In the Great SDN Expectations blog, the first in a series discussing the technologies of the future service network, we introduced the basic constructs of what Software Defined Networks mean to the platforms at the heart of the Communication Service Provider (CSP) network today. SDN itself represents a deconstructed view of current switching and routing architectures providing the abstractions of control from dataplane. With this newfound control, the industry is leveraging multiple virtualization strategies to build out public, private, and hybrid cloud architectures, delivering multiple new application services to the market.
In fairness, virtualization is not a new subject. In fact, you could even say it has been with us for several decades when originally referred to as ‘Time-Sharing’ in mini and mainframe computing. Time-Sharing was driven to extract the greatest amount of value from the platform investment, which at the time was costly in a batch processing world. Today, virtualization is occurring at multiple levels and reaches across applications and network elements guided by both the European Telecommunications Standards Institute (ETSI) Network Function Virtualization programme and Internet Engineering Task Force (IETF) NFVRG (Network Function Research Group).
Perhaps this could be described as the perfect storm of data center cloud transformation: micro-services application design, middleware systems, containerization and network operating systems building connections across all these domains under direction from service lifecycle orchestration platforms.
All of these subjects are quickly influencing how the industry leverages virtualization to blur the lines from the back-office through the network edge to once again extract the greatest amount of value from the platform investment in what is now a cloud-based world, regardless of where you operate your cloud.
Just past the network edge, the customer premises equipment itself has been the next target for virtualization. Two trends in approach have been presented in the industry to address virtual CPE (vCPE) services.
The gateway compute or gateway hosted application model is where a network service function can be deployed and run in a CSP controlled manner within the subscriber gateway device. Using the gateway in this manner to run network functions as isolated applications has been deployed in the commercial customer space, collapsing costly access routers, load balancers, voice and security boxes into software running on a single physical device. This model has proven to be cost-effective and operationally beneficial.
The other approach is to move all the software functions into the CSP cloud, where through network isolation each subscriber continues to have a rich set of features delivered via a simplified, often lower cost demarc device at the premises. This approach has also shown itself to be valuable and cost-effective in commercial vCPE models.
In the residential gateway, the same two functional models are being explored. Initiatives to research network functions in more ‘CPU rich’ gateways creates the opportunity to introduce ‘cloud-to-ground’ services to subscribers which are run from within their premises while controlled by the CSP.
Alternatively, collapsing much of the gateway functions and terminating subscriber services within the CSP network supports lower-cost gateways, and re-uses many existing gateways to leverage the compute power of the CSP data center.
The virtualization building blocks are here. An application function can be delivered to any subscriber and run from the most cost-efficient location for the CSP business. In upcoming blogs we will discuss the role of controllers, network functions, and how orchestration brings organization and service across all these platforms.