Repository navigation
feat: arrival co_correct (zrk-style CO correction) + zrk in bench - #45
Merged
Merged
Conversation
`arrival.co_correct: true` queues permits that find the pool saturated instead of dropping them, so a stall surfaces as latency in `engine_arrival_iteration_ms` / `engine_arrival_queue_wait_ms` (measured from the scheduled start) rather than vanishing into dropped_iterations. The dispatcher also spins cooperatively for the last millisecond before each start, so tokio's ~1ms timer granularity no longer reads as queue wait on an idle target (actix @10k rps: p50 1.69ms -> 0.43ms, zrk 0.43ms). The permit channel is now unbounded (in_flight stays the admission gate in drop mode), and permits still queued when the workers stop are counted in dropped_iterations in both modes. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Anthropic Claude <noreply@anthropic.com>
zrk (HttpArena's open-model, coordinated-omission-corrected generator) joins the throughput and saturation tables in --closed mode, like gcannon. The new `latency` suite holds LAT_RATE req/s (default 10000) over LAT_CONNS connections for LAT_DURATION and compares latency measured from each request's scheduled start: zrk -R vs perfscale `arrival.co_correct: true` vs the default drop-on-saturation arrival, with a Dropped column for the latter. bench.yml installs the zrk 2.3.0 release binary (verified against the release SHA256SUMS.txt) and adds zrk/latency to the default APPS/SUITES. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Anthropic Claude <noreply@anthropic.com>
|
Strix is installed on this repository, but we couldn't run this PR security review because this workspace's trial has ended. Add a card to resume code reviews here. |
🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Anthropic Claude <noreply@anthropic.com>
Merged
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.
Stacked on #44 (gcannon) — retarget to
mainonce #44 merges.arrival.co_correct: trueengine_arrival_iteration_ms/engine_arrival_queue_wait_ms(measured from the scheduled start) instead of vanishing intodropped_iterations.yield_now) for the last millisecond before each start, so tokio's ~1ms timer granularity no longer reads as queue wait on an idle target.in_flightremains the drop-mode gate); permits still queued at the deadline count asdropped_iterationsin both modes.false(k6-compatible drop semantics). Schema + en/ru docs + skill reference updated; new testarrival_co_correct_queues_instead_of_dropping.Measured against HttpArena's actix at 10k rps / 1024 conns (macOS host):
co_correctCost: the spin costs ~¼ core at 5k rps (CPU 5.8s vs 3.5s over 9s).
Bench: zrk
APPSgainszrk:--closedrows in throughput and saturation.latencysuite (LAT_RATE/LAT_CONNS/LAT_DURATION, default 10000/256/20s): zrk-Rvs perfscale-co vs perfscale-drop, CO-corrected p50/p90/p99/max + Dropped + CPU.bench.ymlinstalls the zrk 2.3.0 release binary, verified againstSHA256SUMS.txt.Local smoke run (
SUITES="throughput saturation latency" APPS=zrk, 5k rps): zrk p50 0.15 / p99 1.70 ms, perfscale-co 0.06 / 1.25 ms, perfscale-drop 1.01 / 4.89 ms.🤖 Generated with Claude Code