[action] [PR:16164] [dualtor] Improve mux_simulator
#16369
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
What is the motivation for this PR? The dualtor nightly suffers from mux simulator timeout issue, and there are always HTTP timeout failures observed. This PR tries to improve the mux simulator performance: improve the all mux toggle performance. improve the mux simulator read/write throughput. PR 1522 was a quick fix to address, but it was a temporary quick fix. How did you do it? run mux simulator with gunicorn instead of its own built-in HTTP server. The mux simulator is running with Flask's own built-in HTTP server. Previously, the mux simulator is running with single-threaded mode, which limits its performance && throughput; and the mux simulator is observed stuck in reading from dead connection; PR 1522 proposes a temporary by running mux simulator in threaded mode. The throughput is improved with the threaded approach, but the built-in server limits the tcp listen backlog to 128, and it is designed for development/test purpose and not recommended to be deployed(Flask's deployment doc). So let's run the mux simulator with gunicorn: better performance/throughput with customized worker count increased tcp listen backlog use thread pool to parallel the toggle requests. The mux simulator handles the toggle-all request by toggling each mux port one by one, let's use a thread pool to parallel run thoses toggles to further decrease the response time. How did you verify/test it? Run the following benchmarks on a dualtor-120 testbed, and compare the performance of: A: the original mux simulator, with Flask built-in server in single-thread mode. B: the mux simulator with Flask built-in server in threaded mode. C: the mux simulator with this PR. toggle mux status for all mux ports(one request to toggle one mux port): 20 concurrent users, repeated 2000 times mux simulator version A B C elapse time 96s 37s 36s toggle mux status for all mux ports(one request to toggle all mux ports): 1 user, repeated 1 time. mux simulator version A B C elapse time 16s 16s 7s To summarize, mux simulator with this PR has the best performance in toggles. Any platform specific information? Supported testbed topology if it's a new test case? Signed-off-by: Longxiang Lyu <[email protected]>
|
/azp run |
8 tasks
|
Original PR: #16164 |
|
Azure Pipelines successfully started running 1 pipeline(s). |
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.
Description of PR
Summary:
Fixes # (issue)
Type of change
Back port request
Approach
What is the motivation for this PR?
The dualtor nightly suffers from mux simulator timeout issue, and there are always HTTP timeout failures observed.
This PR tries to improve the mux simulator performance:
PR 1522 was a quick fix to address, but it was a temporary quick fix.
How did you do it?
gunicorninstead of its own built-in HTTP server.The mux simulator is running with
Flask's own built-in HTTP server. Previously, the mux simulator is running with single-threaded mode, which limits its performance && throughput; and the mux simulator is observed stuck in reading from dead connection; PR 1522 proposes a temporary by running mux simulator in threaded mode. The throughput is improved with the threaded approach, but the built-in server limits the tcp listen backlog to 128, and it is designed for development/test purpose and not recommended to be deployed(Flask's deployment doc).So let's run the mux simulator with
gunicorn:The mux simulator handles the toggle-all request by toggling each mux port one by one, let's use a thread pool to parallel run thoses toggles to further decrease the response time.
How did you verify/test it?
Run the following benchmarks on a dualtor-120 testbed, and compare the performance of:
Flaskbuilt-in server in single-thread mode.Flaskbuilt-in server in threaded mode.To summarize, mux simulator with this PR has the best performance in toggles.
Any platform specific information?
Supported testbed topology if it's a new test case?
Documentation