Skip to content

Neton beta22: default NetonStream engine and Hyper4k comparison - #1517

Merged
MDA2AV merged 3 commits into
MDA2AV:mainfrom
zoujiaqing:netonstream-dual-engine
Oct 3, 2026
Merged

MDA2AV merged 3 commits into
MDA2AV:mainfrom
zoujiaqing:netonstream-dual-engine

Conversation

@zoujiaqing

Copy link
Copy Markdown
Contributor

Changes

  • Switch neton to the Kotlin/Native neton-http-netonstream adapter.
  • Add neton-hyper using neton-http-hyper4k.
  • Both entries use Neton 1.0.0-beta21, Kotlin 2.4.0, identical Main.kt routes, GC settings, configuration and profile subscriptions. Only the engine factory and adapter dependency differ.
  • The new stack uses published com.netonstream:io/http/tls/openssl:0.1.0 (upstream OpenSSL remains 4.0.2).
  • No changes to benchmark scripts, scoring or recorded results.

Validation

  • New adapter: 54 tests on macOS arm64 and Linux arm64, both epoll and io_uring; io_uring repeated successfully.
  • Hyper4k adapter: 35 macOS tests.
  • Both entries passed actual-binary mixed baseline smoke tests on macOS and Linux arm64: GET, Content-Length POST, chunked POST, 128 concurrent clients, no incorrect responses or 503.
  • Linux x64, Linux arm64 and macOS release linking checked with staged signed release dependencies.

Release Gate

Foundation 0.1.0 artifacts are publicly available. Neton beta21 publication and final Central-only consumer builds are being completed before this PR is marked ready.

Benchmark Scope

First compare baseline for both entries. No performance improvement is claimed before Arena measurements. Other existing profile subscriptions retained; new-engine high-load behavior is not yet validated.

@zoujiaqing

Copy link
Copy Markdown
Contributor Author

/benchmark-multiple -f neton,neton-hyper -t baseline

@github-actions

Copy link
Copy Markdown
Contributor

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

@zoujiaqing
zoujiaqing marked this pull request as ready for review September 30, 2026 04:18
@github-actions

Copy link
Copy Markdown
Contributor

Benchmark Results

Frameworks: 2 | Test: baseline

✅ neton

Test Conn RPS CPU Mem Δ RPS Δ Mem
baseline 4096 372,848 3386.0% 1.8GiB -43.2% +50.0%

✅ neton-hyper

Test Conn RPS CPU Mem Δ RPS Δ Mem
baseline 4096 1,034,982 5282.7% 1.1GiB ~0% ~0%

@zoujiaqing

Copy link
Copy Markdown
Contributor Author

/benchmark-multiple -f neton,neton-hyper -t baseline

@github-actions

Copy link
Copy Markdown
Contributor

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

@zoujiaqing zoujiaqing changed the title Neton beta21: default NetonStream engine and Hyper4k comparison Neton beta22: default NetonStream engine and Hyper4k comparison Sep 30, 2026
@github-actions

Copy link
Copy Markdown
Contributor

Benchmark Results

Frameworks: 2 | Test: baseline

✅ neton

Test Conn RPS CPU Mem Δ RPS Δ Mem
baseline 4096 276,446 2508.4% 1.7GiB -57.9% +41.7%

✅ neton-hyper

Test Conn RPS CPU Mem Δ RPS Δ Mem
baseline 4096 1,034,545 5271.7% 1.1GiB ~0% ~0%

@MDA2AV
MDA2AV merged commit 5b111b8 into MDA2AV:main Oct 3, 2026
5 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants