scripts/bench-campfire — workers=1, c=16, 3×20s runs per cell after 20s warmup; the median of the 3 runs is reported (the run-to-run spread is in the stability section below). Quiet machine throughout. Full hardware, toolchain versions, and the exact invocation are in the environment appendix below.
rails and the emit lanes BOTH serve campfire's message fragments from a cache — Rails from Redis (cached: true), the emitted tree from its own process-local store — so the deployed lanes compare like with like. Until 2026-09-07 the emit had no fragment cache and this page carried a rails-nocache column so the comparison was renderer against renderer rather than emit against Redis; that column is now a diagnostic, requested by name, and is not published here.
The deployed lanes are what each project would run on this box. rails is once-campfire AS ITS OWN DOCKERFILE RUNS IT: Thruster in front of Puma, (processor_count * 0.666).ceil workers x 5 threads from its own config/puma.rb, jemalloc preloaded, resque-pool alongside. ruby is the CRuby emit in the same Puma shape with the same allocator; spinel is the binary with its workers autodetected.
ruby IS A MATCHED-SHAPE RUNTIME COMPARISON, NOT A DEPLOYMENT CLAIM. It runs Rails' worker count so the number is not a core count, but the emitted tree's Action Cable registry and its job queue are both in-process: at more than one worker a broadcast reaches only the subscribers sharing a process with the poster, so campfire could not actually run this shape. Its deployable CRuby shape today is one worker; one process serving every core (Falcon on Ractors, the shape spinel already has) is the planned replacement. spinel needs none of this: one process, green threads, one shared registry — which is why it is the emit lane the WebSocket tier can measure at its full worker count.
ruby-falcon is the SAME emitted tree and config.ru as ruby, served by Falcon (the emit's falcon.rb: forked workers, the same count and allocator, cache hits answered ahead of Falcon's Rack adapter) instead of Puma: the server tobi runs. It isolates what the app server costs. Falcon is added to the emit's bundle by this harness, at the version tobi's lock pins; the emitted Gemfile names only Puma, and the WebSocket phase still runs Puma.
Known omissions, stated rather than discovered: the WebSocket phase's rails lane runs Puma with campfire's worker count and its allocator but WITHOUT Thruster and resque-pool, and no lane terminates TLS.
MEMORY IS THE SERVER PROCESS TREE, and the rails lanes keep part of themselves outside it. Their fragment cache, their Action Cable adapter and Resque's queue all live in a system Redis (~16 MB RSS here) that is nobody's descendant and so is in no lane's number — while the emit lanes hold their equivalent inside the measured process. Throughput is unaffected: the rails lanes pay for every Redis round trip in their req/s. Counted for rails: Thruster, the Puma master and its workers, resque-pool and its workers.
tobi is tobi/campfire-once-ruby-ractor: campfire ported BY HAND to plain Ruby, no Rails, served by Falcon from worker Ractors in one process, with in-process job and cable Ractors. It is the measured answer to "how fast can careful Ruby serve this app", and the target for the emitted ruby lane. It reads the same seeded database. Its deliberate differences from Rails (header-only CSRF, so its pages carry no tokens; part-hashed ETags; fewer cookie writes) are listed in its README.
The CRuby fragment store is PER PROCESS, so the ruby lane holds one cache per worker where spinel holds one for the whole binary.
spinel runs the same Roundhouse-emitted framework as the CRuby ruby cell, but compiled by Spinel behind its thread-per-connection server (a green thread per connection on the runtime's autodetected OS worker count) rather than CRuby + Puma — so the ruby → spinel gap is a runtime swap, not a framework or source change.
Each endpoint is its own chart. Bars are log-scaled; raw req/sec is shown at the right.
Same Ruby interpreter, same YJIT, same Puma — the only variable is whether the framework runtime is Rails or Roundhouse-emitted. The multiplier under each pair is the lift from the lowerer pipeline. Bars are log-scaled, so the multiplier, not bar height, is the ratio.
Throughput normalized by memory footprint — req/sec divided by RSS in GB. This reorders the raw throughput charts above: targets with small working sets rise and high-RSS targets fall. Bars are log-scaled. How much the metric matters depends on the deployment shape — most on metered or serverless surfaces, least on bare metal with memory headroom.
Max PSS observed across all endpoints, per target, sorted low to high.
PSS, not a sum of RSS: adding RSS across a process tree counts every shared page once per process, which reads high for a clustered lane and changes nothing for a single-process one. PSS divides each shared page among its sharers.
