It’s possible that OpenFlow is the next big thing. It’s more likely that Software Defined Networking, with or without OpenFlow, is among the next big things. But it’s dead certain that Single Point of Management is a damn big thing and top of mind for cloud builders everywhere. And this is what makes SDN so exciting.
Applications are moving to the cloud, desktops are becoming virtualized, video is streaming end-to-end across the entire mess. As a result campus networks, data center networks, metro networks and the Internet itself are growing at a rate and scale never seen before. And the rate of change is increasing rapidly. Where it used to be merely expensive and inefficient to manually configure, monitor, troubleshoot and update the large numbers of network devices, it is becoming impossible to do so for a number of devices that is growing by orders of magnitude. Network builders everywhere are screaming for relief and all agree that they need a simple interface for describing the network that can support their workloads and they only want to tell us once. They want to connect to exactly one “management system” and configure the (virtual, abstracted) network by defining the requirements for their applications and get on with the thing that actually makes them money.
This describes the hottest trend, begging for the biggest disruption in networking in decades and SDN/OpenFlow haven’t been part of the description. Why then is SDN something viewed as so promising and important? Because it is a great architectural approach to delivering the single point of management sought by the market. Us geeks are constantly accused of falling in love with solutions looking for a problem. SDN is the solution to THE problem. It’s not the only possible solution but it fits the problem nicely, has big industry momentum, and is blessed and promoted by some seriously clever propeller-heads. OpenFlow has captured the “flow processing” core of the problem very nicely and looks to have have enough smart engineers aimed at it through ONF and other initiatives to end up being a great wire protocol. However it can never be more than that; the southbound API between controller and commodity switch. Keep in mind that this lives a couple of rungs down the abstraction stack from the big user problem we’re trying to solve here.
Big incumbent vendors can be expected to respond to the SDN/OF threat, which could commoditize the switches and squeeze their profit margins, by building single-point-of-managent solutions that can be called SDN and don’t necessarily use OpenFlow. In the campus, Brocade is touting new solutions delivering a “single point of management” without any mention of SDN or OF. The Hybrid OpenFlow switch, combining traditional L2/L3 protocols with OF based overlay flow handling, is a truly compelling concept. It’s also one that will appeal to tier one network equipment manufacturers that sell an L3/L4 switch for six figures today and are terrified by commodity switches running openflow and available for four figures straight from Taiwan. Cisco’s not-very-secret spin-in, Insieme Networks, can be expected to deliver something that meets the definition of SDN, yet requires proprietary hardware with juicy margins.
The race is on, but remember that the finish line for many end-users building clouds is delivering network virtualization through a single point of management. If the BSNs (Big Switch Networks) and Nicira’s can’t finish that story before the usual suspects turn it into a feature on their latest really expensive boxes, many cloud architects may go back to the comfort and safety of their incumbent vendor. The revolution could be called on account of rain.