mridulm commented on a change in pull request #32388:
URL: https://github.com/apache/spark/pull/32388#discussion_r655778959
##########
File path:
common/network-shuffle/src/main/java/org/apache/spark/network/shuffle/ExternalBlockHandler.java
##########
@@ -264,6 +265,8 @@ private void checkAuth(TransportClient client, String
appId) {
private final Timer registerExecutorRequestLatencyMillis = new Timer();
// Time latency for processing finalize shuffle merge request latency in ms
private final Timer finalizeShuffleMergeLatencyMillis = new Timer();
+ // Block transfer rate in blocks per second
Review comment:
In spark read side, we treat a single read of `ShuffleBlockBatchId` as a
single block read (from metric point of view).
A [recent
discussion](https://github.com/apache/spark/pull/32140#discussion_r652733447)
of this in push based shuffle here for context (see `ShuffleBlockBatchId` in
`ShuffleBlockFetcherIterator`).
Given that, I am fine with treating a batch read as a single read - given it
would be contiguous (effectively, similar to reading a 'large' block).
Thoughts @dongjoon-hyun ?
--
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]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
