What is a Supercomputer?
A supercomputer is a collection of individual computing nodes working together as a single, unified system. Rather than one extremely powerful machine, modern "supercomputers" are almost always clusters — many commodity machines connected by a fast network and managed by orchestration software.
The Raspberry Pi cluster in this guide is a miniature version of the same architectural concept used in data centres and HPC (High Performance Computing) facilities worldwide.
Key Concepts
| Concept | What it means |
|---|---|
| Node | A single physical computer in the cluster (one Raspberry Pi) |
| Cluster | The group of all nodes working together |
| Orchestration | Software that decides which node runs which workload |
| Container | An isolated, portable package containing an application and all its dependencies |
| Image | The read-only template a container is created from |
| Service | A long-running task defined by an image + configuration + replica count |
| Replica | One running instance of a service (one container) |
| Stack | A named group of services deployed together via a docker-compose.yml — the top-level unit of Swarm deployment |
Docker
Docker is a platform for building, running and shipping containers. On a single machine it's already very useful. Across a cluster with Docker Swarm, it becomes a distributed orchestrator — automatically placing replicas across nodes, load-balancing incoming traffic, and restarting failed containers.
Wikipedia: Docker (software)
Kubernetes
Kubernetes (K8s) is the industry-standard container orchestrator for large-scale production workloads. This guide focuses on Docker Swarm — which is simpler to set up — but the same docker-compose.yml files can be adapted to Kubernetes with tools like Kompose.
Docker Swarm is an excellent starting point. Once you understand the concepts (services, networks, replicas, health checks), moving to Kubernetes is a natural next step.
Wikipedia: Kubernetes
How Our Cluster Compares
| Our Pi Cluster | A Cloud VM cluster | HPC Supercomputer | |
|---|---|---|---|
| Nodes | 4 | 10–1000+ | Thousands |
| CPU cores | 16 | Hundreds–thousands | Millions |
| RAM | 8 GB | GBs–TBs | PBs |
| Cost | ~€80 | €/month | €millions |
| Orchestration | Docker Swarm | Kubernetes / ECS | Slurm / PBS |