In November 2014, we started work on the next phase of our network expansion and evolution.
Entitled Project Awesome, it would be the biggest change to our network and the way we provision customer connections since the company’s inception in 2008.
We had already seen through substantial improvements to our network, such as moving to a dual datacentre virtualised platform, and seeing the addition of FTTC, EFM and LLU access technologies.
We also have always had a focus from Day 1 on monitoring and analytics, building data collection, analysis and presentation functions into the fabric of the network, and deeply integrated into sales and support provision.
Constant live analysis of throughput, circuit sync rates, latency and SNR margins, as well as core router CPU usage and other network metrics have been crucial to building a scalable, self-healing platform.
But this was something different, and so fundamental a change, that we took the view that we should build our new core infrastructure from scratch.
By Spring 2015, every new connection was being provisioned on this network. A complete replacement built in parallel and adopting new principles of network design and management.
In the same way that servers have redundant arrays of disks (RAID), so has our entire network been designed.
Over engineered. Redundant. Built at scale.
A pioneering use of Network Function Virtualisation (NFV), with the ability to provision quickly with commercial off the shelf hardware, and with the raw power of the protocol that binds the internet together (BGP), this platform is the first in the industry in the UK.
By November 2015, only a year since the planning phase, we had effected a full migration of our entire customer estate onto this new platform.
And that’s just the core network, now resilient across 4 geographical locations (with no limit), and fully cloud based.
Along with this migration came a wholesale change in the way we provision broadband.
A move from a single broadband network to 3 distinct DSL platforms. Each with isolated LNS clusters and RADIUS, and dedicated internet transit.
Each DSL platform is managed independently, is uncontended, and none of them share any points of presence, either in their wholesale connectivity with BT and LLU partners, or to each in the location of their LNSs.
By the end of 2015 every customer connection had the benefits of multi-platform DSL and next generation cloud core routing infrastructure.
No other network is like this in the world, and no one else is provisioning connectivity as we do.
And because of this huge leap in design, we were able to ride out the major UK network events of the last 18 months.
Sovereign House, Harbour Exchange, Telehouse North; all experienced big outages, but our customers didn’t experience any downtime at all.
And now we come to our latest improvement in network circuit diagnostics; an evolution in our AI.
Systems that automatically diagnose hardware and line faults without the need for human engineers.
With direct control of every network node, constant data feeds of telemetry and northbound integration by API to each DSL platform, we have not only eliminated the need for human intervention for a majority of broadband faults, we have delivered tangible performance benefits.
Less circuit downtime.
More bandwidth available.
More time for the connection to do what the customer needs it to do.
This is the UK’s first Software Defined Access Network.
Allowing us to take Software Defined Networking to the customers’ premises.
Future proof, over engineered, and always on.