What's OpenFlow's killer app?
- 21 October, 2011 01:35
PALO ALTO, Calif. -- As most participants at the inaugural Open Networking Summit extolled the virtues of software-defined networking (SDN) and the OpenFlow protocol, others are waiting for key features to emerge before considering it for their networks.
OpenFlow and SDN allow a network to be programmed as if it were a computer, proponents of the technology say. It provides a layer of abstraction between the physical network to the control element, allowing the network to be configured or manipulated through software, which then opens it up to further innovation.
FAQ: What is OpenFlow?
But missing from OpenFlow, some presenters said, is support for IPv6, MAC-in-MAC Provider Backbone Bridging, Q-in-Q VLAN stacking, QoS and traffic shaping capabilities, fault tolerance and scalability. Some said these gaps limit the OpenFlow protocol's, and SDN's, applicability to their networks.
"We're very much looking forward to IPv6," said Stephen Stuart, distinguished engineer at Google. "Current SDN implementations don't speak the standard protocols" such as IPv6, MPLS, IS-IS with TE, RSVP-TE and BGP. "More people would try out OpenFlow and SDN if they spoke standard protocols."
IPv6 will be one of a number of features coming in Version 1.2 of OpenFlow in December, said Dan Pitt, vice chairman of the summit and executive director of the Open Networking Foundation, an organization driving standardization and adoption of OpenFlow and SDN, and a co-sponsor of the summit.
Others said OpenFlow and SDN are more useful to huge data center-driven companies like Google and network operators than they are to the enterprise.
"I completely embrace the disruptive potential but I'm still puzzled on how it will impact the enterprise," said Peter Christy, co-founder of the Internet Research Group, a Los Altos, Calif., consultancy. "The enterprise is the biggest part of the networking market. Building a reliable SDN controller is a challenging task. Customers don't value reproducing the same thing with a new technology. There have been few OF/SDN 'killer' apps so far. SDN is not ready for the enterprise market yet."
Christy spoke during a panel session on opportunities and challenges with OpenFlow and SDN in the market. Others on that panel took issue with his assessment.
SDN will increase network agility and virtualization in the enterprise, and serve as a common control plane for both wired and wireless LAN infrastructures, said Geng Lin, CTO of the networking business at Dell. "We need the hypervisor and the physical network to work together, collaborate, influence each other" through SDN, Lin said.
Amin Vahdat, a principal engineer at Google, said, "I think the killer app is out there: Stanford is running the Gates building on OpenFlow." Vahdat was referring to Stanford University's use of the protocol it helped create in it computer science department. "It's doing lots of things that are really, really hard right now."
Indeed, these issues are not stopping research and academia from putting OpenFlow and SDN through the gauntlet. Stanford, Indiana University and Internet 2 are engaged in the Network Development and Deployment Initiative (NDDI), touted as the first production deployment of OpenFlow technology for an SDN that will allow researchers to experiment with Internet protocols and architectures, while at the same time enabling scientists to conduct research with collaborators worldwide.
MORE SUMMIT COVERAGE: Ex-Facebook exec pushes software-defined networks
NDDI will support projects ranging from the global exchange of massive datasets from the Large Hadron Collider, radio astronomy and climate modeling experiments, to large-scale network research initiatives like the National Science Foundation's GENI project and Princeton University's PlanetLab, the NDDI website says.
"If we create the capacity for innovation, get it widely dispersed and let people play, SDN will turn into something wild," said Rob Vietzke, executive director of network services for Internet 2.
"Sometimes you have to challenge the status quo," added Matt Davy, Indiana University's chief network architect.
That's what Verizon plans to do with SDN, though not for something wild -- something practical. The carrier sees SDN as a way to reduce cost in its infrastructure.
"Why does Verizon care? It's economics," said Stuart Elby, chief technologist of Verizon Digital Media Services. "The reason is for cost. There's a tremendous reason to do this quickly based on cost savings."
Vertically integrated products are not following the trend of revenue, Elby says. SDN will help balance that out by integrating those products horizontally through virtualization and a common control plane for programmability.
OpenFlow could help Verizon optimize its network for video delivery by enabling traffic steering for long-lived flows and cut-through switching to reduce cost, and for hybrid cloud computing services to enable bandwidth-on-demand for data center interconnection.
But like other speakers, Elby said OpenFlow itself will need to be optimized for the task.
"OpenFlow requires several enhancements to work effectively in virtualized cloud environments with lots of legacy infrastructure," he said.
Read more about lan and wan in Network World's LAN & WAN section.