Feature Overview

Although performance is one of the key design principles of distributed systems, it is usually taken for granted and treated as an afterthought.

With HyperShift being a different architectural pattern, it is important to understand the behavior of the different control-plane and data-plane pieces, especially, how they perform at scale.

Background, and strategic fit

The customer always considers how a system behaves at scale before making a decision. Questions about performance that usually get asked are:

• How many clusters can I host per management cluster?
• How many nodes does it take to host one control-plane?
• With the control-plane separation, what are the acceptable network latency boundaries?
• ...

Goal(s)

• Identify Important KPIs for tracking HyperShift's performance.
• Integrate with CI periodically track performance (e.g, CI test for cluster create tracks resource consumption of control plane in mgmt cluster namespace)
• Provide means to visually track performance over time (e.g., Dashboards)