Skip to content

Getting started Updated 2026-05-30

Concepts and terminology

A concise reference defining the core vocabulary of Stack Harbor PaaS — every term you will encounter across the documentation.

This page defines the vocabulary used throughout the Stack Harbor PaaS documentation. Each term is defined once, at the precision level you need to read any other page confidently. Where a deeper page exists, the term links to it.

What this covers

  • The structural units of the platform: environment, layer, node, container.
  • The resource unit and how it maps to actual capacity: cloudlet, reserved cloudlet, dynamic cloudlet.
  • Infrastructure geography: region, cluster.
  • Topology and the wizard used to build it.
  • Networking primitives: shared load balancer, endpoint, public IP.
  • Environment-level organisation: alias, environment group.

Core terms

Environments and structure

Environment — the top-level unit of work on the platform. An environment is a named, logical group of containers that run one application in one region. It holds all the tiers of that application — web server, application server, database, cache — and has its own network namespace, domain entry-point, and settings. See Environment management.

Layer — a functional role within an environment. The platform groups containers by role: a load-balancer layer, an application-server layer, a database layer, and so on. Each layer can hold multiple nodes. Scaling a layer means changing its node count (horizontal scaling) or changing the size of each node (vertical scaling). See Scaling and clustering.

Node — a single running container inside a layer. When a layer has three replicas, it has three nodes. The node is the unit you restart, redeploy, SSH into, and observe in logs and metrics.

Container — the isolated runtime for a workload. On Stack Harbor PaaS, containers come in two varieties (see below). Every node is a container; the two words are often used interchangeably when the context makes the type clear. See Container types and use cases.

System container — the OS-level isolation boundary that the platform places on each physical host. A system container has its own filesystem, network interfaces, and process tree, but shares the host kernel — there is no guest OS and no hypervisor overhead. Your application containers run inside a system container. This nesting is what makes live vertical resize possible: the orchestrator adjusts resource limits on the system container without restarting it. See Architecture overview.

Application container — a Docker container or Kubernetes pod running inside a system container. This is where your code lives. The platform manages the system container around it; you manage (or choose from the platform’s catalogue) what runs inside. See Container types and use cases.

Resources

Cloudlet — the platform’s unit of RAM and CPU. Every container is sized in cloudlets. A cloudlet is 128 MiB of RAM and 200 MHz of CPU.

Reserved cloudlet — the guaranteed minimum allocation for a container. The container is always given at least this many cloudlets, regardless of actual load. Reserved cloudlets are billed continuously, at a lower unit rate than dynamic cloudlets. Setting the reserved floor prevents cold-start latency when a container wakes from low utilisation.

Dynamic cloudlet — the burst ceiling above the reserved floor. When load rises, the platform allocates additional cloudlets up to this ceiling automatically. When load falls, those cloudlets are released. Dynamic cloudlets are billed only for actual consumption, not for the ceiling.

Infrastructure geography

Region — a geographic deployment location offered by the platform. Each region is backed by dedicated hardware in a specific data centre. You choose a region when you create an environment; the environment lives in that region for its lifetime unless explicitly migrated.

Cluster — the physical and logical unit that backs one region. A cluster has its own hardware capacity, its own internal network, and its own orchestrator instance. “Region” and “cluster” refer to the same boundary from different perspectives: region is the name you see in the dashboard; cluster is the infrastructure concept behind it. See Architecture overview.

Topology

Topology — the shape of an environment: how many layers it has, which software stack each layer runs, and how many nodes and cloudlets each layer is allocated. The topology of an environment is visible and editable at any time through the dashboard.

Topology Wizard — the dashboard interface for defining or changing an environment’s topology. You use the Topology Wizard when creating a new environment (to pick stacks, set cloudlet ranges, and name the environment) and when modifying an existing one (to add layers, change node counts, or adjust cloudlet limits). See Architecture overview and Auto-clustering via topology wizard.

Networking

Shared load balancer (SLB) — the platform-managed ingress front-end that terminates TLS and routes inbound traffic to the correct environment. Environments that do not have a dedicated public IP are reached through the SLB. The SLB is shared across all environments in a cluster; it is not a layer you provision inside your environment. See SSL and TLS overview.

Endpoint — a mapped port on the shared load balancer that provides direct access to a specific container port without requiring a public IP. Endpoints are used to expose non-HTTP services — databases, custom TCP or UDP ports — to the internet. Each endpoint gets a hostname and a randomly assigned external port on the SLB. See Endpoints: direct access.

Public IP — a dedicated IP address attached to a specific container layer. Traffic arrives directly at that container’s network interface, bypassing the shared load balancer entirely. A public IP is required for custom SSL termination at the container level and for any protocol that demands a fixed, predictable IP. See Public IP.

Environment organisation

Environment alias — a human-readable label attached to an environment, shown alongside the technical environment name in the dashboard. Aliases do not affect routing or infrastructure; they make long lists of environments easier to scan. See Environment aliases.

Environment group — a named collection of environments displayed together in the dashboard. Groups are organisational only — they do not share networking or resources. A single environment can belong to more than one group. See Environment groups — overview.

Where to go next