Latencies are at c=16 concurrent connections; treat p50 as the median per-connection wait and p99 as the tail. Lanes are ordered by measured median p50 across endpoints, fastest first.
| route | ruby-falcon | ruby | spinel | tobi | rails | |||||
|---|---|---|---|---|---|---|---|---|---|---|
| p50 | p99 | p50 | p99 | p50 | p99 | p50 | p99 | p50 | p99 | |
/rooms/1 | 1.3 | 3.3 | 1.3 | 3.5 | 1.6 | 4.9 | 3.8 | 9.0 | 8.0 | 20.6 |
/rooms/1/messages?before=50 | 1.1 | 3.5 | 1.2 | 3.6 | 1.4 | 4.2 | 1.6 | 5.7 | 7.7 | 40.5 |
/users/me/sidebar | 0.8 | 2.2 | 0.9 | 2.7 | 1.3 | 4.5 | 1.6 | 4.0 | 4.7 | 12.6 |
/searches?q=coffee | 1.4 | 3.5 | 1.2 | 4.0 | 1.3 | 4.3 | 4.1 | 10.6 | 7.1 | 46.3 |
/users/:avatar_token/avatar | 0.2 | 0.9 | 0.3 | 0.9 | 0.8 | 3.6 | 0.2 | 3.2 | 0.6 | 2.5 |
a stylesheet (/assets/*.css) | 0.1 | 0.8 | 0.3 | 0.8 | 0.2 | 1.5 | 0.1 | 1.9 | 0.4 | 1.5 |
/up | 0.4 | 1.3 | 0.5 | 2.3 | 0.3 | 1.7 | 0.2 | 2.3 | 3.4 | 9.1 |
POST /rooms/2/messages | 14.9 | 95.9 | 15.3 | 57.8 | 13.4 | 79.5 | 12.1 | 65.8 | 44.6 | 227.6 |
All values in milliseconds. Lower is better.
Time from exec to a server that answers, and to a signed-in /rooms/1 that comes back 200. Median of 3 boots per lane, all on the shape named in the row. The room page is the number that matters: it contains the first render, and on a fragment-cached lane a restart keeps the Redis its own Procfile runs in a separate process, so this is a warm-cache restart rather than a cold deployment.
| target | answers (ms) | room page (ms) |
|---|---|---|
| spinel | 324 | 355 |
| ruby-falcon | 1210 | 1472 |
| tobi | 1372 | 1534 |
| ruby | 1314 | 1597 |
| rails | 4673 | 5562 |
Milliseconds. Lower is better. A deployed Rails lane boots a proxy, a Puma master and one worker per (processor_count × 0.666).ceil; the binary boots one process.
Sockets held open, subscribed to a room, then idled and driven: each lane pinned to the same cores, one client, exact delivered/expected frame counts. The idle column is the cost of holding connections that are doing nothing, which is what a chat deployment does most of the time. Read the fan-out latency against the per-room column — one message costs one write per subscriber in its room, so a room of 200 and a room of 50 are different workloads at the same socket count.
| target | sockets | per room | connect storm | idle (cores) | PSS (MB) | fan-out p50 / p99 | delivering (cores) | frames |
|---|---|---|---|---|---|---|---|---|
| spinel | 1,000 | 50 | 0.4 s | 0.009 | 260 | 19 / 51 ms | 0.022 | 1,500/1,500 |
| ruby | 1,000 | 50 | 1.2 s | 0.006 | 224 | 32 / 138 ms | 0.032 | 1,500/1,500 |
| rails | 1,000 | 50 | 2.5 s | 0.013 | 646 | 38 / 221 ms | 0.078 | 1,500/1,500 |
The driver is one thread reading every socket; its longest single reading pass in this run was 58 ms. A fan-out p99 near that number is the client's backlog rather than the server's fan-out.
Memory is PSS where the lane is multi-process, RSS otherwise. Lower is better in every column except frames, which must be exact.
