[CELEBORN-1300] Optimize CelebornInputStreamImpl's memory usage #2348
Conversation
This file contains 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
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## main #2348 +/- ##
==========================================
- Coverage 48.90% 48.85% -0.04%
==========================================
Files 207 207
Lines 12965 12965
Branches 1113 1113
==========================================
- Hits 6339 6333 -6
- Misses 6220 6224 +4
- Partials 406 408 +2 ☔ View full report in Codecov by Sentry. |
waitinfuture
changed the title
SteNicholas
reviewed
Mar 4, 2024
client/src/main/java/org/apache/celeborn/client/read/CelebornInputStream.java
Outdated
Show resolved
Hide resolved
SteNicholas
reviewed
Mar 4, 2024
client/src/main/java/org/apache/celeborn/client/read/CelebornInputStream.java
Outdated
Show resolved
Hide resolved
SteNicholas
reviewed
Mar 4, 2024
tests/spark-it/src/test/scala/org/apache/celeborn/tests/spark/SparkTestBase.scala
Outdated
Show resolved
Hide resolved
pan3793
reviewed
Mar 4, 2024
tests/spark-it/src/test/scala/org/apache/celeborn/tests/spark/SparkTestBase.scala
Outdated
Show resolved
Hide resolved
CodingCat
approved these changes
Mar 4, 2024
client/src/main/java/org/apache/celeborn/client/read/CelebornInputStream.java
Outdated
Show resolved
Hide resolved
tests/spark-it/src/test/scala/org/apache/celeborn/tests/spark/SparkTestBase.scala
Outdated
Show resolved
Hide resolved
…nputStream.java Co-authored-by: Nicholas Jiang <programgeek@163.com>
…SparkTestBase.scala Co-authored-by: Nicholas Jiang <programgeek@163.com>
|
Comments addressed, PTAL @CodingCat @SteNicholas @pan3793 |
SteNicholas
approved these changes
Mar 5, 2024
SteNicholas
reviewed
Mar 5, 2024
client/src/main/java/org/apache/celeborn/client/read/CelebornInputStream.java
Show resolved
Hide resolved
SteNicholas
approved these changes
Mar 5, 2024
|
LGTM |
pan3793
reviewed
Mar 5, 2024
| @@ -32,6 +33,10 @@ import org.apache.celeborn.service.deploy.MiniClusterFeature | |||
|
|
|||
| trait SparkTestBase extends AnyFunSuite | |||
| with Logging with MiniClusterFeature with BeforeAndAfterAll with BeforeAndAfterEach { | |||
|
|
|||
| val Spark3OrNewer = SPARK_VERSION >= "3.0" | |||
There was a problem hiding this comment.
not a big deal, in some cases, SPARK_VERSION returns "unknown", and
scala> "unknown" >= "3.0"
res0: Boolean = true
pan3793
approved these changes
Mar 5, 2024
waitinfuture
added a commit
that referenced
this pull request
Mar 5, 2024
### What changes were proposed in this pull request? To avoid too much memory usage when CelebornShuffleReader creates input streams. This PR does the following: 1. Constructor of `CelebornInputStream` does not fetch chunk 2. `compressedBuf` and `rawDataBuf` are created first time `fillBuffer` is called 3. When `fillBuffer` returns false, which means the inputstream is exhausted, `close` is called and resource released 4. `CelebornFetchFailureSuite` is only run for Spark 3.0 and newer ### Why are the changes needed? ditto ### Does this PR introduce _any_ user-facing change? No ### How was this patch tested? GA and e2e test. Closes #2348 from waitinfuture/1300. Lead-authored-by: zky.zhoukeyong <zky.zhoukeyong@alibaba-inc.com> Co-authored-by: Keyong Zhou <waitinfuture@gmail.com> Signed-off-by: zky.zhoukeyong <zky.zhoukeyong@alibaba-inc.com> (cherry picked from commit 8b6bc35) Signed-off-by: zky.zhoukeyong <zky.zhoukeyong@alibaba-inc.com>
|
Thanks! Merged to main(0.5.0)/branch-0.4(v0.4.1) |
cfmcgrady
pushed a commit
to cfmcgrady/incubator-celeborn
that referenced
this pull request
May 31, 2024
To avoid too much memory usage when CelebornShuffleReader creates input streams. This PR does the following: 1. Constructor of `CelebornInputStream` does not fetch chunk 2. `compressedBuf` and `rawDataBuf` are created first time `fillBuffer` is called 3. When `fillBuffer` returns false, which means the inputstream is exhausted, `close` is called and resource released 4. `CelebornFetchFailureSuite` is only run for Spark 3.0 and newer ditto No GA and e2e test. Closes apache#2348 from waitinfuture/1300. Lead-authored-by: zky.zhoukeyong <zky.zhoukeyong@alibaba-inc.com> Co-authored-by: Keyong Zhou <waitinfuture@gmail.com> Signed-off-by: zky.zhoukeyong <zky.zhoukeyong@alibaba-inc.com> (cherry picked from commit 8b6bc35) Signed-off-by: zky.zhoukeyong <zky.zhoukeyong@alibaba-inc.com>
RexXiong
pushed a commit
that referenced
this pull request
Jul 18, 2024
…ream ### What changes were proposed in this pull request? #2348 avoids fetching first chunk in the constructor of `CelebornInputStreamImpl`, but in some cases, i.e. coalescing 3000 partitions into one in Spark, it can be beneficial to do so for performance. This PR adds back prefetching with knobs default to false. ### Why are the changes needed? ### Does this PR introduce _any_ user-facing change? Yes, two configs are added. ### How was this patch tested? Extended `MemorySkewJoinSuite` and `ReusedExchangeSuite`, and manual test. Closes #2549 from waitinfuture/1446. Authored-by: zky.zhoukeyong <zky.zhoukeyong@alibaba-inc.com> Signed-off-by: Shuang <lvshuang.xjs@alibaba-inc.com>
RexXiong
pushed a commit
that referenced
this pull request
Jul 18, 2024
…ream ### What changes were proposed in this pull request? #2348 avoids fetching first chunk in the constructor of `CelebornInputStreamImpl`, but in some cases, i.e. coalescing 3000 partitions into one in Spark, it can be beneficial to do so for performance. This PR adds back prefetching with knobs default to false. ### Why are the changes needed? ### Does this PR introduce _any_ user-facing change? Yes, two configs are added. ### How was this patch tested? Extended `MemorySkewJoinSuite` and `ReusedExchangeSuite`, and manual test. Closes #2549 from waitinfuture/1446. Authored-by: zky.zhoukeyong <zky.zhoukeyong@alibaba-inc.com> Signed-off-by: Shuang <lvshuang.xjs@alibaba-inc.com> (cherry picked from commit d692e49) Signed-off-by: Shuang <lvshuang.xjs@alibaba-inc.com>
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.
What changes were proposed in this pull request?
To avoid too much memory usage when CelebornShuffleReader creates input streams.
This PR does the following:
CelebornInputStreamdoes not fetch chunkcompressedBufandrawDataBufare created first timefillBufferis calledfillBufferreturns false, which means the inputstream is exhausted,closeis called and resource releasedCelebornFetchFailureSuiteis only run for Spark 3.0 and newerWhy are the changes needed?
ditto
Does this PR introduce any user-facing change?
No
How was this patch tested?
GA and e2e test.