[CELEBORN-2062] Introduce splitNum to optimize shuffle write#3363
Open
littlexyw wants to merge 7 commits intoapache:mainfrom
Open
[CELEBORN-2062] Introduce splitNum to optimize shuffle write#3363littlexyw wants to merge 7 commits intoapache:mainfrom
littlexyw wants to merge 7 commits intoapache:mainfrom
Conversation
|
This PR is stale because it has been open 20 days with no activity. Remove stale label or comment or this will be closed in 10 days. |
added 2 commits
August 20, 2025 17:12
|
This PR is stale because it has been open 20 days with no activity. Remove stale label or comment or this will be closed in 10 days. |
|
This PR is stale because it has been open 20 days with no activity. Remove stale label or comment or this will be closed in 10 days. |
|
This issue was closed because it has been staled for 10 days with no activity. |
|
This PR is stale because it has been open 20 days with no activity. Remove stale label or comment or this will be closed in 10 days. |
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
4 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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?
Introduce splitNum to optimize shuffle write.
Why are the changes needed?
Currently, map tasks can only write data to one partition split at a time, which will cause a significant performance regression in the case of partition skew. Therefore, the splitNum configuration is introduced to support the generation of multiple partition splits while revive and assign these partition splits to different map tasks to write, so as to improve the writing concurrency.For example, if splitNum is 2 and num mappers is 8, then it turns to
These leaf nodes represent the latestPartitionLocations. Only leaf nodes can be written.
When the splitStart of a partition split equals splitEnd, the splitNum becomes 1. Additionally, a maxWriteParallelism is introduced to control the maximum concurrency. By setting the initial partition split's splitEnd to maxWriteParallelism - 1, the maximum number of leaf nodes can be controlled to be maxWriteParallelism. If maxWriteParallelism is not configured, the actual maximum number of leaf nodes will equal the number of mappers.
When a map task requests to revive a PartitionLocation, the driver first checks if this partition split is already undergoing revival. If true, it records the request. If not, the driver checks for child partition splits. If children exist, it randomly selects one child node (recursively traversing down the hierarchy until a leaf node is found) and returns it to the map task. If no children exist, the driver initiates the revive process.
Does this PR introduce any user-facing change?
No
How was this patch tested?
ut