One authenticated GET per lane per endpoint, captured before any load. Equal tag counts are what makes the throughput numbers above a comparison rather than two measurements.
| lane | endpoint | status | tags | bytes |
|---|---|---|---|---|
| rails | /rooms/1 | 200 | 3,639 | 400,734 |
| rails | /rooms/1/messages?before=50 | 200 | 3,375 | 367,595 |
| rails | /users/me/sidebar | 200 | 330 | 37,354 |
| rails | /searches?q=coffee | 200 | 3,544 | 390,527 |
| rails | /users/:avatar_token/avatar | 200 | 7 | 631 |
| rails | a stylesheet (/assets/*.css) | 200 | 0 | 1,218 |
| rails | /up | 200 | 2 | 73 |
| ruby | /rooms/1 | 200 | 3,639 | 396,226 |
| ruby | /rooms/1/messages?before=50 | 200 | 3,375 | 365,275 |
| ruby | /users/me/sidebar | 200 | 330 | 35,295 |
| ruby | /searches?q=coffee | 200 | 3,544 | 386,163 |
| ruby | /users/:avatar_token/avatar | 200 | 7 | 631 |
| ruby | a stylesheet (/assets/*.css) | 200 | 0 | 1,218 |
| ruby | /up | 200 | 2 | 73 |
| ruby-falcon | /rooms/1 | 200 | 3,639 | 396,226 |
| ruby-falcon | /rooms/1/messages?before=50 | 200 | 3,375 | 365,275 |
| ruby-falcon | /users/me/sidebar | 200 | 330 | 35,295 |
| ruby-falcon | /searches?q=coffee | 200 | 3,544 | 386,163 |
| ruby-falcon | /users/:avatar_token/avatar | 200 | 7 | 631 |
| ruby-falcon | a stylesheet (/assets/*.css) | 200 | 0 | 1,218 |
| ruby-falcon | /up | 200 | 2 | 73 |
| spinel | /rooms/1 | 200 | 3,639 | 396,226 |
| spinel | /rooms/1/messages?before=50 | 200 | 3,375 | 365,275 |
| spinel | /users/me/sidebar | 200 | 330 | 35,295 |
| spinel | /searches?q=coffee | 200 | 3,544 | 386,163 |
| spinel | /users/:avatar_token/avatar | 200 | 7 | 631 |
| spinel | a stylesheet (/assets/*.css) | 200 | 0 | 1,218 |
| spinel | /up | 200 | 2 | 73 |
| tobi | /rooms/1 | 200 | 3,639 | 395,861 |
| tobi | /rooms/1/messages?before=50 | 200 | 3,375 | 362,755 |
| tobi | /users/me/sidebar | 200 | 330 | 37,365 |
| tobi | /searches?q=coffee | 200 | 3,544 | 385,127 |
| tobi | /users/:avatar_token/avatar | 200 | 7 | 631 |
| tobi | a stylesheet (/assets/*.css) | 200 | 0 | 1,218 |
| tobi | /up | 200 | 2 | 73 |
Each cell is timed 3 times and the charts above report the median run. This section reads per-run.json directly to show how far the individual runs strayed from it. Across all 40 cells the median run-to-run coefficient of variation in req/sec is 0.73% — the timed runs of a given cell agree to a fraction of a percent, so the reported medians aren't masking noise.
The 8 cells that vary by more than 3% are concentrated in the lowest-throughput cells, where a small absolute swing is a larger fraction of the rate; in 1 of them the first timed run is the slowest, consistent with residual warmup the fixed 20s warmup doesn't fully absorb:
| target | endpoint | run 1 | run 2 | run 3 | CV |
|---|---|---|---|---|---|
| tobi | /up | 55,957 | 67,038 | 44,255 | 16.7% |
| tobi | {css} | 77,934 | 59,325 | 87,941 | 15.8% |
| ruby-falcon | /users/me/sidebar | 19,502 | 18,605 | 16,733 | 6.3% |
| ruby-falcon | {css} | 87,040 | 79,988 | 92,875 | 6.1% |
| ruby-falcon | /users/{avatar}/avatar | 76,276 | 69,397 | 78,713 | 5.3% |
| ruby-falcon | /searches?q=coffee | 10,683 | 9,703 | 10,822 | 4.8% |
| ruby-falcon | /rooms/1 | 10,177 | 11,345 | 10,851 | 4.4% |
| ruby-falcon | /up | 39,665 | 37,393 | 36,607 | 3.4% |
req/sec per timed run; CV = standard deviation ÷ mean.
