tgravescs commented on a change in pull request #23695:
[SPARK-26780][CORE]Improve shuffle read using ReadAheadInputStream
URL: https://github.com/apache/spark/pull/23695#discussion_r280150222
##########
File path:
core/src/main/scala/org/apache/spark/storage/ShuffleBlockFetcherIterator.scala
##########
@@ -474,6 +475,12 @@ final class ShuffleBlockFetcherIterator(
// TODO: manage the memory used here, and spill it into disk in
case of OOM.
Utils.copyStream(input, out, closeStreams = true)
input = out.toChunkedByteBuffer.toInputStream(dispose = true)
+ } else {
+ val readAheadEnabled = SparkEnv.get != null &&
+ SparkEnv.get.conf.get(config.SHUFFLE_READ_AHEAD)
+ if (readAheadEnabled) {
+ input = new ReadAheadInputStream(input, 1024 * 1024)
Review comment:
yes I think we should document this behavior, also agree with some of the
other comments not sure we need the enabled config, perhaps the buffer size one
and if its set to 0 then disabled. How did you choose the buffer size, just
take it from the existing spill reader config? did you try different buffer
sizes?
----------------------------------------------------------------
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]
