Skip to content

[Rust] Add tokio-ws - WebSocket#814

Merged
MDA2AV merged 2 commits into
mainfrom
add-tokio-ws
Jun 6, 2026
Merged

[Rust] Add tokio-ws - WebSocket#814
MDA2AV merged 2 commits into
mainfrom
add-tokio-ws

Conversation

@MDA2AV
Copy link
Copy Markdown
Owner

@MDA2AV MDA2AV commented Jun 6, 2026

Description

A WebSocket echo server written directly on raw tokio, with the WebSocket protocol hand-rolled — no tokio-tungstenite, no wtx, no WS library.

Added as an engine-tier entry: the lowest-level way to serve echo-ws on the tokio reactor, so the leaderboard shows what the runtime itself does once framing is reduced to the minimum.

Hand-rolled

  • RFC 6455 handshake — request parsing, Sec-WebSocket-Accept derivation, and the 101 reply, with from-scratch SHA-1 + base64. The only dependencies are tokio and socket2.
  • Frame codec — streaming parser for 7/16/64-bit lengths, client→server unmasking, and partial frames split across reads. Echoes are re-emitted as unmasked server frames preserving FIN + opcode (so fragmentation passes through transparently). Ping→Pong, Close echoed.

Serving model

One current_thread runtime per core, each binding :8080 with SO_REUSEPORT (kernel-sharded accept, no cross-core work-stealing). TCP_NODELAY per connection; outgoing echoes are batched per read so a pipelined burst flushes in one write. Same shape as the other one-thread-per-core engine entries (e.g. rust-epoll).

Tests

Subscribes to echo-ws and echo-ws-pipeline. Passes validate-ws.py (7/7) locally and via the Docker build.

🤖 Generated with Claude Code

@MDA2AV @claude
[Rust] Add tokio-ws - WebSocket
e7579d4 
Hand-rolled WebSocket echo server on raw tokio, with no WebSocket
library. The RFC 6455 handshake (including from-scratch SHA-1 + base64),
frame parser/masking, and echo write path are implemented directly on a
tokio TcpStream. Serving model is one current_thread runtime per core
with SO_REUSEPORT sharding, matching the other engine-tier entries.

Subscribes to echo-ws and echo-ws-pipeline. Passes validate-ws.py (7/7)
locally and via the Docker build.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@MDA2AV
Copy link
Copy Markdown
Owner Author

MDA2AV commented Jun 6, 2026

/benchmark -f tokio-ws

@github-actions
Copy link
Copy Markdown
Contributor

github-actions Bot commented Jun 6, 2026

👋 /benchmark request received. A collaborator will review and approve the run.

@github-actions
Copy link
Copy Markdown
Contributor

github-actions Bot commented Jun 6, 2026

Benchmark Results

Framework: tokio-ws | Test: all tests

Test Conn RPS CPU Mem Δ RPS Δ Mem
echo-ws 512 4,439,600 6178.4% 44MiB NEW NEW
echo-ws 4096 4,504,517 6413.0% 173MiB NEW NEW
echo-ws 16384 4,220,007 6404.0% 592MiB NEW NEW
echo-ws-pipeline 512 66,801,988 6204.4% 40MiB NEW NEW
echo-ws-pipeline 4096 67,381,971 6423.2% 168MiB NEW NEW
echo-ws-pipeline 16384 62,045,219 6382.1% 593MiB NEW NEW
Full log 
  Thread Stats   Avg      p50      p90      p99    p99.9
    Latency    123us    116us    187us    264us    489us

  329150112 frames sent in 5.00s, 329150079 frames received
  Throughput: 65.81M req/s
  Bandwidth:  439.33MB/s
  WS upgrades: 512
  WS frames:   329150079
  Latency samples: 329149040 / 329150079 responses (100.0%)
[info] CPU 6322.9% | Mem 43MiB

[run 3/3]
gcannon v0.5.3 [WS]
  Target:    localhost:8080/ws
  Threads:   64
  Conns:     512 (8/thread)
  Pipeline:  16
  Req/conn:  unlimited (keep-alive)
  Expected:  200
  Duration:  5s


