The future of Armada; three use cases

[severinson]

Future use cases for Armada

Here, we outline three use cases Armada could be evolved to address. They are distinct in the sense that an organisation may be interested in only one of them, but are overlapping in terms of the features necessary to support them.

1) Armada as a high-throughput resource creation engine

Kubernetes objects are submitted directly to Armada, which handles creating them in a suitable cluster, similar to how Armada currently operates. However, we significantly expand the range of resources Armada is able to manage.

2) Armada as a pod scheduler enforcing cross-cluster fairness

This use-case is outlined in #1925

3) Armada as a virtual cluster of apparently infinite size

Each user has a single "home" cluster, in which the user interacts with the kube-api just as with a regular cluster. Armada controls the resources available in each cluster by making available nodes residing in one cluster for scheduling in another cluster, e.g., via virtual kubelets.