Most of our assumptions have outlived their uselessness.
The cloud-native data center network comes in many different sizes and shapes, ranging from the ones run by the hyperscalars such as Amazon, Google, and Microsoft to the networks of much smaller organizations, anywhere from 20 to 50 switches. But the common goals across all of them are reliability and cost-efficiency. Operational cost-efficiency is much harder to achieve than cost efficiency of purchasing a router. In my experience in dealing with a wide swath of organizations, cloud-native data center networks achieve reliable, cost-efficient networks by adhering to the following design principles:
Standard, simple building blocks
A reconsideration of network failures
Ruthless focus on simplicity
Each of these principles overturned fundamental network design principles on which enterprise and data center networks had thus far been built. I touch upon these aspects many times over in the course of this book. However, dedicating a short chapter to these principles before we move from the architecture and technology to actual deployment will, I predict, serve the reader well. Missing one or more of these principles trips up organizations that are trying to evolve and lands them back in the old morass.
To be sure, these three concepts draft off each other to paint a more complex picture than the way I present them as separate concepts. Also, not having been an ...