  Thread Stats   Avg      p50      p90      p99    p99.9
    Latency    121us    117us    173us    227us    541us

  334009968 frames sent in 5.00s, 334009942 frames received
  Throughput: 66.78M req/s
  Bandwidth:  445.82MB/s
  WS upgrades: 512
  WS frames:   334009942
  Latency samples: 334008672 / 334009942 responses (100.0%)
[info] CPU 6204.4% | Mem 40MiB

=== Best: 66801988 req/s (CPU: 6204.4%, Mem: 40MiB) ===
[info] saved results/echo-ws-pipeline/512/tokio-ws.json
httparena-bench-tokio-ws
httparena-bench-tokio-ws

==============================================
=== tokio-ws / echo-ws-pipeline / 4096c (tool=gcannon) ===
==============================================
[info] ws-only framework — skipping HTTP probe (sleep 2s for startup)

[run 1/3]
gcannon v0.5.3 [WS]
  Target:    localhost:8080/ws
  Threads:   64
  Conns:     4096 (64/thread)
  Pipeline:  16
  Req/conn:  unlimited (keep-alive)
  Expected:  200
  Duration:  5s


  Thread Stats   Avg      p50      p90      p99    p99.9
    Latency    961us    950us   1.16ms   1.35ms   1.49ms

  336851312 frames sent in 5.00s, 336789136 frames received
  Throughput: 67.33M req/s
  Bandwidth:  449.57MB/s
  WS upgrades: 4096
  WS frames:   336789136
  Latency samples: 336789136 / 336789136 responses (100.0%)
[info] CPU 6123.6% | Mem 151MiB

[run 2/3]
gcannon v0.5.3 [WS]
  Target:    localhost:8080/ws
  Threads:   64
  Conns:     4096 (64/thread)
  Pipeline:  16
  Req/conn:  unlimited (keep-alive)
  Expected:  200
  Duration:  5s


  Thread Stats   Avg      p50      p90      p99    p99.9
    Latency    972us    963us   1.11ms   1.32ms   4.00ms

  336909856 frames sent in 5.00s, 336909856 frames received
  Throughput: 67.34M req/s
  Bandwidth:  449.65MB/s
  WS upgrades: 4096
  WS frames:   336909856
  Latency samples: 336908288 / 336909856 responses (100.0%)
[info] CPU 6423.2% | Mem 168MiB

[run 3/3]
gcannon v0.5.3 [WS]
  Target:    localhost:8080/ws
  Threads:   64
  Conns:     4096 (64/thread)
  Pipeline:  16
  Req/conn:  unlimited (keep-alive)
  Expected:  200
  Duration:  5s


  Thread Stats   Avg      p50      p90      p99    p99.9
    Latency    975us    957us   1.15ms   1.36ms   4.10ms

  335876320 frames sent in 5.00s, 335876013 frames received
  Throughput: 67.15M req/s
  Bandwidth:  448.25MB/s
  WS upgrades: 4096
  WS frames:   335876013
  Latency samples: 335874528 / 335876013 responses (100.0%)
[info] CPU 6139.4% | Mem 173MiB

=== Best: 67381971 req/s (CPU: 6423.2%, Mem: 168MiB) ===
[info] saved results/echo-ws-pipeline/4096/tokio-ws.json
httparena-bench-tokio-ws
httparena-bench-tokio-ws

==============================================
=== tokio-ws / echo-ws-pipeline / 16384c (tool=gcannon) ===
==============================================
[info] ws-only framework — skipping HTTP probe (sleep 2s for startup)

[run 1/3]
gcannon v0.5.3 [WS]
  Target:    localhost:8080/ws
  Threads:   64
  Conns:     16384 (256/thread)
  Pipeline:  16
  Req/conn:  unlimited (keep-alive)
  Expected:  200
  Duration:  5s


  Thread Stats   Avg      p50      p90      p99    p99.9
    Latency   4.24ms   4.21ms   4.63ms   5.43ms   13.60ms

