[SPARK-54134][SQL] Optimize Arrow memory usage #52747
Closed
Conversation
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
viirya
force-pushed
the
release_buffers
branch
from
309d66f to
14ff1f9
Compare
cloud-fan
reviewed
Oct 29, 2025
sql/catalyst/src/main/scala/org/apache/spark/sql/execution/arrow/ArrowWriterWrapper.scala
Outdated
Show resolved
Hide resolved
viirya
changed the title
viirya
force-pushed
the
release_buffers
branch
from
b0320db to
9f57f37
Compare
cloud-fan
reviewed
Oct 30, 2025
sql/core/src/main/scala/org/apache/spark/sql/execution/arrow/ArrowConverters.scala
Outdated
Show resolved
Hide resolved
viirya
changed the title
viirya
commented
Nov 2, 2025
sql/core/src/main/scala/org/apache/spark/sql/execution/arrow/ArrowConverters.scala
Outdated
Show resolved
Hide resolved
viirya
changed the title
viirya
force-pushed
the
release_buffers
branch
from
77fc1b2 to
0e402d0
Compare
viirya
force-pushed
the
release_buffers
branch
from
f1e7921 to
c3d5a77
Compare
dongjoon-hyun
approved these changes
Nov 3, 2025
Member
dongjoon-hyun
left a comment
dongjoon-hyun
left a comment
There was a problem hiding this comment.
+1, LGTM. Thank you, @viirya and @cloud-fan .
Merged to master/4.1.
dongjoon-hyun
pushed a commit
that referenced
this pull request
Nov 3, 2025
### What changes were proposed in this pull request? This patch proposes some changes to optimize memory usage on Arrow in Spark. It compress Arrow IPC data when serializing. ### Why are the changes needed? We have encountered OOM when loading data and processing them in PySpark through `toArrow` or `toPandas`. The same data could be loaded by PyArrow directly but fails to load through `toArrow` or `toPandas` into PySpark due to OOM issues. ### Does this PR introduce _any_ user-facing change? No ### How was this patch tested? Unit tests. Manually test it locally. ### Was this patch authored or co-authored using generative AI tooling? Generated-by: Claude Code v2.0.13 Closes #52747 from viirya/release_buffers. Authored-by: Liang-Chi Hsieh <[email protected]> Signed-off-by: Dongjoon Hyun <[email protected]> (cherry picked from commit 02ba89f) Signed-off-by: Dongjoon Hyun <[email protected]>
Member
Author
|
Thank you @cloud-fan @dongjoon-hyun |
Contributor
allisonwang-db
left a comment
allisonwang-db
left a comment
There was a problem hiding this comment.
Thanks for adding this!
| val codecType = new Lz4CompressionCodec().getCodecType() | ||
| factory.createCodec(codecType) | ||
| case other => | ||
| throw new IllegalArgumentException( |
Contributor
There was a problem hiding this comment.
Should be SparkException.internalError
Member
Author
There was a problem hiding this comment.
Yea, that would be better.
Member
Author
There was a problem hiding this comment.
I will change to SparkException when I extend this to pandas udf.
zhengruifeng
reviewed
Nov 5, 2025
|
|
||
| val ARROW_EXECUTION_COMPRESSION_CODEC = | ||
| buildConf("spark.sql.execution.arrow.compressionCodec") | ||
| .doc("Compression codec used to compress Arrow IPC data when transferring data " + |
Contributor
There was a problem hiding this comment.
does this optimization take effect in pandas udf?
Member
Author
There was a problem hiding this comment.
I think no, it is currently applied on toArrow and toPandas which is on the reported issue. It should be also available to arrow udf and pandas udf. I will try to extend this to such cases.
dbtsai
reviewed
Nov 5, 2025
| case "zstd" => | ||
| val factory = CompressionCodec.Factory.INSTANCE | ||
| val codecType = new ZstdCompressionCodec().getCodecType() | ||
| factory.createCodec(codecType) |
Member
There was a problem hiding this comment.
Would be great that we can have an option to add compression levels
Member
Author
There was a problem hiding this comment.
Okay, we can add compression level option together.
Member
Author
There was a problem hiding this comment.
