By: Adam Dunstan

Broadband service begins with edge access. It stands to reason that the process of virtualizing broadband capabilities should also start at the edge. Of course, there are things that just cannot be virtualized, such as physical connectivity, which remains a fact of life in fixed broadband. But, the edge is arguably the largest component of any service provider network, so it is the logical place to start.

Leaving aside the in-premise connectivity equipment, as its upgrade is a costly logistical problem, the next component after physical connectivity is packet processing at the broadband edge, where the process of delivering services begins in the network. Service is provisioned here with Broadband Network Gateways (BNGs) and an array of OSS software components.  Beginning by virtualizing the broadband edge unlocks the opportunity to move customers’ connectivity directly into the cloud and provides the opportunity to use the cloud to virtualize other network and infrastructure services.

Two profoundly different operating models

Operators throughout the world have virtualization projects underway, and the marketing around the use of virtualization is largely the same as it has always been – lower costs and new services.  While these are legitimate benefits, there is a much more profound reason for the urgency to move to virtualization. This reason is something called webscale.

Webscale is a term used to describe the automation used to build large-scale cloud data centers that deliver infrastructure services such as those provided by Amazon, or applications such as those provided by Google and Facebook. This automation has resulted in a fundamentally different business-operating model compared to the one used by broadband service providers. Thanks to automation, each operator at Google and Facebook manages 20-30K server instances – resulting in orders of magnitude operating cost savings compared to broadband service providers’ operations, where engineers still need to perform manual processes to manage servers and network equipment. 

However, a more profound difference is the fact that most of their engineers aren’t tasked with managing an equipment vendor’s boutique equipment. Because only a relatively small team of engineers is needed to operate the complex, most of the engineers come to work each day and dedicate their time to developing products for customers.  The exact opposite is true at broadband service providers, where the majority of engineers spend the majority of their time babysitting complex, proprietary systems. So, the competitive advantage enjoyed by companies like Google and Facebook is both proven and significant, leaving operators with little choice but to move to webscale or face an existential threat they cannot overcome.

The components used to develop webscale infrastructure also represent a significantly different model than the one used in the development of existing network equipment – a difference that draws parallels to the structural changes that occurred during the migration from mainframes and mini-computers to personal computers 30 years ago. Webscale technologies leverage a “software-server-ecosystem” development methodology. Instead of a “waterfall” hardware development cycle that starts with ASIC development and has a three- to four-year duration, this new model leverages developments being undertaken in the ecosystem. Unlocked by the availability of cost-effective high-performance multicore processor technology, the primary components of this ecosystem are Intel architecture servers and the Linux operating system. By developing solutions using these components, network vendors and their customers can leverage far greater economies of scale.  

From a hardware perspective, the boutique, purpose-built computers we call network devices are replaced with general-purpose servers. In addition to yielding cost benefits, the volume of production of these servers results in processor and platform innovation that are up to five times faster than boutique network equipment. The Linux software ecosystem is equally important, because it provides the capability to leverage software components common to all computers (including network equipment), reducing development costs and, more importantly, allowing vendors to simply work on developing functionality where they have specialist skills, and avoid most of the overhead associated with re-creating infrastructure components. Taken one step further, this ecosystem allows vendors and operators to re-imagine and redesign the components and the network used to provide broadband services.