  294874976 frames sent in 5.00s, 294612832 frames received
  Throughput: 58.90M req/s
  Bandwidth:  393.57MB/s
  WS upgrades: 16384
  WS frames:   294612832
  Latency samples: 294612832 / 294612832 responses (100.0%)
[info] CPU 5880.8% | Mem 553MiB

[run 2/3]
gcannon v0.5.3 [WS]
  Target:    localhost:8080/ws
  Threads:   64
  Conns:     16384 (256/thread)
  Pipeline:  16
  Req/conn:  unlimited (keep-alive)
  Expected:  200
  Duration:  5s


  Thread Stats   Avg      p50      p90      p99    p99.9
    Latency   4.19ms   4.20ms   4.61ms   5.32ms   10.60ms

  310488240 frames sent in 5.00s, 310226096 frames received
  Throughput: 62.02M req/s
  Bandwidth:  414.41MB/s
  WS upgrades: 16384
  WS frames:   310226096
  Latency samples: 310226096 / 310226096 responses (100.0%)
[info] CPU 6382.1% | Mem 593MiB

[run 3/3]
gcannon v0.5.3 [WS]
  Target:    localhost:8080/ws
  Threads:   64
  Conns:     16384 (256/thread)
  Pipeline:  16
  Req/conn:  unlimited (keep-alive)
  Expected:  200
  Duration:  5s


  Thread Stats   Avg      p50      p90      p99    p99.9
    Latency   4.22ms   4.19ms   4.58ms   7.42ms   10.70ms

  308432640 frames sent in 5.00s, 308170496 frames received
  Throughput: 61.61M req/s
  Bandwidth:  411.72MB/s
  WS upgrades: 16384
  WS frames:   308170496
  Latency samples: 308170496 / 308170496 responses (100.0%)
[info] CPU 5961.9% | Mem 615MiB

=== Best: 62045219 req/s (CPU: 6382.1%, Mem: 593MiB) ===
[info] saved results/echo-ws-pipeline/16384/tokio-ws.json
httparena-bench-tokio-ws
httparena-bench-tokio-ws
[info] rebuilding site/data/*.json
[updated] /home/diogo/actions-runner/_work/HttpArena/HttpArena/site/data/frameworks.json
[updated] /home/diogo/actions-runner/_work/HttpArena/HttpArena/site/data/echo-ws-16384.json
[updated] /home/diogo/actions-runner/_work/HttpArena/HttpArena/site/data/echo-ws-4096.json
[updated] /home/diogo/actions-runner/_work/HttpArena/HttpArena/site/data/echo-ws-512.json
[updated] /home/diogo/actions-runner/_work/HttpArena/HttpArena/site/data/echo-ws-pipeline-16384.json
[updated] /home/diogo/actions-runner/_work/HttpArena/HttpArena/site/data/echo-ws-pipeline-4096.json
[updated] /home/diogo/actions-runner/_work/HttpArena/HttpArena/site/data/echo-ws-pipeline-512.json
[updated] /home/diogo/actions-runner/_work/HttpArena/HttpArena/site/data/current.json
[info] done
[info] restoring loopback MTU to 65536

@MDA2AV
Copy link
Copy Markdown
Owner Author

MDA2AV commented Jun 6, 2026

/benchmark -f tokio-ws --save

@github-actions
Copy link
Copy Markdown
Contributor

github-actions Bot commented Jun 6, 2026

👋 /benchmark request received. A collaborator will review and approve the run.

@github-actions
Copy link
Copy Markdown
Contributor

github-actions Bot commented Jun 6, 2026

Benchmark Results

Framework: tokio-ws | Test: all tests

Test Conn RPS CPU Mem Δ RPS Δ Mem
echo-ws 512 4,352,944 6332.2% 43MiB NEW NEW
echo-ws 4096 4,433,246 6228.7% 176MiB NEW NEW
echo-ws 16384 4,143,535 6399.5% 590MiB NEW NEW
echo-ws-pipeline 512 65,856,262 6216.1% 42MiB NEW NEW
echo-ws-pipeline 4096 66,930,970 6240.6% 173MiB NEW NEW
echo-ws-pipeline 16384 61,865,334 6397.0% 592MiB NEW NEW
Full log 

