For fans of software defined networking (SDN) — especially those paid to make sense of each vendor claim and then advise clients — this week brought a bonanza of news. Both Cisco and Juniper, the largest players in the networking space and the two with the most to lose as networking becomes more about software and less about smart boxes, both outlined their strategy for coping with SDNs.
Both companies’ announced strategies that involve downplaying the role of the OpenFlow protocol and the division of the control and data planes that OpenFlow allows, saying most of their customers don’t really aspire to running so-called “dumb networks” filled with commodity gear. This was, frankly, unsurprising. Flatter networks aren’t for everyone.
It’s clear that Cisco and Juniper recognize the data center switching market as commoditized, and are aiming their efforts at helping smaller firms, corporate networks and service providers make their networks more agile using versions of what each dub SDNs — although their customers will pay for that privilege.
On Tuesday, Juniper explained its strategy at a series of events including its financial analyst day. Juniper plans to support OpenFlow, BigSwitch’s open-source OpenFlow controller called Floodlight, and will open up its software development kits so others can program to its hardware. It will still offer pricier and more high-margin gear for service providers inside their core and edge networks, but is taking a pretty customer-led point of view there.
Basically, Juniper’s willing to offer everything to everyone. It understands that some of its business will get commoditized, but it still believes there are customers and markets willing to pay for specialized hardware. It’s a pretty realpolitik view of the world, which is smart considering that Juniper in many ways saw the bottleneck that the network would become a few years back.
The following day, Cisco executives took the stage at Cisco live in San Diego,Calif. to make sure the world knew its story around SDNs. Cisco has been explaining that it would offer service provider customers a fatter not flatter network with programmability for the last few months, but more details on the overall strategy were unveiled at the event.
For example, Cisco says it is developing a controller, seen as an essential element to creating a SDN and as the key to undercutting Cisco’s proprietary boxes. Nicira and BigSwitch, for instance, are both selling controller software that runs on commodity hardware, where the network intelligence would reside.
That version of a flatter network doesn’t appeal to Cisco, which is still a hardware company despite its efforts to write software. So other aspects of its plan, such as an open API for developers to write networking apps, make more sense. With the API Cisco gives customers more agility because they can program its boxes more easily and tie them to their applications. But they will also be programming to the Cisco switches and routers, as opposed to a controller.
Cisco’s hope is that it can convince its users that they don’t want to invest in the IT resources to support a truly software defined network as startups such as Nicira or BigSwitch (or even web companies like Google or NTT) are deploying. Instead Cisco will offer ways to give customer programmability, while still selling them smart boxes.
Together these two events and stories from networking’s biggest players will help shape the discussion around software-defined networks for years to come. Whether that’s a good thing or a bad thing depends on where you sit in the ecosystem. It’s still very early in the emergence of this technology, and the route to achieving a software defined networks isn’t yet mapped. So far most of the landscape and imagery has been filled in by webscale customers and startups. Now it’s time to see what everyone else wants.