raptor 0.13.0 → 0.14.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: b7a531c4a4d9cfe93de0c4e8d255d9eac6a09d59f9af6d3884231cd0a383aca5
-  data.tar.gz: 3155fd5b27d53b5a65a9be9cc36d139da801ad3ce33a188ea9a46c77289ee97e
+  metadata.gz: 90014b6262f682d50ed48e31fea60c483aeb217fa28f758bcef6eeada39baf98
+  data.tar.gz: 37b4e90995cc4e76319318f14a6c1d73a6d1ab841c97f25c5b0543ef07f52c87
 SHA512:
-  metadata.gz: 53bfe180b9db4a6d9f7fdb3ea2f11bf2f19b5fa81c54beb09e74caa189381a52587756c6acae9b18ce16a5500a278a0b737934bc9b5e129fe3002605ee2ac5c1
-  data.tar.gz: a235a90619e1a8c1935208d1c516280bf911335e075bc6e91d5be78d134c5cc385381dba899328ebc2f8da213753d31618f5328e5eec8d609a06cfb1c30b5cb8
+  metadata.gz: ba970eef596750d5e2cc885018bc6d981edb0ef1479dfade73d66264e54693f513e52e5ec4a630f198b1ba241f121ca976588717b9c15e34360198a8c8bbcb75
+  data.tar.gz: 6aa65a255c612c094b568b252c8f5ac42debabf92e72896a341fd0ac5031960e56255774c750bdafa2e6658261dc2e3104dc0a546e839d99ce1be773d42cd26e

data/CHANGELOG.md

@@ -1,5 +1,10 @@
 ## [Unreleased]
 
+## [0.14.0] - 2026-08-09
+
+- Cap eager HTTP/2 frame consumption at thread pool saturation
+- Bundle HTTP/2 frame writes into a single socket write
+
 ## [0.13.0] - 2026-07-21
 
 - Tune BPF reuseport routing to spread bursts across near-idle workers

data/Dockerfile

@@ -18,6 +18,8 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
     && rm -rf /var/lib/apt/lists/*
 
 COPY --from=hey-builder /go/bin/hey /usr/local/bin/hey
+COPY --from=hey-builder /usr/local/go /usr/local/go
+ENV PATH=/usr/local/go/bin:${PATH}
 
