Github user manku-timma commented on a diff in the pull request:
https://github.com/apache/spark/pull/18174#discussion_r125386319
--- Diff:
core/src/main/java/org/apache/spark/shuffle/sort/UnsafeShuffleWriter.java ---
@@ -360,12 +368,10 @@ void forceSorterToSpill() throws IOException {
final OutputStream bos = new BufferedOutputStream(
new FileOutputStream(outputFile),
- (int)
sparkConf.getSizeAsKb("spark.shuffle.unsafe.file.output.buffer", "32k") * 1024);
+ outputBufferSizeInBytes);
--- End diff --
Just to understand what is happening.
1. Shuffle records are written to a serialisation buffer (1M) after
serialisation
2. The serialised buffer is written to in-memory-sorterâs buffer
3. once in-memory sorterâs buffer is full, the data is copied to
sorterâs disk buffer (1M)
4. the sorterâs disk buffer is written out to a buffered output stream
(buffer = 32k)
I am guessing reducing the sorterâs disk buffer (in step 3) is helping
because it triggers fewer writes/allocations in a single call at step 4 (and
allowing more parallelism in writing back to disk and copying of data).
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
