NetworkPolicy scale tests

[jianjuns]

Describe what you are trying to solve
Do scale tests and performance evaluation of Antrea NetworkPolicy implementation. This is mainly for NetworkPolicy computation and dissemination to Agents.

Describe the solution you have in mind
We need to cover:

1. Controller NetworkPolicy computation performance and CPU/memory consumption.
2. Controller to Agent dissemination.
   Odeally we should support the K8s scale of 5K Nodes.
   As a first step, we could start from 1K Nodes and 30K Pods, and at least verify Antrea does not perform very bad.

Describe the main design/architecture of your solution
As we might not be able to build a testbed with >1K Nodes and > 30K Pods, we might need some simulation here. Simulation of Nodes to stress Controller dissemination might be simple - we just need some concurrent clients to watch the NetworkPolicy resources from the Controller. Simulation of Pods and NetworkPolicies might require more work.

[jianjuns]

@abhiraut @tnqn @weiqiangt

[jianjuns]

@tnqn and @weiqiangt: I am assigning this issue to you as you worked on the scale tests and already made some progress. Maybe you want to update the issue and add progress?

[github-actions]

This issue is stale because it has been open 180 days with no activity. Remove stale label or comment, or this will be closed in 180 days