Avoid rate limiter resource count update from iterating over all instances #208
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
kthui
changed the title
rmccorm4
reviewed
May 26, 2023
rmccorm4
reviewed
May 27, 2023
rmccorm4
reviewed
May 27, 2023
rmccorm4
reviewed
May 27, 2023
rmccorm4
reviewed
May 30, 2023
src/rate_limiter.cc
Outdated
| resource_manager_->AddModelInstance(pair_it.first->second.get()); | ||
| const auto& status = resource_manager_->UpdateResourceLimits(); | ||
| const auto& status = | ||
| resource_manager_->AddModelInstance(pair_it.first->second.get()); | ||
| if (!status.IsOk()) { | ||
| resource_manager_->RemoveModelInstance(pair_it.first->second.get()); |
There was a problem hiding this comment.
@kthui it looks like we throw away the status returned by RemoveModelInstance, which in turn calls several other methods that can return detailed status errors.
Can we propagate the error statuses back up via LOG_ERROR in the places where we call this method? It looks like we call it in a few places.
There was a problem hiding this comment.
Yes. Updated: Minor fix and return error on instance remove failure
kthui
force-pushed
the
jacky-rate-limiter-resource
branch
2 times, most recently
from
64f80d6 to
6460553
Compare
tanmayv25
reviewed
Jun 2, 2023
kthui
force-pushed
the
jacky-rate-limiter-resource
branch
2 times, most recently
from
88f7622 to
8fa3e9a
Compare
kthui
force-pushed
the
jacky-rate-limiter-resource
branch
2 times, most recently
from
afb3c1b to
b12ea16
Compare
kthui
force-pushed
the
jacky-rate-limiter-resource
branch
from
b12ea16 to
e664d04
Compare
tanmayv25
approved these changes
Jun 14, 2023
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Outdated suggestions cannot be applied.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 optimize the logic on rate limiter instance max resource update, after an instance is added or removed. The previous logic will iterate over all instances once for each instance added or removed. For adding an instance, the new logic will only compare the resource limit between the new instance and the max resource, and update the max resource if the new instance has a higher limit. For removing an instance, the resource on the instance being removed is compared with the max resource. If the max resource is greater than the resource on the instance being removed, then no update will be performed. If the removed instance has higher or equal limit to the max resource, then the max resource is discarded and re-computed, which iterates over the instances once.
Server PR: triton-inference-server/server#5885