Spawn VID with spawn_blocking
#2369
Merged
Conversation
This file contains 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
shenkeyao
previously approved these changes
Jan 10, 2024
| let vid_disperse = vid.disperse(encoded_transactions.clone()).unwrap(); | ||
|
|
||
| let vid_disperse = spawn_blocking(move || { | ||
| let vid = VidScheme::new(chunk_size, num_quorum_committee, &srs).unwrap(); |
There was a problem hiding this comment.
Can we add the note below (copied from the PR description) as a comment here?
Note that the unwrap for the tokio version just passes on any panics that happen during the thread, so it is "safe", i.e it's not adding a new place we can panic.
shenkeyao
approved these changes
Jan 10, 2024
bfish713
changed the title
spawn_blocking WIPspawn_blocking
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
Closes #2349
This PR:
Spawn the VID calculation on a different thread from the async executor to not block other tasks. From this blog https://ryhl.io/blog/async-what-is-blocking/ from the tokio creator we should not be going more than ~100 micro seconds. VID disperse takes an order of magnitude or longer than this so we should run it on it a dedicated thread for blocking. The blog recommends rayon for cpu tasks but also says if we don't have a lot of CPU work than
spawn_blockingis okI also increased the async threads to two in the just file because I believe the CI infra has 2 threads (except for maybe self hosted). Personally I think we should remove this restriction entirely.
This PR does not:
Look at other places where we block async tasks for cpu heavy work. We should probably audit the code for other places where we might go a long time between awaits in our code. Potentially signature aggregation, qc validation, and serialization.
It also doesn't make the VID disperse run on multiple threads. As mentioned in the task we should not be dividing up the block on the consensus side. The further parallelization will be done in jellyfish.
Key places to review:
vid.rs where we now spawn a task. Note that the
unwrapfor the tokio version just passes on any panics that happen during the thread, so it is "safe", i.e it's not adding a new place we can panic.This should improve the consistency of our CI since we should be able to run faster. I think a key issue with our tests was that all nodes would block each other from running while the leader calculated the vid disperse and blocked the main executor. In a single node setting it's probably not so bad since a leader would probably not have other work to do while it was blocked on vid calculation.