TTFT aware Latency Scorer - #188
Open
Mohammad-nassar10 wants to merge 29 commits into
Open
Conversation
Signed-off-by: Mohammad <mohammad.nassar@ibm.com>
Signed-off-by: Mohammad <mohammad.nassar@ibm.com>
Mohammad-nassar10
marked this pull request as draft
June 22, 2026 11:43
|
Your PR is large. Please consider breaking it into multiple PRs. The |
Signed-off-by: Mohammad <mohammad.nassar@ibm.com>
Signed-off-by: Mohammad <mohammad.nassar@ibm.com>
Signed-off-by: Mohammad <mohammad.nassar@ibm.com>
Signed-off-by: Mohammad <mohammad.nassar@ibm.com>
Signed-off-by: Mohammad <mohammad.nassar@ibm.com>
aradhalevy
reviewed
Jun 24, 2026
aradhalevy
reviewed
Jun 24, 2026
ronenkat
reviewed
Jun 28, 2026
ronenkat
left a comment
Contributor
There was a problem hiding this comment.
The code is missing algorithm style documentation to help understand the intent of the code.
Signed-off-by: Mohammad <mohammad.nassar@ibm.com>
Signed-off-by: Mohammad <mohammad.nassar@ibm.com>
Signed-off-by: Mohammad <mohammad.nassar@ibm.com>
Signed-off-by: Mohammad <mohammad.nassar@ibm.com>
Signed-off-by: Mohammad <mohammad.nassar@ibm.com>
Signed-off-by: Mohammad <mohammad.nassar@ibm.com>
Signed-off-by: Mohammad <mohammad.nassar@ibm.com>
Signed-off-by: Mohammad <mohammad.nassar@ibm.com>
ronenkat
reviewed
Jul 2, 2026
Signed-off-by: Mohammad-nassar10 <79787844+Mohammad-nassar10@users.noreply.github.com>
Signed-off-by: Mohammad <mohammad.nassar@ibm.com>
Signed-off-by: Mohammad <mohammad.nassar@ibm.com>
Signed-off-by: Mohammad <mohammad.nassar@ibm.com>
Signed-off-by: Mohammad-nassar10 <79787844+Mohammad-nassar10@users.noreply.github.com>
Signed-off-by: Mohammad <mohammad.nassar@ibm.com>
Signed-off-by: Mohammad <mohammad.nassar@ibm.com>
Signed-off-by: Mohammad <mohammad.nassar@ibm.com>
Signed-off-by: Mohammad <mohammad.nassar@ibm.com>
Signed-off-by: Mohammad <mohammad.nassar@ibm.com>
|
This PR is marked as stale after 21d of inactivity. After an additional 14d of inactivity (7d to become rotten, then 7d more), it will be closed. To prevent this PR from being closed, add a comment or remove the |
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
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.
What type of PR is this?
/kind feature
What this PR does / why we need it:
Adds a new
ttft-aware-scorerthat routes each request to the model with the lowest predicted TTFT under current load.It reads the per-model snapshot published by the
ttft-percentile-extractor— the service floorP10Low, the operating pointsP25/P50, and their in-flight anchorsinflightAtP25/inflightAtP50— and predicts the effective TTFT for a request arriving now as a line through the high operating point(inflightAtP50, P50)and a low anchor blended between the in-cloud point(inflightAtP25, P25)and the load-free floor(0, P10Low):Models are then scored
(maxTTFT - effectiveTTFT) / (maxTTFT - minTTFT). Under-observed (uncalibrated) models are seeded optimistically at the best observed TTFT; with optionalexplorationRate, each under-observed model is independently given that probability of being forced to the top score so it receives a calibration probe — preventing a cold pool from being starved by an already-calibrated one.Parameters:
explorationRate(default0, disabled) andanchorGapScale(default2).Depends on PR #204.
Which issue(s) this PR fixes:
Fixes #54.
Release note (write
NONEif no user-facing change):