Skip to content

cefothe/devoxx-uk-2026

Repository files navigation

Devoxx UK 2026 — Achieving Microsecond Latencies with Java

Companion repository for Stefan Angelov's 50-minute Devoxx UK 2026 talk.

What's in here

devoxx-uk/
├── submission/        Conference submission package (title, abstract, bio, outline, narrative)
├── benchmarks/        Maven JMH benchmark suite — every claim in the talk is backed by code
├── results/           Expected results, raw JSON output, and analysis-and-narrative.md
└── slides/            Marp-formatted slide deck + speaker notes

Reading order

If you're Stefan preparing for the talk:

  1. submission/02-abstract.md — the audience-facing pitch
  2. submission/06-talk-outline.md — minute-by-minute structure
  3. submission/07-narrative-arc.md — the story you tell
  4. benchmarks/BENCHMARK-METHODOLOGY.md — how the numbers are produced
  5. benchmarks/scripts/run-all.sh — run the suite, capture JSON
  6. results/analysis-and-narrative.md — what each number means
  7. slides/slide-deck.md — the Marp deck

If you're a teammate (review, fact-check, port, etc.):

  1. benchmarks/README.md — how to build & run
  2. benchmarks/BENCHMARK-METHODOLOGY.md — JMH config + caveats
  3. Each benchmark file has a header comment explaining the claim it backs

Source corpus this is built from

This is not a from-scratch effort. The talk distils:

  • ../final-article/FINAL-ARTICLE-FOR-SUBMISSION.md — JavaPro-published article (~28.5k words)
  • ../article-outline.md — full structural outline
  • ../three-techniques-complete.md — the climax: off-heap, single-threaded, pooling
  • ../Low latency Java systems.pptx (1).pdf — the original deck

Reference production system: exchange-core (https://github.com/exchange-core/exchange-core). 5M ops/sec, 0.5 µs median, p99 ~42 µs.

Hardware target

Benchmarks are designed to run on:

  • AMD EPYC 7282 @ 2.8 GHz, 3 physical cores, 24 GB RAM, 180 GB NVMe (publishable numbers)
  • Mac M2/M3 (dev iteration only — affinity benchmarks skip themselves on macOS)

JDK 25 (OpenJDK).

Slide rendering

Slides are written for Marp. To preview/export:

npm i -g @marp-team/marp-cli
marp slides/slide-deck.md --preview        # live preview
marp slides/slide-deck.md --pdf            # export PDF
marp slides/slide-deck.md --pptx           # export PowerPoint
marp slides/slide-deck.md --html           # export self-contained HTML

About

No description, website, or topics provided.

Resources

Stars

1 star

Watchers

1 watching

Forks

Releases

No releases published

Packages

 
 
 

Contributors