The way to overcome the structural limitations of today's internet is to change its architecture, believes the team behind the SPARC project. The alternative would be more reliable, fully automated and cheaper to manage.

The internet currently connects about two billion people worldwide, delivering a plethora of services like banking, e-commerce, social networking, media and content storage. Only 20% of the global population has access, but this figure is steadily growing. Despite its appeal, the internet can at times have service, mobility and security issues. New technologies to break the current 'vendor lock-in' are needed to cope with this increasing demand and to help deliver a top-notch service.

The SPARC project studied the feasibility of introducing a so-called 'split' to the current architecture of internet components. This split would support network design and operation in large-scale networks requiring a high degree of automation and reliability. These networks would also be cheaper and simpler to manage than those currently operating.

The team was the first to take the relatively undeveloped concept of split architecture and try elevating it to carrier grade. The work involved splitting the traditionally monolithic and inflexible IP router architecture into separate forwarding (hardware) and control (software) elements. In short, the network control plane was physically separated from the forwarding plane and then programmed to control several devices simultaneously. Beyond this forwarding function, they proposed a data-processing plane that can be software controlled, following the same approach.

This new architecture is manageable, cost-effective and adaptable, making it ideal for the high-bandwidth, dynamic nature of current and future applications, believes the EU-funded team.

Testing the theory

SPARC also developed a prototype based on the OpenFlow concept – an approach originally developed by Stanford University in its 'clean slate programme'. With future internet research demanding virtualised ICT resources, the OpenFlow method provides virtualisation by nature and is a good basis for testing. It is also needed to move network control out of proprietary network switches and into control software that is open source and locally managed.

The team's investigations showed how introducing split architecture into real networks would have a profound impact on the way they are built and operated. Overall, this would mean significantly less dependency on legacy protocols and design principles.

“With barriers to entry mostly diminished, business opportunities would automatically open up for new players in the telecommunications network element market,” says project coordinator Andreas Gladisch of Deutsche Telekom. “The new architecture would also optimise network and functional building blocks like network application modules and controller software.”

The findings also point to increased scalability – enhancing the internet's ability to process large amounts of information.

The new architecture would finally end the rigid vendor lock-in, which makes a customer dependent upon one non-open source vendor and hampers the internet's reliability and development. In addition, flexible and open access to hardware under the OpenFlow concept would accelerate the development and adoption of new services.

“The system would be highly reliable and robust – a big improvement on current standards. Its high automation means it could cope with the expected increase in internet connections across the globe,” says Gladisch. “The system also has a high level of interoperability.”

Crucially for EU competitiveness, the SPARC team says its approach would also encourage a dynamic ecosystem of companies developing new control applications.

SPARC finished in 2012, and as its results filter through the developer community, they may well change the internet landscape as we know it.

“The system would be highly reliable and robust – a big improvement on current standards …”

More info