Architecture

Deis uses a service oriented architecture with Components grouped into a Control Plane, Data Plane and Router Mesh.

System Diagram

Deis System Diagram

Operators use the Deisctl API to stand up the cluster’s Control Plane, Data Plane and Router Mesh. End-users of the platform interact with the Control Plane using the Deis API.

The Control Plane dispatches work to the Data Plane via a scheduler. The Router Mesh is used to route traffic to both the Control Plane and Data Plane. Because the router mesh is usually connected to the public Internet, it is often connected to a front-end load balancer.

Control Plane

Deis Control Plane Architecture

The Control Plane performs management functions for the platform. Control plane components (in blue) are all implemented as Docker containers.

The Store component consists of a number of smaller components that represent a containerized Ceph cluster which provides a blob storage API and POSIX filesystem API for the control plane’s stateful components:

  • Registry - a Docker registry used to hold images and configuration data
  • Database - a Postgres database used to store platform state
  • Logger - a syslog log server that holds aggregated logs from the data plane

End-users interact primarily with the Controller which exposes an HTTP API. They can also interact with the Builder via git push.

Data Plane

Deis Data Plane Architecture

The Data Plane is where Containers (in blue) are run on behalf of end-users.

The platform scheduler is in charge of placing containers on hosts in the data plane. Deis also requires a few lightweight components on these hosts:

Topologies

For small deployments you can run the entire platform – Control Plane, Data Plane and Router Mesh – on just 3 servers. For larger deployments, you’ll want to isolate the Control Plane and Router Mesh, then scale your data plane out to as many servers as you need.

The Deis Control Plane, Data Plane, and Router Mesh components all depend on an etcd cluster for service discovery and configuration. For larger deployments, you’ll want to isolate this as well.

Those building larger clusters should reference Isolating etcd for further details.