Repository navigation
perf(detection): charge the scan budget per pass and cut monitor overhead - #152
Merged
Merged
Conversation
…e engine The statistical check used the absolute z-score, so a faster-than-average sample was flagged. It is now one-sided (z > threshold) with the early return at or below the mean, reads the minimum sample count from detectionMinSamplesForAnomaly (default 30, range 10 to 1000, matching the reference) instead of a fixed 10, and computes mean and standard deviation in allocation-free loops.
…attern The per-request scan time budget read the thread CPU clock around every pattern. It now charges per pass: one read when a pass opens, one at each 2 ms wall-clock slice boundary and one when the pass closes. The per-pattern check reads only a cheap monotonic clock and compares the charged total, so the budget lags the true spend by at most one slice plus one pattern.
github-actions
Bot
added
area: detection-engine
This was referenced Oct 16, 2026
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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 #151
Port of the reference fix in Guard-Core/guard-core#173. ScanTimeBudget charges CPU per pass: one clock reading when a pass opens, a charge at each yieldIfDue slice boundary (charged before the yield and restarted after, so other tasks' CPU is never charged), around the awaits of the opt-in worker pool, and a final charge when the pass closes; a budget without a cap never reads the clock. The monitor now matches the reference: it flags only slower-than-average samples, honours detectionMinSamplesForAnomaly (default 30, 10 to 1000, window max(min samples, 100)), computes the mean once and the deviation in one allocation-free loop, and skips the sanitized copy when no callback is registered. On a ~1 KB JSON-like value plus 10 short header values detect went from 5.11 ms to 4.05 ms of CPU. Tests cover the clock reads per pass, a long pass stopping within about one slice, suspended time not being charged, and the one-sided z-score.