| target | endpoint | req/sec | p50 (ms) | p99 (ms) | PSS (MB) | req/sec/GB |
|---|---|---|---|---|---|---|
| spinel | /rooms/1 | 8,347 | 1.65 | 4.94 | 99 | 86,033 |
| spinel | /users/me/sidebar | 10,823 | 1.29 | 4.54 | 103 | 107,350 |
| spinel | /searches?q=coffee | 9,886 | 1.34 | 4.27 | 102 | 98,312 |
| spinel | /rooms/1/messages?before=50 | 9,884 | 1.36 | 4.25 | 101 | 99,902 |
| spinel | /users/{avatar}/avatar | 16,365 | 0.81 | 3.61 | 107 | 156,038 |
| spinel | {css} | 53,011 | 0.24 | 1.47 | 110 | 490,911 |
| spinel | /up | 37,877 | 0.33 | 1.72 | 110 | 351,958 |
| spinel | POST:/rooms/2/messages | 1,123 | 13.40 | 79.54 | 176 | 6,528 |
| tobi | /rooms/1 | 3,849 | 3.75 | 9.01 | 400 | 9,831 |
| tobi | /users/me/sidebar | 9,348 | 1.65 | 4.05 | 236 | 40,475 |
| tobi | /searches?q=coffee | 3,571 | 4.08 | 10.58 | 663 | 5,511 |
| tobi | /rooms/1/messages?before=50 | 8,027 | 1.60 | 5.70 | 384 | 21,394 |
| tobi | /users/{avatar}/avatar | 72,285 | 0.16 | 3.24 | 568 | 130,235 |
| tobi | {css} | 77,934 | 0.14 | 1.89 | 163 | 486,659 |
| tobi | /up | 55,957 | 0.21 | 2.28 | 155 | 368,228 |
| tobi | POST:/rooms/2/messages | 1,114 | 12.07 | 65.79 | 983 | 1,159 |
| ruby-falcon | /rooms/1 | 10,851 | 1.32 | 3.30 | 942 | 11,792 |
| ruby-falcon | /users/me/sidebar | 18,605 | 0.82 | 2.18 | 731 | 26,037 |
| ruby-falcon | /searches?q=coffee | 10,683 | 1.39 | 3.54 | 1,020 | 10,718 |
| ruby-falcon | /rooms/1/messages?before=50 | 12,091 | 1.06 | 3.49 | 1,015 | 12,195 |
| ruby-falcon | /users/{avatar}/avatar | 76,276 | 0.18 | 0.87 | 742 | 105,251 |
| ruby-falcon | {css} | 87,040 | 0.14 | 0.79 | 705 | 126,293 |
| ruby-falcon | /up | 37,393 | 0.42 | 1.26 | 701 | 54,587 |
| ruby-falcon | POST:/rooms/2/messages | 932 | 14.86 | 95.86 | 1,174 | 812 |
| ruby | /rooms/1 | 10,922 | 1.31 | 3.50 | 513 | 21,774 |
| ruby | /users/me/sidebar | 17,101 | 0.86 | 2.71 | 708 | 24,707 |
| ruby | /searches?q=coffee | 11,795 | 1.20 | 3.99 | 625 | 19,299 |
| ruby | /rooms/1/messages?before=50 | 11,964 | 1.18 | 3.64 | 559 | 21,898 |
| ruby | /users/{avatar}/avatar | 45,527 | 0.33 | 0.86 | 1,277 | 36,506 |
| ruby | {css} | 47,918 | 0.31 | 0.80 | 1,412 | 34,744 |
| ruby | /up | 30,782 | 0.47 | 2.30 | 702 | 44,895 |
| ruby | POST:/rooms/2/messages | 945 | 15.32 | 57.79 | 2,189 | 441 |
| rails | /rooms/1 | 1,852 | 8.05 | 20.63 | 1,427 | 1,328 |
| rails | /users/me/sidebar | 3,112 | 4.68 | 12.56 | 1,591 | 2,002 |
| rails | /searches?q=coffee | 1,983 | 7.14 | 46.28 | 1,560 | 1,301 |
| rails | /rooms/1/messages?before=50 | 1,913 | 7.70 | 40.54 | 1,500 | 1,305 |
| rails | /users/{avatar}/avatar | 21,782 | 0.64 | 2.53 | 1,561 | 14,283 |
| rails | {css} | 34,388 | 0.44 | 1.51 | 1,560 | 22,571 |
| rails | /up | 4,372 | 3.37 | 9.10 | 1,611 | 2,778 |
| rails | POST:/rooms/2/messages | 305 | 44.60 | 227.60 | 1,919 | 163 |
| hostname | showcase.party |
|---|---|
| OS | Ubuntu 24.04.5 LTS |
| kernel | Linux showcase.party 6.8.0-147-generic #147-Ubuntu SMP PREEMPT_DYNAMIC Thu Sep 24 11:33:16 UTC 2026 x86_64 x86_64 x86_64 GNU/Linux |
| board | ASRock B450 Pro4 R2.0 |
| CPU | AMD Ryzen 5 3600 6-Core Processor |
| topology | 6 cores / 12 threads |
| clock | governor=performance, boost disabled, 3600 MHz max |
| memory | 64,228 MB |
| app | campfire |
|---|---|
| revision | edbc779f4dfc |
| Rails | 8.2.0.alpha (e3d5c569d217) |
| Ruby | 4.0.5 |
| served | production |
| seed | 50 users, 5 rooms, 100 messages/room |
| app server (rails lanes) | puma 8.0.2 |
| app server (emit lanes) | puma 8.0.1 |
Recorded at boot, not copied from intent: a lane that asked for jemalloc and did not get one says so here.
