TL/UCP: add rtt ca for onesided a2a #1242
Draft
+170
−16
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Contributor
|
Skipped: This PR does not contain any of your configured labels: ( |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
What
Extends congestion avoidance already present in onesided alltoall to include RTT to skip and revisit potentially congested peers.
Why ?
The original congestion avoidance calculates a set of tokens to limit the amount of messages in flight, which is useful for medium messages, but has limited effect for large messages.
How ?
When token calculation results in less than 1 token, the algorithm segments messages to calculate a RTT and determine if the RTT is beyond a particular threshold (25%) of the estimated latency. This causes the sender to stop sending to the congested rank and move to the next available rank for a set number of times before forcing the send of the message.