BRKRST-3014 – Policy, Complexity, and Modern Control Planes

Presented by: Russ White, LinkedIn

Networks are complex. How do we measure complexity? How do we measure scale? What’s the unit of measure?

You can’t “solve” complexity.

Alderson D and J Dole “complexity in highly organized systems arises primarily from design strategies intended to create robustness to uncertainty.” There’s a point on the complexity scale where robustness actually drops. “Robust but fragile”.

Dunning Kruger effect?

What is complexity?

  • Anything you don’t understand?
  • Anything with many parts?
  • Anything with unintended consequences?
  • Something that can’t be solved; can’t be easily defined.
  • We need to develop a model to understand complex systems quickly

“If you haven’t found the trade off, you haven’t looked hard enough” — Russ

The model:

  • Ask: Why?
  • Ask: What & How?
    • State
    • Optimization
    • [Interaction] Surface (where two components interact with each other [which could be, and often is, human on CLI])
    • No matter how you’re analyzing a network (protocols, applications, whatever), you’ll find these 3 things.
  • Ask: This is like what? (what’s it similar to?)
  • Good examples of questions to ask for each point in different use cases are in the deck

3-way trade off:

  • Quick/Cheap/Quality, or
  • State/Surface/Optimization

“Adding more state to the system should result in an increase in optimization. If not, why are you adding state?” — Russ

“Leaky abstraction”

Bad questions:

  • “How complex is this?”
  • “Will this scale?”
  • How the hell do you answer these questions?!

This was a fascinating session and will require watching the recording and thumbing through the slides at a slower pace. 

Leave a Reply

Your email address will not be published. Required fields are marked *

Would you like to subscribe to email notification of new comments? You can also subscribe without commenting.