A New Beginning in Networking – By Mike Volpi
When Guido and Kyle were looking for a name for their newly founded company, they were struggling to come up with something that captured the breadth of their vision for what they foresaw as the future of the networking business. They could have found another randomly-generated five or six letter Latin-sounding name, but that didn’t seem to do the vision or the company justice.
IP networking, the technology that powers the Internet, was originally invented by DARPA (an agency of the United States Department of Defense responsible for the development of new technologies for use by the military). Its principles were simple and, not surprisingly, designed for military use. The Internet was built using relatively autonomous devices that exchanged information with each other through pre-established protocols. This information exchange would allow for the devices to send and receive data while being resilient to one or more of them going down or being taken over (think a big bomb blowing up your router). The resulting network had some very robust attributes, which we all enjoy today as we share files on Dropbox, buy some clothes on NastyGal, or enjoy some music on Sonos.
IP networks were highly scalable (just add another device to the network and have it talk to the others); highly resilient (if a system goes down, just find another path to the destination); upgradeable (just pop in new systems one at a time), and so forth. But, all things considered, the design of Internet was brilliant and has served us well since the 1970s.
Unfortunately, any architecture is a series of tradeoffs, and even the best ones do have some drawbacks. In case of IP networks, there were three main issues that we still cope with today. First, management of these networks was a nightmare. Having to centrally administer a disparate range of elements with different operating systems, protocol implementation, etc. was awfully hard. Not surprisingly, VCs have backed an endless stream of network management companies over the years. Second, networks services (like quality of service) had been hard to implement across a network because each device had a diverse set of both “open” and proprietary features. And, last, IP networks were “shared”, by many users, but, endpoints were assume to be physical – either servers or clients.
Fast forward these issues to today, and, we find some of the most challenging problems that network managers face. Networks are hard to manage and configure. The “Noah’s Ark” of devices strewn across networks make it impossible to consistently implement features. Finally, the advent of virtualization and cloud are massive headaches for networkers who never imagined the notion of a virtual server or cloud service. This latter issue is the most acute for today’s IT departments and is perhaps the proverbial “first bowling pin.”
But, what if the network could be controlled and managed as if it were one big programmable system and still maintain its scalable architecture? Could we build a big platform layer that spanned across individual devices – both physical and virtual – and offers a universal programming interface? If we did that, the network would be one Big Switch. The vision of the company became its name.
This core value proposition has evolved over the last two years in both nomenclature – from OpenFlow to Open SDN – as well as maturity. But, the key principles are clear:
(1) A consistent range of APIs that can be used by developers and network managers alike to create network applications that are deployed across diverse network elements
(2) An underlying control plane that leverages both physical and virtual network elements to implement the applications on a packet-by-packet basis
(3) The creation of an ecosystem of applications for networks. Think iTunes and iOS and the millions of applications that run on iDevices. A network that could deploy firewalls or load balancers as a network service on demand.
The popular perception of SDN is that it will “commoditize” the routing and switching devices and usher in a new breed of low cost devices as happened with Linux and servers in the computing world. Switches may, in fact, become more cost effective with the advent of SDNs. But, the real transformation for the networking industry lies in the third principle above. And, in fact, the first beneficiary of this approach will be the cloud-based and virtualized environments which are already well-equipped for SDNs.
If we stretch our sights to the horizon, we can see how just as the enterprise application business has been transformed from premise-based software to SaaS, the networking business would undergo an analog metamorphosis where the network services would be sold and deployed as a utility rather than software running on the devices themselves. In parallel, the virtual switching universe would benefit from the use of the same services and the same architecture. This is Big Switch’s vision. To transform an industry that has grown up on pushing boxes into an entirely new era, just like what we have witnessed in the computing world over the last five years.
Today, Big Switch Networks is formally introducing itself to the world. Big Switch is a young entrepreneurial venture, but what they represent is a path into a new era for the networking and communications industry. Big Switch’s open source SDN controller – Floodlight – has been downloaded over 10,000 times since January. Microsoft, Citrix, Palo Alto Networks, Arista Networks, F5, and many other amazing companies are voting with their feet in support of Big Switch. The fledgling academic project called OpenFlow has come a long way.
We want to congratulate Guido, Kyle, Howie, Jason, Mansour, and David and the whole team there on their big day and a delighted to support them in their journey forward.