Dependability is the watchword for any telecommunications operator. Sound monitoring goes a long way to ensure continuity of service, but typically this relies on known existing metrics to be effective. When the network changes and these metrics are invalidated, it can be challenging to confirm there is no ongoing service impact for the end user.

This was the challenge that REANNZ faced in the past year as we undertook a hardware refresh affecting our entire domestic backhaul. Every service we deliver was required to be rehomed and reimplemented according to a new network architecture. To catalogue every possible service and establish individually the evidence that would be required to prove their continuity after migration would require an enormous time-investment and be prone to error, so we instead developed a generalised service model and implemented automation with Python to reason from live data pulled from our network. New configuration was generated consistent with the requirements of our new hardware, and feedback produced at the time of implementation describing in clear detail a comparison between old and new running state. This model was iterated upon based on our learnings over time until it could accurately verify even highly bespoke configurations.

In this presentation, I propose to demonstrate the workings of this system with case studies from our experience. I will explain the challenges we encountered and how they were overcome, and conclude how ultimately this work proved its worth for our membership.