Needed: An Open SDN Platform

Tuesday Jul 8th 2014 by Guest Author
Share:

Software defined networking vendors would do themselves and customers a favor if they focused on open interoperability.

Guest editorial by Neela Jacques

Also by Neela Jacques: Is 'Open' the 'Organic' of the IT Industry?

Battles between competitors can be either good or really bad for customers. It’s good when it leads to innovation of course, but what we are seeing right now in software defined networking (SDN) is the potential for destructive competition that will lead nowhere.

There was a fascinating tweet storm when Nexus’s Colin McNamara showed his displeasure with Cisco and VMware for their battle. The problem in this case is while Cisco and VMware are putting out fundamentally different visions of networking, most customers just don’t want to or frankly can’t chose between the two.

The reality is Cisco doesn’t make hypervisors and VMware doesn’t sell hardware. In many cases a company will view both VMware and Cisco as strategic vendors. So what type of “open” do end users want from Cisco,  VMware and other vendors? They want to be able to pick solutions that work and are supported by their vendors. They want technologies that work, but more importantly technologies that work together.

I want to take a moment to double click on these two major players. Both VMware and Cisco have been talking a lot lately about how “open” they are. VMware has been touting open standards for a while, either via IETF/DMTF standards, the open (but little used) vCloud API, or more recently through their investment in and support for OpenStack. Cisco has recently created a business unit to invest in open source called Noiro Networks. Cisco also touts a long history of participation in the IETF. Both companies, of course, are members of OpenDaylight; Cisco at the Platinum level and VMware at the Gold level.

So let’s dive right in, starting with Cisco. Are their new Group Policy and OpFlex projects “open”? It’s a very important question to Cisco because if they are, then by association CISCO APIC might be deemed “open.” In fact it may mean Cisco APIC is <i>more</i> open than VMware NSX. Given that customers are clearly saying that openness is a key requirement for their SDN solutions, this one question may be worth billions in the long run.

Here’s my opinion. The fact that Cisco is taking what might otherwise be a proprietary standard is a good thing. It’s good for end users and it’s good for Cisco. But only time will tell if it’s truly “open.”

Here is the litmus test for me: Are the IETF and OpenDaylight submissions take-it-as-is or not at all? It’s hard to believe that Cisco got everything right the first time. Part of the reason we all want open standards and open source is that many different minds work better than a few. Is Cisco open to other people’s suggestions and contributions? I would be very concerned if the IETF passed OpFlex without one substantive amendment.

On the OpenDaylight side it is very encouraging that IBM, Midokura and Plexxi have signed up to be involved, so we’ll want to watch how much they’re able to influence Mike Dvorkin and Kyle Mestery of Noiro Networks. Will the project draw other great minds to the table and become a true collaboration among industry heavyweights, or remain one vendor’s view of the world?

Knowing Mike and Kyle, I have full faith in their desire to do this right, but it takes two to tango. I invite folks to jump in and have the debate with us. OpenDaylight is THE place where the biggest questions in networking are being debated – with code not PowerPoint. I urge the thought leaders in the industry to jump into the debate either by joining the Group Policy Plug-in project, by building out the Open vSwitch (OVS) Database Integration or even by starting a new project. I urge Cisco to seek out ever-greater diversity for their OpenDaylight projects. Diversity of contributors, but also of approaches. Consider nothing sacred. Demonstrate your willingness to build something that doesn’t look exactly like what you’ve already built. 

At the same time I am very interested in the Congress project that Martin Casado of VMware has been proposing to OpenStack. It’s an ambitious project to bring Policy-as-a-Service to cloud services. Is this VMware’s way of getting their view of the world accepted as a standard, or are they truly willing to collaborate with others to develop one common way to describe the application policy needs that we all sorely need?

I wonder if it will be possible to bring Group Policy and Congress together. They aren’t the same things of course, but there’s plenty of room to bring the two projects together.

On a similar note, why do we have so many vSwitches in the world? When Nicira was locked out of the VMware ecosystem by a proprietary vSphere architecture, they built a great open vSwitch, which is appropriately called OVS. It is now at the heart of VMware’s NSX multi-hypervisor offering and is gaining tremendous mindshare. But the only way to use OVS with over 60 percent of the VMs in the world is to buy NSX, as VMware continues to cut out the open community from the vSphere party. I love VMware for so many reasons. I love vSphere, I love OVS and I love how they are investing in making vSphere a real choice in an OpenStack infrastructure. I call on VMware to give OVS equal status (kernel access) as vDS, and the proprietary vSwitches they’ve let Cisco, IBM and HP build. Be open. Let the next Niciras innovate.

I’m putting a spotlight on Cisco and VMware, but I challenge every vendor with a proprietary SDN controller to either clearly articulate why they have found a game-changing technology no one else has, or work with the open source community to improve the one we’ve got and make it core to their offering. “Open” isn’t just a word to dress-up product data sheets. Open represents a real set of attributes that really matter to end users. Interoperability isn’t easy, it’s hard.

We don’t need 30 SDN controllers in the world. They all do pretty much the same thing and customers don’t want to pay a dime for the controller. The value is in the network services, the hardware and the management layer. Not the controllers.

We are on the cusp of something big. Now is the time for vendors of networking technology to walk the walk. As an industry we sorely need a common SDN platform. I see a world where customers have a great range of technologies, built around a range of valid models, but that are all based on one common code base. I believe that newer technologies and approaches will need to co-exist with older technologies and protocols.

This is why we need an SDN platform with the ability to work with a wide range of protocols, not just OpenFlow, OVSDB or Netconf. I have yet to hear a compelling reason why this is not the right vision for the industry, just a lot of comments that it’s hard to pull off. No kidding. But the OpenDaylight community is well on the way to achieving this vision. We will achieve it if we continue to get more contributors and more people sharing opinions, but also if more people listen and change their opinions. Together we can change the world.

Neela Jacques is Executive Director at OpenDaylight Project

Share:
Home
Mobile Site | Full Site
Copyright 2017 © QuinStreet Inc. All Rights Reserved