Realizing the SDN dream

Phil Marshall/Tolaga Research
16 Oct 2013

In the 1970s the computer industry was concerned with what was referred as the "software crisis". During this crisis, the power of computer hardware was increasing at Moore's Law rates, creating tremendous complexities for software programming. It wasn't until standardized operating systems and "high level" programming languages emerged, that software could be abstracted from the complexities of the underlying hardware upon which it was deployed, to enable advanced computer systems to thrive.

In many respects the computer industry of the early 1970s is a suitable metaphor for expressing the challenges associated with communication networks today. To deliver the necessary performance, most network infrastructure has traditionally required purpose-built hardware, with each network element being optimized for the set of functions that it delivers. This created siloed network architectures and necessitated operational models and operational support systems (OSS) focused primarily on managing, as opposed to abstracting network complexity. More recently, the performance of merchant silicon has improved to the extent that it can deliver real-time line speed functionality for some communication network elements, such as those located at the network peripheries. Network operators have been implementing these elements to lower capital costs, however, in most cases they have not yet had a meaningful impact on operational models.

In recent years, network operators have come under tremendous competitive pressure from players that offer over-the-top web services. To respond to this competition, operators must dramatically reduce their capital and operational costs, and improve their customer experience and service innovation capabilities. It is these competitive pressures coupled with massive growth in network traffic demand that has fuelled interest in SDN. The aim of SDN architectures is to abstract complexity so that networks are intelligent, open, programmable and application aware. This is achieved by separating network control and data-plane functions, with early implementations leveraging standards based protocols such as OpenFlow as a means of standardizing the interface between the control and data-planes.

While the potential benefits of SDN architectures are far reaching, their implementation is complicated and can be particularly disruptive to legacy operations. SDN implementations within data-centers are gaining momentum and initial solutions in the wide area network environments are being implemented by service providers like AT&T, Telstra and Verizon. Data center implementations benefit from virtualized IT infrastructure and operations that are relatively centralized and automated. In particular, large data centers are typically run by a small number of highly skilled staff who benefit greatly from operational solutions that abstract away IT complexity. In a vision for SDNs, wide area networks could be operated in a similar manner as data centers and see BSS/OSS being implemented as part of huge distributed data center functions.

Related content

No Comments Yet! Be the first to share what you think!