| rails | jemalloc=/usr/lib/x86_64-linux-gnu/libjemalloc.so.2, resque-pool=pid 1177296, thruster=front on :19100 -> puma :20100, workers=campfire's own formula (WEB_CONCURRENCY unset) |
|---|---|
| ruby | jemalloc=/usr/lib/x86_64-linux-gnu/libjemalloc.so.2, workers=8 |
| ruby-falcon | server=falcon 0.57.0 via falcon.rb (forked), jemalloc=/usr/lib/x86_64-linux-gnu/libjemalloc.so.2, workers=8 |
| spinel | workers=autodetected |
| tobi | root=/home/rubys/git/campfire-once-ruby-ractor, revision=353e899, ruby=4.0.5, jemalloc=/usr/lib/x86_64-linux-gnu/libjemalloc.so.2, workers=8 ractors (one process) |
| cargo | cargo 1.98.1 (797e8a9bc 2026-08-05) |
|---|---|
| crystal | Crystal 1.20.2 [2482c62c1] (2026-05-15) |
| curl | curl 8.5.0 (x86_64-pc-linux-gnu) libcurl/8.5.0 OpenSSL/3.0.13 zlib/1.3 brotli/1.1.0 zstd/1.5.5 libidn2/2.3.7 libpsl/0.21.2 (+libidn2/2.3.7) libssh/0.10.6/openssl/zlib nghttp2/1.59.0 librtmp/2.3 OpenLDAP/2.6.10 |
| dotnet | 10.0.301 |
| go | go version go1.26.4 linux/amd64 |
| jruby | jruby 10.1.1.0 (4.0.0) 2026-07-22 aaa95d57e1 OpenJDK 64-Bit Server VM 25.0.3+9-LTS on 25.0.3+9-LTS +indy +jit [x86_64-linux] |
| mise | 2026.5.15 linux-x64 (2026-05-23) |
| node | v26.3.0 |
| python3 | Python 3.14.6 |
| redis-server | Redis server v=7.0.15 sha=00000000:0 malloc=jemalloc-5.3.0 bits=64 build=d064f9f48c901a92 |
| ruby | ruby 4.0.5 (2026-05-20 revision 64336ffd0e) +PRISM [x86_64-linux] |
| rustc | rustc 1.98.1 (48a229cea 2026-09-01) |
| shards | Shards 0.20.0 [b2b98ca] (2025-12-19) |
| spinel | spinel 2026.09.12+10496 (77198b78d) [gcc 13.3.0 (cc)] |
| sqlite3 | 3.45.1 2024-01-30 16:01:20 e876e51a0ed5c5b3126f52e532044363a014bc594cfefa87ffb5b82257ccalt1 (64-bit) |
| truffleruby | truffleruby 40.0.0 (2026-09-17), like ruby 4.0.2, Oracle GraalVM Native [x86_64-linux] |
| uv | uv 0.11.22 (x86_64-unknown-linux-musl) |
| wrk | wrk debian/4.1.0-4build2 [epoll] Copyright (C) 2012 Will Glozer |
| command | scripts/bench-campfire --app /home/rubys/once-campfire --port 19100 --out bench/results/campfire/20261017-111349 rails ruby ruby-falcon spinel tobi |
|---|---|
| workers | 1 |
| concurrency | 16 |
| runs | 3 × 20s after 20s warmup |
| wrk threads | 2 |
| endpoints | /rooms/1 /rooms/1/messages?before=50 /users/me/sidebar /searches?q=coffee /users/{avatar}/avatar {css} /up POST:/rooms/2/messages |
| targets | rails, ruby, ruby-falcon, spinel, tobi |
| commit | cceae7c37e05aba2978071eab74202a4bf913f42 |
|---|---|
| branch | main |
| subject | Merge pull request #1024 from dai199/ci-spinel-timings |
| load average | 0.52 / 3.86 / 6.23 |
|---|---|
| uptime | 12:44:53 up 1 day, 23:29, 1 user, load average: 0.52, 3.86, 6.23 |