Network virtualisation startup Big Switch has landed a $6.5m investment from Intel Capital Investments.
The company, founded in 2010, is part of a new generation of network virtualisation startups eyeing networking in the datacentre and supposedly posing a threat to Cisco's grip on the switch and router market. The primary goal of software-defined networking (SDN) is to enable software control of the network and replace some functions currently served by routers and switches.
Intel's $6.5m investment brings Big Switch's total funding to $45m via three rounds, beginning with $13.75m from Index Ventures and Khosla Ventures, followed by a $25m Series B round from Redpoint Ventures and Goldman Sachs.
"Datacentre operators need programmable and cost effective merchant silicon-based networking architectures to meet their datacentre economic and operational objectives," Bryan Wolf, managing director of Intel Capital, said in a statement.
The company's three main products launched last November are its open SDN platform, Big Network Controller, its datacentre network virtualisation application, Big Virtual Switch, and unified network monitoring application, Big Tap. The underlying technology behind BigSwitch is the network communications protocol OpenFlow, which the company's CEO and founder Guido Appenzeller, helped develop at Stanford University.
The acquisition that sparked more widespread interest in the SDN sector was VMware's $1.26bn acquisition of Nicira. Other players that have since staked their own claims in the hyped SDN market include Citrix, Oracle, Brocade and Juniper.
Last August Cisco claimed it was already an SDN player, pointing to its Nexus 1000V virtualised switch, but for some observers, the company had been skirting around the subject of a full-blown SDN strategy.
This month however, it gave a few more details about its SDN plans, announcing its open network strategy based on OpenFlow.
But it has been an uncomfortable subject for Cisco executives. Fielding questions about Nicira at the Citrix Synergy conference in Barcelona late last year, Cisco CTO Padmasree Warrior belittled the efforts of SDN startups as only focusing on separating the control plane from the data plane, which she said added complexity to the network and slowed network provisioning.
"Typically what VMware is trying to do when they bought Nicira and the approach startups are taking with SDN are very limited because all they focus on is separating the control plane from the data plane. When you do that you're not actually able to provision the network any faster. To provision the network faster you need to have programmability, capability in the network. So in a way that approach of simple separation of control and data plane adds more complexity versus removes complexity," said Warrior.
Warrior's comments were viewed as a red herring by some.
"Cisco has recognized that SDN is about software control of services first and foremost. The problem is that they've not really addressed how they would achieve more than the current network devices can achieve in meeting the rest of SDN goals," Tom Nolle, president of analyst CIMI Corp, told ZDNet in an email at the time.
"I at least believe that separating the "control plane and data plane" is something that's been done by router vendors (including Cisco) for well over a decade. What SDN proposes is to reduce the dependence of networks on 'adaptive behavior' that's derived from how each device interprets network control plane data to determine where to send its packets. It doesn't necessarily mean that control planes change one way or the other."
Nolle also took exception to the claim that faster network provisioning required programmability.
"What you need to provision the network faster is orchestration, meaning the ability to convert a set of abstract service goals into specific network changes. Programmability can introduce disorder and slow provisioning because direct application control over services can create collisions of goals inside the network. Do you think the average network operator or enterprise wants individual applications taking their own shot at network control? Not likely! They want some organised operational process that starts by harmonizing goals and ends by orchestrating network behavior to match the harmony."