This page lists some thoughts on network design and architecture issues, in general terms, called heuristics, because they are rules of thumb that may be challenged by any number of engineering & business tradeoffs made in a specific implementation. These heuristics are in part based on the principles / goals outlined in On Information & Information Networks. Those principles / goals are abstract in nature, this page brings them one step forward to being more applicable to actual networks.
Protocol information should be sufficient, clear, and accurate
- Aggregate information to the forwarding plane, from the packet header and the control plane, should enable the best fowarding decision: complete, necessary, available when needed, without noise, without redundancy, without error, and truthful.
- Control plane information should be necessary to the intended outcome, no more, no less.
Networks are subject to the laws of physics
- Transmission signals propagate at a speed dictated by physics.
- The amount of bits/symbols that propagate per second is a function of the serialization rate.
- When distance or velocity changes, time changes.
- Information is distributed, leading to a different understanding of when an event occurred. We often design around such “relativity” by adding information, for example a timestamp.
- No router or controller is omnipotent.
- Complexity cannot be reduced further than necessary information.
- More loosely, complexity cannot be reduced for the SAME capability.
The value of networks
- As networks grow larger, communication between all nodes becomes more difficult.
- As networks grow larger, information hubs contribute to scale and expansion.
Automation & Autonomy
- Automation & autonomy increases productivity and scale