viirya commented on a change in pull request #26085: [SPARK-29434] [Core]
Improve the MapStatuses Serialization Performance
URL: https://github.com/apache/spark/pull/26085#discussion_r335747691
##########
File path: core/src/main/scala/org/apache/spark/MapOutputTracker.scala
##########
@@ -822,18 +830,42 @@ private[spark] object MapOutputTracker extends Logging {
} {
objOut.close()
}
- val arr = out.toByteArray
+
+ val arr: Array[Byte] = {
+ val zos = new ZstdOutputStream(compressedOut)
+ Utils.tryWithSafeFinally {
+ compressedOut.write(DIRECT)
+ // `out.writeTo(zos)` will write the uncompressed data from `out` to
`zos`
+ // without copying to avoid unnecessary allocation and copy of byte[].
+ out.writeTo(zos)
+ } {
+ zos.close()
+ }
+ compressedOut.toByteArray
+ }
if (arr.length >= minBroadcastSize) {
// Use broadcast instead.
// Important arr(0) is the tag == DIRECT, ignore that while
deserializing !
val bcast = broadcastManager.newBroadcast(arr, isLocal)
// toByteArray creates copy, so we can reuse out
Review comment:
This comment is out-of-dated now. Can be removed or updated, I think.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
