The talk
Clusterf**ctional — Haarlem, October 21, 2026, 40 minutes including Q&A. Devs who have only done cross-service communication over HTTP; people curious how distribution works deeper in the stack; anyone running services in Kubernetes.
Act by act
Section titled “Act by act”| Part | Time | Research |
|---|---|---|
| Act 1 — EPMD & Erlang distribution | 3 min | Erlang distribution |
| Act 1 — libcluster | 2 min | libcluster |
| Act 1 — Live demo, two nodes | 3 min | Experiments |
| Act 2 — Service registration, client pattern, retries, two-tier messaging | 15 min | NETZlive in production |
| Act 3 — Dependency tax, boilerplate, when not to cluster | 10 min | Trade-offs |
| Closing — four principles | 2 min | — |
The Alm metaphor
Section titled “The Alm metaphor”The abstract frames distributed Elixir through herding goats on a Swiss Alm. Each mapping should hold up technically, or be flagged on stage as a stretch. Track them here.
| Alm | BEAM | Holds up? |
|---|---|---|
| Looking for the next unmilked goat | Node.list() |
|
| Herd splits on the way up the mountain | Netsplit | |
| Goats spread out so no spot gets overgrazed | Process distribution | Abstract already calls this a stretch |