In the Beginning Was the Server - AtlasOfLivingAustralia/documentation GitHub Wiki

Servers were like tractors: massive, clunky, and entirely yours.
Physical machines, metal beasts with personal names: pegaso, loki, olea.
If something broke, you knew exactly why.
Bare metal. One machine, one person, one guilt.
It was simple — and that’s why it worked.
Everyone tended their server the way a farmer tends their field.
Knowledge lived in people’s heads and hands.

Then came virtual machines (VM), and suddenly you could drive tractors with trailers.
One engine could pull dozens of instances.
It was more efficient, sure, but also more abstract.
It wasn’t enough to know hardware anymore; now you had to understand other things hypervisors, shared storage, virtual networks.
Still, the setup wasn’t truly efficient — each VM carried more software than it needed, like an engine idling too loud.

Then Docker burst onto the scene, promising efficiency through simplicity.
It was the era of camper vans: light, portable, self-contained.
Developers could now pack up their environment and take it anywhere.
Knowledge became horizontal: anyone could “deploy.”
Companies thought they had discovered perpetual autonomy.
In reality, they had merely moved complexity sideways — from the server to the container, from the sysadmin to the developer.

Docker Compose was the family phase: the motorhome.
You no longer traveled alone — the database, cache, backend, and frontend all rode along, defined neatly inside one YAML file.
It was the dream of self-management: small, agile teams able to spin up entire environments in minutes.
But long trips required coordination — or rather, orchestration: who empties the tank, who cooks, who drives?
And soon it became clear: that wouldn’t scale well.

Then came Docker Swarm, promising order through automation.
But it wasn’t an “automatic highway” — it was more like a convoy of camper vans linked by walkie-talkies, trying to stay in formation down the freeway.
When it worked, it was beautiful.
But one lost signal and the whole group scattered across the map.
Swarm was our first reminder that coordination is expensive, and that simplicity doesn’t scale well.

And then came Kubernetes, bringing with it the international airport.
You don’t drive anymore.
Your applications are airplanes: taking off, landing, changing runways — without anyone touching the controls.
The technical staff are no longer mechanics or drivers but air traffic controllers.
They need to know less about wires and more about abstractions: pods, services, namespaces, ingress controllers.
You need a control tower, security teams, observability, audits.
Every flight is safe, traceable, and perfectly bureaucratic.
Knowledge has become a shift system.

Until ECS and its sibling EKS appeared — Amazon’s (and other companies) managed skies.
You don’t even have an airport anymore: you rent one.
The staff are still there, but wearing someone else’s vests.
Your teams write manifests and pay invoices; the runway, tower, and radar belong to another company.
It’s comfortable. Also irreversible.
One day you try to land back on solid ground and realize you no longer have pilots — only ticket sellers and route planners.

Summary

Bare metal: One person, one machine, 100% local knowledge.
VMs: Small teams, growing specialization.
Docker / Compose: Individual freedom, collective chaos.
Swarm: Early dreams of automated order.
Kubernetes: Massive coordination, new roles, less direct control, more complexity.
ECS/EKS: Total outsourcing, comfort traded for dependency.

Each technological leap reduces friction — and understanding.
Every layer of abstraction removes a pain, but also a skill.
The real cost isn’t hardware anymore: it’s the type of human team you need to keep the lights on.

At first, we were mechanics, then air traffic controllers, later ticket sellers.

Note: As of October 2025, our community still runs portals at every level — from bare servers without virtualization to full Kubernetes clusters on Amazon EKS.

Inspired by
In the Beginning was the Command Line by Neal Stephenson (1999)