NRE-Bytes

Morsels of NRE Wisdom

Building an Edge Traffic Controller - Part 2

Practical demo of using the controller to detour traffic from overloaded interfaces

In Part 1 of this series, we ran through the technical details of what it would take to build an edge traffic controller to steer traffic away from overloaded edge links. In this blog, I will try to demonstrate the controller in action by simulating traffic flows using our virtualized topology. We will also look into other real world considerations such as operational monitoring and metrics. Initial Setup To recap, this is what our topology looks like:

Building an Edge Traffic Controller - Part 1

A Proof of Concept implementation of a Software Defined Edge traffic controller using Sflow and GoBGP

2017 was the year of the Software Defined Network (SDN). Apart from other things like new players jumping onto the SDN space and a bunch of new SD-WAN offerings, two prominent innovation leaders - Google and Facebook - both released blogs and papers on their software defined edge network. Google’s solution; named Espresso, is likely a more battle tested and production-hardened solution owning to years of R&D and testing. It is, however (or at least in my opinion) a lot more complex than Facebook’s Edge Fabric that uses a much simpler approach to solving the same problem - which is to overcome BGP’s inability to take link performance ( which translates to application performance) into account for its routing decisions.