Virtualization has revolutionized the data center, improving the utilization, efficiency, and reliability of IT, but its progress has been stymied by inefficiency in networks. Software-defined networking (SDN), considered as an alternative to legacy networks, transfers intelligence in switches and routers to software running on a server that dictates how networking equipment applies rules and forwards data packets.

The Open Networking Summit, held Monday through Wednesday at Stanford University, brought together academia, network equipment vendors, and others to discuss OpenFlow, the emerging standard for how SDNs would work.

In a virtualized environment, virtual machines (VMs) can be created in short order and VM workloads can be moved around from one physical server to another as needed, but existing Ethernet or TCP-IP networks can limit that flexibility based on the configuration of each router or switch.

[Check out 4 steps to improve the security of your virtualization infrastructure with the Virtualization Security Checklist.]

In the same way that server virtualization abstracts the operating system and the VM from the physical server, OpenFlow-based SDN abstracts the network "control plane" from the physical network hardware, said Stanford's Guru Parulkar, chair of the first-ever summit on SDN technology. The control plane is the set of instructions that direct how packets are managed on the network.

"With OpenFlow and SDN, the promise is that the networking interface will also mature to the point that you can use it in a plug-and-play way so that when you are provisioning an application or a service ... they are able to create a virtual network of their own specification," Parulkar said.

While networking industry giants like Cisco Systems and HP were represented, the conference also featured startups like Big Switch Networks, which is developing OpenFlow-based products, and Arista Networks, whose networking technology is used in some high-performance computing environments.

Ken Duda, founder and CTO of Arista, said a typical network of 400 switches, 2,000 IP subnets, 1,600 transit links, and multiple different configurations is too complex and unwieldy to manage and maximize the benefits of virtualization.

"This architecture is just made-to-order for some kind of central control automated for provisioning, automated for some kind of administration. This scenario is doable with today's switches," Duda said. While there is still some debate as to how much of a game changer SDN is, because it replaces the configuration done in switches and routers it is seen as a competitive threat to Cisco, whose highly engineered equipment comes at a high price tag for customers and a healthy profit margin for the company.

But Cisco is well aware of what's coming and is a member of the OpenFlow Foundation, which supports the development of the OpenFlow standard.

"[Cisco] folks get this and how to react to it is what's being formulated right now," said David Meyer, a Cisco fellow, in answer to a question from the audience at the summit. He prefaced his remarks as his own and not an official Cisco statement.

"It's very obvious to everyone that something's going on here and the question is how to react to it in a way that everybody can live with," Meyer said.

Charles Clark, research director in HP's networking business, said at the summit that he prefers some type of hybrid switch that has its own configuration, but can also be controlled by the SDN controller. His rationale is that a campus LAN administrator "won't tolerate a forklift upgrade, they aren't going to replace all of their equipment."

Creating cloudlike enterprise data centers means pooling, virtualizing, and sharing servers, networks, and storage. In this report, we examine HP's Matrix to discover how it knits together the components of a converged architecture promising new levels of efficiency, flexibility, and scalability. Download it now. (Free with registration.)