By Casimer DeCusatis Ph.D.
“The Beatles have no future in show business. Guitar groups are on the way out.” – quote following the Beatles now-infamous Decca Records audition.
There’s little question that 2013 will be considered “The Year of SDN” by the vast majority of the networking community and their clients, and with good reason. Every major networking equipment vendor, and a host of startups, brought a wide range of SDN enabled products to market this past year, officially moving the technology from slideware to actual deployments. Some analysts are predicting an incredible 60% annual growth rate for the SDN market between 2012 and 2018, culminating in a $3.5 B global market. The networking industry is going through a disruption which may be the most significant change since the invention of Ethernet itself. For example, a quick glance at the advance program for OFC 2014 shows dozens of contributed and invited talks, workshops, tutorials, and symposia devoted to this topic.
Like most overnight success stories (including the rock group mentioned in my title quote), software-defined networking (SDN) has actually been around for quite some time. The concept of separating the data plane and management/control plane is nothing new to the telecommunications industry; solutions like GMPLS have been around for decades, and have demonstrated some interesting practical applications (for a discussion on whether SDN is going to displace GMPLS, be sure to attend the OFC panel discussion on this topic). So if SDN is such a good idea, why hasn’t it already taken over the transport network?
Just ask any celebrity whose career seems to have taken off “overnight”; most of them will tell you about the many years of hard work they invested in developing their craft before being “discovered” in a breakout role. For an aspiring rock group from Liverpool, their big break arguably came on the Ed Sullivan show with a song that launched the British Invasion. Just as in the music world, having a good idea alone isn’t enough; the market needs to be receptive before a new technology gets “discovered”. For SDN, that breakout role was re-kindling innovation within the data center network. Most data center switches each contain a full copy of the vendor’s proprietary operating system - tens of thousands of lines worth of code which doesn’t interoperate well with other products. Lacking an open, standards-based API to program end-to-end traffic flows, network virtualization has lagged behind server and storage virtualization, even in large data centers. The discrepancy between provisioning a new virtual server (a few minutes) and a new network flow (days or weeks) has become a significant pain point for clients, driving adoption of a more programmable, dynamic network. This in turn has ignited a fresh round of innovation among established network equipment providers and startups alike, as noted at the beginning of this article.
There might be another breakout role for SDN waiting in the transport network. It’s well known that revenue per subscriber has not been keeping pace with a rapidly growing subscription base for some time. Despite various efforts to address this problem, many large transport providers lose a little revenue every time they add a new client. SDN offers the opportunity for short-term reductions in capital and operating expense which help reverse this trend. Further, SDN is complimentary to other cost saving measures, such as replacing physical devices with virtual ones (known as network function virtualization, or NFV). We know from many years of experience in developing server and storage virtualization for data centers that this approach yields a much more economical total cost of ownership than non-virtualized environments.
But cost savings is far from the whole story. SDN in the transport network also has the potential to enable new service offerings which would dramatically change the transport landscape. Modern telecommunications infrastructure is facing a unique series of challenges, including changing bandwidth and latency requirements from new workloads (social media, wireless, mobile, and big data analytics). Most of the existing transport network is statically provisioned; it takes days or weeks for highly skilled network administrators to re-provision new service offerings. This makes it difficult to build a business case that will cost justify such offerings, or write service agreements that depend on rapid re-provisioning in response to changing workload conditions. As a result, innovation tends to stagnate, and the cost of a new service remains high because it needs to include over-provisioning for peak bandwidth requirements, rather than typical bandwidth. By addressing these obstacles, SDN makes it possible to offer new services which were previously not economically or technically viable.
Rapid, successful implementation of these new services will be a defining trait for the next generation of transport network providers. Many companies have already begun to embrace this change; automated provisioning of service chains across the entire Layer 0-7 stack have been demonstrated in research environments and a number of compelling proof of concept test beds. There’s growing evidence that SDN for transport networks is not just desirable, but inevitable if the industry is to maintain projected cost/performance through the end of the decade.
Of course, this position isn’t without its critics, who cite the historically slow adoption rates for new transport technologies, relative immaturity of SDN, and other factors. But there were also a lot of influential people who said The Beatles would never be a hit, either. What do you think ? The symposium on SDN Transport at OFC 2014 provides a unique opportunity to review the facts about SDN and related technologies. If you think SDN is more than just the latest fad, then you can’t afford to miss this session. I hope to see you there; meanwhile, feel free to keep the discussion going by adding your comments below, or look for me on Twitter (@Dr_Casimer).
Posted: 13 January 2014 by
Casimer DeCusatis Ph.D.
| with 0 comments