February 15, 2018 by Michel Langlois

Who Are You Building Your Network For?


Check the headlines of any industry news publication and you can’t help but be bombarded with messages of networking salvation:

  1. Virtualization + open source software+ commodity hardware will dramatically reduce costs, solve network-wide compute and scale problems, and empower service providers to turn themselves into systems integrators with the freedom of Do It Yourself
  2. DevOps methodologies will deliver new services overnight while offsetting the inertia of decades of complex, disjoint, and labor-intensive workflows via automation
  3. Domain-based SDN service-chaining will not only peel off layers of legacy OSS/BSS integration, but is now coupled with “intent-based networking” that will provide adaptive networking and welcome in an era of autonomous networking

While there are nuggets of truth in these headlines, they are not in themselves addressing the specific business challenges that broadband service providers face:

  • Monetization of an access network while locked into a war selling bandwidth speed increments with sky rocking costs of ownership, compounded by the fact that there are three dedicated access networks – one each for business, residential, and wireless
  • New competitive threats from OTT application providers taking over the consumer’s home and therefore owning the subscriber experience and a large share of the associated consumer wallet
  • Because of a lack of real ownership in solving smart home and WIFI complexity, consumers are turning their back on service providers, making them the default bad guys because they provide the connectivity to the Internet

The access network architecture needs a strategic refresh, but the outcome will depend on your point of view.

Building out from the core. Over decades, we have been conditioned to build networks from the core out. Multiple factors can explain that thinking. First the Internet, web data center, and the distributed enterprise networks meant that core transport bandwidth, reachability, and capacity were the problems to solve. Traffic patterns and data consumption was largely a web download client-server model build around the concept of aggregation of flows, sessions, and uplinks/downlink distribution.

Then edge functions started to emerge to regulate traffic, policies, access authentication, SLAs, and lawful intercept, among other things, and spawned an investment feast where the networking space got populated with CORE routers, Edge and PE router, Route Reflectors, multi-service edge (MSE) routers, service aggregation switches… you name it.

At the same time, the large incumbent vendors did an amazing job in raising complexity of choice, giving rise to the term “End-to-End solution” concocted to reward vendor specific extensions and one shop architectures while also creating an industry of certified engineers and network architects well rewarded professionally to be the only ones capable of figuring out the network design.

By the time we got to the central office, 90% of the capital budget was spent and the access network was reduced to a speed, reach, and distribution problem with no inherent intelligence. “Smart core, dumb pipes” makes great sense if you sell routers. Incumbency was fast to conclude that the fate of access evolution was plain connectivity funded by broadband stimulus. And if you follow that thinking, why not move all control plane intelligence from all network elements to a centralized model?

Data center thinking. Enter the data center architects, with visions of web scale compute and an adaptive pool of storage, ready to virtualize applications across commodity server farms and inspired to build new access architectures such as we see with ONF’s OpenCORD implementation. New network-wide orchestration, control, and data planes will add virtual services and edge compute throughout the network. But access networks are fundamentally constrained distribution networks with peculiar, mandatory PHY layers – and the data center grand vision gets challenged when you try to implement it beyond greenfield networks.

Put the subscriber first. Imagine a clean sheet of paper on which you design a subscriber-centric network starting with subscriber experience and working back to the core. Change your perspective to a vision of elastic networks build upon software platforms decoupled from the physical access network. Like thinking about the wireless network not as a collection of smartphones and tablets, but a distributed collection of iOS and Android software platforms launching personalized subscriber experience applications. Access networks can’t avoid the physical layer, but the innovation engine of the future will be based on software platforms, not silicon.

Before you build a network, think like the subscriber each one of is. What do we all want?

  • Wireless connectivity everywhere and available all the time
  • Orchestrated smart home applications and services
  • Flexible application and service choices with immediate delivery
  • Simple… with knowledgeable full-service support when we need it

Work your way backward from those core subscriber-first requirements as you design your network. What would it look like? How would it operate? Check out our AXOS, EXOS, and Calix Cloud software platforms for inspiration.

Need some validation? See what Verizon, Pilot Fiber, and others are accomplishing with a Calix Software Defined Access network.