 ENV BUNDLE_PATH=/workspace/.bundle
 ENV BUNDLE_APP_CONFIG=/workspace/.bundle

data/README.md

@@ -37,7 +37,7 @@ run proc { |_env| [200, { "content-type" => "text/plain" }, ["Hello, World!"]] }
 ```
 > bundle exec raptor -w 4 -t 3 hello_world.ru
 [Raptor 59013|Main|Main] Cluster initializing:
-[Raptor 59013|Main|Main] ├─ Version: 0.13.0
+[Raptor 59013|Main|Main] ├─ Version: 0.14.0
 [Raptor 59013|Main|Main] ├─ Ruby Version: ruby 4.0.5 (2026-05-20 revision 64336ffd0e) +YJIT +PRISM [arm64-darwin23]
 [Raptor 59013|Main|Main] ├─ Environment: development
 [Raptor 59013|Main|Main] ├─ Master PID: 59013
@@ -193,29 +193,28 @@ Worker 1 (phase 0): pid=91351, requests=1199, busy=1/3, backlog=0, booted, last_
 
 ## (Micro) Benchmarks
 
-Raptor 0.13.0 vs Puma 8.0.2 vs Falcon 0.55.5 across two workload profiles. **IO-bound** is a GET endpoint that
-interleaves 5-10 short sleeps (total 2.5-15ms) with small CPU work, simulating a read path that makes several DB or
-cache calls. **CPU-bound** is a POST endpoint that accepts a small JSON body, interleaves 3-5 chunks of JSON item
-building (total 450-1500 items) with sub-100µs sleeps, and returns the built array, simulating a write path that does
-most of its work in Ruby with a few near-zero-cost cache hits.
+Raptor 0.14.0 vs Puma 8.0.2 vs PumaPlus 0.1.0 vs Falcon 0.57.0 across two workload profiles. **IO-bound** is a GET
+endpoint that interleaves 5-10 short sleeps (total 2.5-15ms) with small CPU work, simulating a read path that makes
+several DB or cache calls. **CPU-bound** is a POST endpoint that accepts a small JSON body, interleaves 3-5 chunks of
+JSON item building (total 450-1500 items) with sub-100µs sleeps, and returns the built array, simulating a write path
+that does most of its work in Ruby with a few near-zero-cost cache hits.
 
 Each cell reports the median throughput and median p95 latency independently across 5 runs, so the two numbers in a row
 may come from different runs. Every run starts a fresh server process so the samples are independent of each other;
 state accumulated in a previous run cannot bias the next. Across the whole table, the widest spread
-((max - min) / 2 / median) between runs of a single cell was ±30.0% for throughput
-and ±22.8% for p95.
-
-| Protocol              | Workload | Raptor req/s | Raptor p95 | Puma req/s  | Puma p95 | vs Puma req/s | vs Puma p95 | Falcon req/s | Falcon p95 | vs Falcon req/s | vs Falcon p95 |
-| --------------------- | -------- | ------------ | ---------- | ----------- | -------- | ------------- | ----------- | ------------ | ---------- | --------------- | ------------- |
-| HTTP/1.1              | IO       | 1.39k req/s  | 50.20 ms   | 0.96k req/s | 66.20 ms | 44.1% higher  | 24.2% lower | 4.54k req/s  | 15.00 ms   | 69.4% lower     | 234.7% higher |
-| HTTP/1.1              | CPU      | 4.55k req/s  | 18.10 ms   | 3.64k req/s | 16.50 ms | 24.9% higher  | 9.7% higher | 3.87k req/s  | 15.00 ms   | 17.5% higher    | 20.7% higher  |
-| HTTP/1.1 (keep-alive) | IO       | 1.42k req/s  | 30.00 ms   | 0.92k req/s | 65.80 ms | 54.8% higher  | 54.4% lower | 4.08k req/s  | 16.70 ms   | 65.3% lower     | 79.6% higher  |
-| HTTP/1.1 (keep-alive) | CPU      | 4.51k req/s  | 11.20 ms   | 3.77k req/s | 17.50 ms | 19.8% higher  | 36.0% lower | 4.00k req/s  | 16.50 ms   | 12.9% higher    | 32.1% lower   |
-| HTTP/2                | IO       | 0.71k req/s  | 85.84 ms   | N/A         | N/A      | -             | -           | 3.88k req/s  | 17.43 ms   | 81.7% lower     | 392.3% higher |
-| HTTP/2                | CPU      | 4.07k req/s  | 21.85 ms   | N/A         | N/A      | -             | -           | 4.33k req/s  | 30.84 ms   | 6.0% lower      | 29.2% lower   |
-
-> ruby 4.0.5 (2026-05-20 revision 64336ffd0e) +YJIT +PRISM [aarch64-linux]
-> 4 worker processes; Raptor and Puma run 3 threads per worker, Falcon runs unbounded fibers per worker;
+((max - min) / 2 / median) between runs of a single cell was ±35.4% for throughput and ±95.1% for p95.
+
+| Protocol              | Workload | Raptor req/s | Raptor p95 | Puma req/s  | Puma p95  | vs Puma req/s | vs Puma p95 | PumaPlus req/s | PumaPlus p95 | vs PumaPlus req/s | vs PumaPlus p95 | Falcon req/s | Falcon p95 | vs Falcon req/s | vs Falcon p95 |
+| --------------------- | -------- | ------------ | ---------- | ----------- | --------- | ------------- | ----------- | -------------- | ------------ | ----------------- | --------------- | ------------ | ---------- | --------------- | ------------- |
+| HTTP/1.1              | IO       | 1.33k req/s  | 52.20 ms   | 0.48k req/s | 140.60 ms | 175.8% higher | 62.9% lower | 0.48k req/s    | 122.30 ms    | 177.0% higher     | 57.3% lower     | 3.41k req/s  | 21.50 ms   | 60.9% lower     | 142.8% higher |
+| HTTP/1.1              | CPU      | 3.93k req/s  | 20.80 ms   | 1.74k req/s | 39.60 ms  | 125.6% higher | 47.5% lower | 2.32k req/s    | 27.60 ms     | 69.2% higher      | 24.6% lower     | 3.60k req/s  | 16.40 ms   | 9.3% higher     | 26.8% higher  |
+| HTTP/1.1 (keep-alive) | IO       | 1.31k req/s  | 46.00 ms   | 0.46k req/s | 133.80 ms | 183.3% higher | 65.6% lower | 0.46k req/s    | 128.20 ms    | 186.0% higher     | 64.1% lower     | 2.06k req/s  | 35.40 ms   | 36.5% lower     | 29.9% higher  |
+| HTTP/1.1 (keep-alive) | CPU      | 3.48k req/s  | 21.50 ms   | 1.73k req/s | 41.00 ms  | 101.5% higher | 47.6% lower | 2.09k req/s    | 30.40 ms     | 66.0% higher      | 29.3% lower     | 3.74k req/s  | 17.70 ms   | 7.0% lower      | 21.5% higher  |
+| HTTP/2                | IO       | 0.47k req/s  | 161.47 ms  | N/A         | N/A       | -             | -           | 0.52k req/s    | 113.14 ms    | 8.1% lower        | 42.7% higher    | 2.12k req/s  | 34.10 ms   | 77.6% lower     | 373.6% higher |
+| HTTP/2                | CPU      | 2.09k req/s  | 40.26 ms   | N/A         | N/A       | -             | -           | 2.23k req/s    | 29.19 ms     | 5.9% lower        | 37.9% higher    | 2.85k req/s  | 36.93 ms   | 26.4% lower     | 9.0% higher   |
+
+> ruby 4.0.6 (2026-07-14 revision 03b6d3f889) +YJIT +PRISM [aarch64-linux]
+> 4 worker processes; Raptor, Puma, and PumaPlus run 3 threads per worker, Falcon runs unbounded fibers per worker;
 > 48 concurrent HTTP/1.1 client connections; 16 concurrent HTTP/2 client connections × 3 streams each
 
 See [bin/benchmark](bin/benchmark) for more details.