Improve network perf test #2695
Closed
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
christarazi
approved these changes
Jul 23, 2024
qmonnet
added
the
needs-rebase
qmonnet
approved these changes
Jul 24, 2024
pchaigno
approved these changes
Jul 28, 2024
maintainer-s-little-helper
bot
added
the
ready-to-merge
michi-covalent
removed
the
ready-to-merge
Previously, each measurement was exported as separate json object. This resulted in poor user experience when displaying these results in perfdash. Now we aggregate latency/throughput/transaction output based on category so it's easier to spot regressions in CI. Signed-off-by: Marcel Zieba <[email protected]>
Now, you can choose tests to run, throughput/crr/rr. Also, adds options to specify message size. Signed-off-by: Marcel Zieba <[email protected]>
Previously, they didn't explicitly say which test was being run. Signed-off-by: Marcel Zieba <[email protected]>
Signed-off-by: Marcel Zieba <[email protected]>
This is next step for separating perf test from regular connectivity tests. Signed-off-by: Marcel Zieba <[email protected]>
Previously, we were also setting it for TCP_STREAM, which is significantly lowering throughput. This was different from all other tests that were being performed in the past, for example in cilium/cilium-perf-networking Signed-off-by: Marcel Zieba <[email protected]>
marseel
force-pushed
the
impr_perf_test
branch
from
f7488ce to
b348e00
Compare
|
rebased and resolved conflict on CODEOWNERS. |
|
"Lock cilium-cli repo / Lock cilium-cli repo (pull_request)" |
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Outdated suggestions cannot be applied.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.
This PR contains various improvements to net perf testing - more details in commit messages.
Most notable improvement is that we no longer use fixed message size for TCP_STREAM test, which results in way more reasonable results as compared to the previous tests.
First of all, we can finally saturate network throughput and also throughput between pods on the same node is way higher than throughput between pods running on different nodes. Example test results for native routing:
Previously native throughput was showing sad 1500 Mb/s for the same Cilium configuration.
How the new output looks like in perfdash:
Makes it much easier to compare host vs pod throughput / latency etc.