  328492192 frames sent in 5.00s, 328529408 frames received
  Throughput: 65.67M req/s
  Bandwidth:  438.33MB/s
  WS upgrades: 512
  WS frames:   328529408
  Latency samples: 328490624 / 328529408 responses (100.0%)
[info] CPU 6359.4% | Mem 42MiB

[run 3/3]
gcannon v0.5.3 [WS]
  Target:    localhost:8080/ws
  Threads:   64
  Conns:     512 (8/thread)
  Pipeline:  16
  Req/conn:  unlimited (keep-alive)
  Expected:  200
  Duration:  5s


  Thread Stats   Avg      p50      p90      p99    p99.9
    Latency    123us    115us    184us    238us    630us

  329281312 frames sent in 5.00s, 329281312 frames received
  Throughput: 65.84M req/s
  Bandwidth:  439.66MB/s
  WS upgrades: 512
  WS frames:   329281312
  Latency samples: 329403056 / 329281312 responses (100.0%)
[info] CPU 6216.1% | Mem 42MiB

=== Best: 65856262 req/s (CPU: 6216.1%, Mem: 42MiB) ===
[info] saved results/echo-ws-pipeline/512/tokio-ws.json
httparena-bench-tokio-ws
httparena-bench-tokio-ws

==============================================
=== tokio-ws / echo-ws-pipeline / 4096c (tool=gcannon) ===
==============================================
[info] ws-only framework — skipping HTTP probe (sleep 2s for startup)

[run 1/3]
gcannon v0.5.3 [WS]
  Target:    localhost:8080/ws
  Threads:   64
  Conns:     4096 (64/thread)
  Pipeline:  16
  Req/conn:  unlimited (keep-alive)
  Expected:  200
  Duration:  5s


  Thread Stats   Avg      p50      p90      p99    p99.9
    Latency    969us    972us   1.14ms   1.28ms   1.63ms

  334087376 frames sent in 5.00s, 334024912 frames received
  Throughput: 66.77M req/s
  Bandwidth:  445.85MB/s
  WS upgrades: 4096
  WS frames:   334024912
  Latency samples: 334024912 / 334024912 responses (100.0%)
[info] CPU 6097.9% | Mem 153MiB

[run 2/3]
gcannon v0.5.3 [WS]
  Target:    localhost:8080/ws
  Threads:   64
  Conns:     4096 (64/thread)
  Pipeline:  16
  Req/conn:  unlimited (keep-alive)
  Expected:  200
  Duration:  5s


  Thread Stats   Avg      p50      p90      p99    p99.9
    Latency    986us    966us   1.16ms   1.32ms   4.10ms

  332086672 frames sent in 5.00s, 332085536 frames received
  Throughput: 66.40M req/s
  Bandwidth:  443.28MB/s
  WS upgrades: 4085
  WS frames:   332085536
  Latency samples: 332084336 / 332085536 responses (100.0%)
[info] CPU 6428.8% | Mem 167MiB

[run 3/3]
gcannon v0.5.3 [WS]
  Target:    localhost:8080/ws
  Threads:   64
  Conns:     4096 (64/thread)
  Pipeline:  16
  Req/conn:  unlimited (keep-alive)
  Expected:  200
  Duration:  5s


  Thread Stats   Avg      p50      p90      p99    p99.9
    Latency    978us    974us   1.17ms   1.32ms   1.49ms

  334648352 frames sent in 5.00s, 334654854 frames received
  Throughput: 66.90M req/s
  Bandwidth:  446.59MB/s
  WS upgrades: 4096
  WS frames:   334644246
  Latency samples: 334656608 / 334654854 responses (100.0%)

  WARNING: 10608/334654854 responses (0.0%) had unexpected status (expected 2xx)
[info] CPU 6240.6% | Mem 173MiB

=== Best: 66930970 req/s (CPU: 6240.6%, Mem: 173MiB) ===
[info] saved results/echo-ws-pipeline/4096/tokio-ws.json
httparena-bench-tokio-ws
httparena-bench-tokio-ws

