Skip to content

[SPARK-35879][Core][Shuffle] Fix performance regression caused by collectFetchRequests #33063

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
yaooqinn wants to merge 9 commits into from
Closed

[SPARK-35879][Core][Shuffle] Fix performance regression caused by collectFetchRequests #33063

Changes from 3 commits
Commits
Show all changes
9 commits
Select commit Hold shift + click to select a range
1eec196
[SPARK-35879][Core][Shuffle] Fix performance regression caused by col…
yaooqinn Jun 24, 2021
8352ac8
[SPARK-35879][Core][Shuffle] Fix performance regression caused by col…
yaooqinn Jun 24, 2021
de42ae5
[SPARK-35879][Core][Shuffle] Fix performance regression caused by col…
yaooqinn Jun 24, 2021
0f0e183
scala
yaooqinn Jun 24, 2021
3813003
fix test
yaooqinn Jun 24, 2021
8e4696f
fix scala 213
yaooqinn Jun 25, 2021
f894220
pass ArrayBuffer not Seq
yaooqinn Jun 25, 2021
32ac63d
Revert "pass ArrayBuffer not Seq"
yaooqinn Jun 25, 2021
dd78ab8
Merge branch 'master' into SPARK-35879
yaooqinn Jun 25, 2021
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
21 changes: 12 additions & 9 deletions core/src/main/scala/org/apache/spark/storage/ShuffleBlockFetcherIterator.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -380,7 +380,10 @@ final class ShuffleBlockFetcherIterator(
hostLocalBlockBytes += mergedBlockInfos.map(_.size).sum
} else {
remoteBlockBytes += blockInfos.map(_._2).sum
collectFetchRequests(address, blockInfos, collectedRemoteRequests)
val (_, timeCost) = Utils.timeTakenMs {
collectFetchRequests(address, blockInfos, collectedRemoteRequests)
}
logDebug(s"Collected remote fetch requests for $address in $timeCost ms")
}
}
val numRemoteBlocks = collectedRemoteRequests.map(_.blocks.size).sum
Expand Down Expand Up @@ -435,26 +438,26 @@ final class ShuffleBlockFetcherIterator(
collectedRemoteRequests: ArrayBuffer[FetchRequest]): Unit = {
val iterator = blockInfos.iterator
var curRequestSize = 0L
var curBlocks = Seq.empty[FetchBlockInfo]
val curBlocks = new ArrayBuffer[FetchBlockInfo]()

while (iterator.hasNext) {
val (blockId, size, mapIndex) = iterator.next()
assertPositiveBlockSize(blockId, size)
curBlocks = curBlocks ++ Seq(FetchBlockInfo(blockId, size, mapIndex))
curBlocks += FetchBlockInfo(blockId, size, mapIndex)
curRequestSize += size
// For batch fetch, the actual block in flight should count for merged block.
val mayExceedsMaxBlocks = !doBatchFetch && curBlocks.size >= maxBlocksInFlightPerAddress
if (curRequestSize >= targetRemoteRequestSize || mayExceedsMaxBlocks) {
curBlocks = createFetchRequests(curBlocks, address, isLast = false,
collectedRemoteRequests)
val retBlocks =
createFetchRequests(curBlocks, address, isLast = false, collectedRemoteRequests)
curBlocks.clear()
curBlocks ++= retBlocks
curRequestSize = curBlocks.map(_.size).sum
}
}
// Add in the final request
if (curBlocks.nonEmpty) {
curBlocks = createFetchRequests(curBlocks, address, isLast = true,
collectedRemoteRequests)
curRequestSize = curBlocks.map(_.size).sum
createFetchRequests(curBlocks, address, isLast = true, collectedRemoteRequests)
Copy link
Contributor

@cloud-fan cloud-fan Jun 24, 2021

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

why can we skip updating curBlocks and curRequestSize here?

Copy link
Contributor

@cloud-fan cloud-fan Jun 24, 2021

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

ah, we are leaving this method and no need to update local variables here.

}
}

Expand Down Expand Up @@ -994,7 +997,7 @@ object ShuffleBlockFetcherIterator {
blocks: Seq[FetchBlockInfo],
doBatchFetch: Boolean): Seq[FetchBlockInfo] = {
val result = if (doBatchFetch) {
var curBlocks = new ArrayBuffer[FetchBlockInfo]
val curBlocks = new ArrayBuffer[FetchBlockInfo]
val mergedBlockInfo = new ArrayBuffer[FetchBlockInfo]

def mergeFetchBlockInfo(toBeMerged: ArrayBuffer[FetchBlockInfo]): FetchBlockInfo = {
Expand Down