I am going to add the option in #52925 along with Pandas UDF support.
dongjoon-hyun
pushed a commit
that referenced
this pull request
Nov 7, 2025
### What changes were proposed in this pull request? This is an extension to #52747. In #52747, we add the support of Arrow compression to `toArrow` and `toPandas` to reduce memory usage. We would like to extend the memory optimization feature to Pandas UDF case. ### Why are the changes needed? To optimize memory usage for Pandas UDF case. ### Does this PR introduce _any_ user-facing change? No ### How was this patch tested? Unit test ### Was this patch authored or co-authored using generative AI tooling? Generated-by: Claude Code v2.0.14 Closes #52925 from viirya/arrow_compress_udf. Authored-by: Liang-Chi Hsieh <[email protected]> Signed-off-by: Dongjoon Hyun <[email protected]>
dongjoon-hyun
pushed a commit
that referenced
this pull request
Nov 7, 2025
### What changes were proposed in this pull request? This is an extension to #52747. In #52747, we add the support of Arrow compression to `toArrow` and `toPandas` to reduce memory usage. We would like to extend the memory optimization feature to Pandas UDF case. ### Why are the changes needed? To optimize memory usage for Pandas UDF case. ### Does this PR introduce _any_ user-facing change? No ### How was this patch tested? Unit test ### Was this patch authored or co-authored using generative AI tooling? Generated-by: Claude Code v2.0.14 Closes #52925 from viirya/arrow_compress_udf. Authored-by: Liang-Chi Hsieh <[email protected]> Signed-off-by: Dongjoon Hyun <[email protected]> (cherry picked from commit 96ed48d) Signed-off-by: Dongjoon Hyun <[email protected]>
zifeif2
pushed a commit
to zifeif2/spark
that referenced
this pull request
Nov 22, 2025
### What changes were proposed in this pull request? This is an extension to apache#52747. In apache#52747, we add the support of Arrow compression to `toArrow` and `toPandas` to reduce memory usage. We would like to extend the memory optimization feature to Pandas UDF case. ### Why are the changes needed? To optimize memory usage for Pandas UDF case. ### Does this PR introduce _any_ user-facing change? No ### How was this patch tested? Unit test ### Was this patch authored or co-authored using generative AI tooling? Generated-by: Claude Code v2.0.14 Closes apache#52925 from viirya/arrow_compress_udf. Authored-by: Liang-Chi Hsieh <[email protected]> Signed-off-by: Dongjoon Hyun <[email protected]>
huangxiaopingRD
pushed a commit
to huangxiaopingRD/spark
that referenced
this pull request
Nov 25, 2025
### What changes were proposed in this pull request? This patch proposes some changes to optimize memory usage on Arrow in Spark. It compress Arrow IPC data when serializing. ### Why are the changes needed? We have encountered OOM when loading data and processing them in PySpark through `toArrow` or `toPandas`. The same data could be loaded by PyArrow directly but fails to load through `toArrow` or `toPandas` into PySpark due to OOM issues. ### Does this PR introduce _any_ user-facing change? No ### How was this patch tested? Unit tests. Manually test it locally. ### Was this patch authored or co-authored using generative AI tooling? Generated-by: Claude Code v2.0.13 Closes apache#52747 from viirya/release_buffers. Authored-by: Liang-Chi Hsieh <[email protected]> Signed-off-by: Dongjoon Hyun <[email protected]>
huangxiaopingRD
pushed a commit
to huangxiaopingRD/spark
that referenced
this pull request
Nov 25, 2025
### What changes were proposed in this pull request? This is an extension to apache#52747. In apache#52747, we add the support of Arrow compression to `toArrow` and `toPandas` to reduce memory usage. We would like to extend the memory optimization feature to Pandas UDF case. ### Why are the changes needed? To optimize memory usage for Pandas UDF case. ### Does this PR introduce _any_ user-facing change? No ### How was this patch tested? Unit test ### Was this patch authored or co-authored using generative AI tooling? Generated-by: Claude Code v2.0.14 Closes apache#52925 from viirya/arrow_compress_udf. Authored-by: Liang-Chi Hsieh <[email protected]> Signed-off-by: Dongjoon Hyun <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
6 participants
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.
What changes were proposed in this pull request?
This patch proposes some changes to optimize memory usage on Arrow in Spark. It compress Arrow IPC data when serializing.
Why are the changes needed?
We have encountered OOM when loading data and processing them in PySpark through
toArrowortoPandas. The same data could be loaded by PyArrow directly but fails to load throughtoArrowortoPandasinto PySpark due to OOM issues.Does this PR introduce any user-facing change?
No
How was this patch tested?
Unit tests. Manually test it locally.
Was this patch authored or co-authored using generative AI tooling?
Generated-by: Claude Code v2.0.13