Plenty of vendors are hard at work on SDN technology, but mostly they're focusing on two areas: enabling southbound control of network gear by SDN controllers and developing the first generation of controllers to perform useful -- not just cool -- work. The problem is there is very little work in creating frameworks that allow disparate SDN controllers to interoperate. Yet without the flexibility to mix and match multivendor controllers, administrators will continue to face the challenge of juggling multiple panes of glass. This may trade GUIs for green screens but won't solve the problem.
Is the current SDN controller interoperability model anti-interoperability?
There's no question that the controller is the shiniest component of software-defined networking. For network admins, it's where we'll brew our magic spells to create programmable networks and take advantage of the underlying SDN standards, such as OpenFlow, that these controllers rely on. For vendors it's the repository of actionable processes in the SDN world, a completely new network component category and, therefore, also the part that provides a new product to add to customer sales engagements.
Different vendors will steep their controllers with the functionalities of their domain expertise. Some will excel at seamless management and migration of virtual machines in the data center. Others provide new levels of global security policy assessment. Some may specialize in managing the edge and distributed elements of remote campuses. But the greatest application of SDN is one where several controllers work as a coordinated team, each lending their charms to a greater whole.
Yet getting controllers from different vendors to interoperate opposes the current model in which controllers speak directly to network devices.
Reviewing the proposed standards, we see little in the way of unilateral and bilateral trust management specifications, distributed authority context, delegated change permissions or security inheritance. It's clear that the real opportunity of peer-to-peer controller cooperation is being ignored, perhaps intentionally, across vendor boundaries. This threatens to lock admins into a single or a handful of vendors -- which means business as usual in networking.
Cisco is bucking SDN interoperability with Internet-of-Everything plans
Perhaps the most interesting message during Cisco Live 2013 was the Internet of Everything strategy -- the proposition that 99% of the world's devices still aren't on the Internet but must eventually be connected. These devices won't be connected using traditional management technologies, but by using SDN and programmable networks instead. Our current technologies are just too manual to scale.
You would think this would be SDN's moment in the sun, but you'd be wrong.
Instead vendors are generally proposing their own paths. Cisco Systems is promoting its next generation ASIC with 4 billion transistors [and] 1.5 million lines of software code. This approach meets the tenants of SDN by providing virtualization and network programmability and is likely to arrive years before an easy-to-assemble, open-standards-based solution. But it is not SDN -- it's Cisco's SDN. Cisco CEO John Chambers underscored this when he said during a Cisco Live panel discussion, "We've moved past SDN."
Closed SDN architectures, missed vendor opportunities
Perhaps this problem is arising because most network vendors come from hardware backgrounds where proprietary approaches rule. Yet the networking industry is losing an opportunity when it refuses to fight against the unknown market effects of open SDN and true interoperability. Cisco, HP Networking, Juniper Networks and others offer different areas of feature excellence. This is the reason why a predominantly Cisco shop may still select Palo Alto routers for specific uses. The same goes in the software world, where no vendor has been all things to all customers.
Instead, what has succeeded time and again is when vendors focus on accessibility in their areas of established expertise. When they get those features to market quickly, they win. They even win if customers must occasionally bolt on specialist products to cover a business-specific use case. And in that scenario, when vendors quickly make established features available in new and innovative ways, who really wins? Everybody. Especially us, the network admins.