==============================================
=== tokio-ws / echo-ws-pipeline / 16384c (tool=gcannon) ===
==============================================
[info] ws-only framework — skipping HTTP probe (sleep 2s for startup)

[run 1/3]
gcannon v0.5.3 [WS]
  Target:    localhost:8080/ws
  Threads:   64
  Conns:     16384 (256/thread)
  Pipeline:  16
  Req/conn:  unlimited (keep-alive)
  Expected:  200
  Duration:  5s


  Thread Stats   Avg      p50      p90      p99    p99.9
    Latency   4.24ms   4.25ms   4.64ms   5.01ms   7.31ms

  296918336 frames sent in 5.00s, 296660016 frames received
  Throughput: 59.30M req/s
  Bandwidth:  396.29MB/s
  WS upgrades: 16384
  WS frames:   296660016
  Latency samples: 296660016 / 296660016 responses (100.0%)
[info] CPU 5918.9% | Mem 552MiB

[run 2/3]
gcannon v0.5.3 [WS]
  Target:    localhost:8080/ws
  Threads:   64
  Conns:     16384 (256/thread)
  Pipeline:  16
  Req/conn:  unlimited (keep-alive)
  Expected:  200
  Duration:  5s


  Thread Stats   Avg      p50      p90      p99    p99.9
    Latency   4.22ms   4.26ms   4.64ms   5.03ms   7.79ms

  309588816 frames sent in 5.00s, 309326672 frames received
  Throughput: 61.85M req/s
  Bandwidth:  413.28MB/s
  WS upgrades: 16384
  WS frames:   309326672
  Latency samples: 309326672 / 309326672 responses (100.0%)
[info] CPU 6397.0% | Mem 592MiB

[run 3/3]
gcannon v0.5.3 [WS]
  Target:    localhost:8080/ws
  Threads:   64
  Conns:     16384 (256/thread)
  Pipeline:  16
  Req/conn:  unlimited (keep-alive)
  Expected:  200
  Duration:  5s


  Thread Stats   Avg      p50      p90      p99    p99.9
    Latency   4.25ms   4.28ms   4.66ms   5.10ms   10.20ms

  306241088 frames sent in 5.00s, 305978944 frames received
  Throughput: 61.16M req/s
  Bandwidth:  408.67MB/s
  WS upgrades: 16384
  WS frames:   305978944
  Latency samples: 305978944 / 305978944 responses (100.0%)
[info] CPU 6077.5% | Mem 618MiB

=== Best: 61865334 req/s (CPU: 6397.0%, Mem: 592MiB) ===
[info] saved results/echo-ws-pipeline/16384/tokio-ws.json
httparena-bench-tokio-ws
httparena-bench-tokio-ws
[info] rebuilding site/data/*.json
[updated] /home/diogo/actions-runner/_work/HttpArena/HttpArena/site/data/frameworks.json
[updated] /home/diogo/actions-runner/_work/HttpArena/HttpArena/site/data/echo-ws-16384.json
[updated] /home/diogo/actions-runner/_work/HttpArena/HttpArena/site/data/echo-ws-4096.json
[updated] /home/diogo/actions-runner/_work/HttpArena/HttpArena/site/data/echo-ws-512.json
[updated] /home/diogo/actions-runner/_work/HttpArena/HttpArena/site/data/echo-ws-pipeline-16384.json
[updated] /home/diogo/actions-runner/_work/HttpArena/HttpArena/site/data/echo-ws-pipeline-4096.json
[updated] /home/diogo/actions-runner/_work/HttpArena/HttpArena/site/data/echo-ws-pipeline-512.json
[updated] /home/diogo/actions-runner/_work/HttpArena/HttpArena/site/data/current.json
[info] done
[info] restoring loopback MTU to 65536

@MDA2AV MDA2AV merged commit c3ac1af into main Jun 6, 2026
@MDA2AV MDA2AV deleted the add-tokio-ws branch June 7, 2026 14:31
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@Kaliumhexacyanoferrat Kaliumhexacyanoferrat Awaiting requested review from Kaliumhexacyanoferrat Kaliumhexacyanoferrat is a code owner

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

1 participant

@MDA2AV