Is software-defined networking (SDN) truly a savior for overburdened, overly complex network systems? Or is it just another a superfluous technology?
The Open Networking User Group (ONUG) has invited two distinguished professors to address both sides of the SDN debate as part of its spring conference, May 5-6 in New York City. Nick Feamster, professor of computer science at Georgia Tech, will represent SDN proponents, and Doug Comer, Internet research pioneer and professor of computer science at Purdue University, will defend his stance that “SDN just isn’t needed.”
Comer, who built software to control switches years before SDN became popular, will address functionality of existing Internet technologies, such as MPLS, which already provide the ability to classify packets onto flows and route each flow along a specific path. Feamster, on the other hand, will argue that many of the problems network operators face with legacy control are what motivated SDN’s central paradigm to separate data and control planes in the first place. His work on network management problems in the AT&T backbone network ultimately led to the routing control platform, a precursor to today’s SDNs.
The ONUG SDN debate aims to highlight options for IT business executives so they can make better decisions about their own network strategy. Learn more about the debate and ONUG Spring 2014 at the ONUG web site, or read the full release below.
ONUG Hosts the Great Debate on SDN:
New Perspective or Empty Promise?
April 10, 2014 – Boston, Mass. – SDN proponents, including Nick Feamster, say that SDN provides more direct control over network functions, making it easier both to develop new functions and to apply advances in formal verification and distributed systems to long standing problems with network management. But, while they may indeed pay less for commodity equipment that arises from merchant silicon, adopters still must either buy or build control software, according to Doug Comer, Internet research pioneer, Professor of Computer Science at Purdue University, and recognized authority on TCP/IP protocols. According to Comer “SDN just isn’t needed.”
Comer, who built software to control switches years before SDN became popular, will address functionality, existing Internet technologies, such as MPLS, which are already providing the ability to classify packets onto flows and route each flow along a specific path, as part of an educational review to help IT users access the right approach for their company.
“Perhaps the most misleading claim of all concerns the assertion that SDN will permit IT departments to guarantee Quality of Service (QoS), including latency, throughput, and jitter for each specific application or each TCP flow,” said Comer. “Have we forgotten that huge amounts of research and development went into ATM back in the 1990s to solve the same problem?”
Nick Feamster, Professor of Computer Science at Georgia Tech, worked on network management problems in the AT&T backbone network and his work ultimately led to the Routing Control Platform, a precursor to today’s SDNs. Feamster has developed several SDN control programs to solve longstanding network management problems in both enterprise and backbone networks. He is now leading the development of an SDN-based Internet exchange point to address shortcomings in the Internet’s interdomain routing infrastructure and teaches the Coursera online “MOOC” course on SDN.
Feamster says that many of the problems network operators face with legacy control are what motivated SDN’s central paradigm to separate data and control planes, that this concept in fact pre-dates “modern-day SDN”, and that the SDN’s direct control makes it easier to solve hard network management problems than trying to “invert” today’s low-level network control plane.
According to Feamster, “Earlier research was unable to make meaningful improvements to network management because the legacy network’s control plane is fundamentally difficult to manage. SDN’s separation of the network’s control plane from the data plane changes the game—this separation makes it easier to apply years of progress in programming languages, formal verification, and distributed systems that previous attempts to improve network management were not able to leverage.”
“For the past two years, vendors in the networking industry have had a tumultuous relationship with SDN, with some rejecting SDN, others embracing it, and yet others being ambivalent. Organizations like ONUG who try to provide insight and a level playing field play a critical role in helping the user community sort through the noise” said Roy Chua, Co-Founder and Partner at SDNCentral. “With these debates and other ONUG events, the user community has an opportunity to engage and learn directly from IT business leaders who have evaluated and deployed early solutions, and academic leaders who will raise the pros and cons of both sides. This empowers end users to make their own decisions about timing and direction for their networking strategy.”
The Open Networking User Group (ONUG), has invited these two distinguished professors to address both sides of the SDN debate as a way to highlight options for IT business executives, along with providing multiple use cases for users to best evaluate different approaches to Software-Defined WANs and other use cases.
The ONUG Spring 2014 agenda also includes early adopters of open networking and open storage solutions, spanning diverse industries. Speakers include IT business leaders from Bank of America, Cigna, Citigroup, Credit Suisse, FedEx, Fidelity Investments, GAP, Inc., Lucera, NTT, Pfizer, Sears, UBS, and others. These IT executives will discuss their requirements, evaluation processes, challenges, and the benefits realized by their open networking and SDN pilots or deployments.
The ONUG Spring conference will take on a different format this year, with conference sessions from IT business leaders spanning both days. The morning keynotes and panels will be held exclusively for IT end users, while both afternoons both will include a combination of speaking sessions and tutorials open to the industry at large. Also new to this spring’s conference, are intimate Fireside Chats with ONUG board members, the ONUG SDN Hackathon, and a VXLAN interoperability demo.
Vendors participating in the open networking solutions demonstration and supporting ONUG Spring 2014 include: Adara Networks, Avi Networks, Cisco, Cleversafe, F5 Networks, Glue Networks, HP, Intel, Ixia, NEC, Nuage Networks, Pluribus Networks, SageCloud, Stateless Networks, vArmour Networks, viptela, and VMware.