coverage 56.12% 59.50% +3.38%
- Complexity 976 1382 +406
============================================
Files 119 167 +48
Lines 11743 15524 +3781
Branches 2251 2575 +324
============================================
+ Hits 6591 9238 +2647
- Misses 4012 4989 +977
- Partials 1140 1297 +157 ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
andygrove
marked this pull request as ready for review
January 12, 2026 15:05
comphead
reviewed
Jan 12, 2026
|
|
||
| This script generates a parquet dataset with a realistic schema (100 columns | ||
| including deeply nested structs, arrays, and maps) for benchmarking shuffle | ||
| operations across Spark, Comet JVM, and Comet Native shuffle modes. |
Contributor
There was a problem hiding this comment.
would be nice to see the generated schema example, it might be separate txt file, WDYT?
Member
Author
There was a problem hiding this comment.
The schema does get printed each time you run the script
comphead
reviewed
Jan 12, 2026
| --conf spark.comet.exec.enabled=true \ | ||
| --conf spark.comet.exec.all.enabled=true \ | ||
| --conf spark.comet.exec.shuffle.enabled=true \ | ||
| --conf spark.comet.shuffle.mode=jvm \ |
Contributor
There was a problem hiding this comment.
Suggested change
| --conf spark.comet.shuffle.mode=jvm \ |
this is hallucination
comphead
reviewed
Jan 12, 2026
Co-authored-by: Oleks V <comphead@users.noreply.github.com>
comphead
approved these changes
Jan 12, 2026
comphead
left a comment
Contributor
There was a problem hiding this comment.
Thanks @andygrove this is a good start!
hsiang-c
reviewed
Jan 12, 2026
| ) | ||
|
|
||
| print(f"Generated schema with {len(df.columns)} columns") | ||
| df.printSchema() |
Contributor
There was a problem hiding this comment.
👍 It's easier to check schema
hsiang-c
reviewed
Jan 12, 2026
| df.printSchema() | ||
|
|
||
| # Write as parquet | ||
| df.write.mode("overwrite").parquet(output_path) |
Contributor
There was a problem hiding this comment.
(nit) We can add zstd compression if you want, not necessary though.
hsiang-c
reviewed
Jan 12, 2026
| " 'country', 'USA'," | ||
| " 'geo', named_struct(" | ||
| " 'lat', cast(id % 180 as double) - 90.0," | ||
| " 'lng', cast(id % 360 as double) - 180.0," |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Which issue does this PR close?
Closes #.
Rationale for this change
Add a realistic benchmark with a wide schema and nested types that can be run in a cluster, in an effort to track down reported issues of Comet producing larger shuffle files than Spark.
This is intended as a starting point for a full suite of standalone benchmarks to cover different use cases.
What changes are included in this PR?
ETL benchmark: read parquet, repartition, write parquet.
How are these changes tested?