Brocade stepped up to the plate once again to present to the assembled delegates at Network Field Day 5. I’ve been constantly impressed with what they bring each time they come to the party. Sometimes it’s a fun demo. Other times its a great discussion around OpenFlow. With two hours to spend, I wanted to see how Brocade would steer this conversation. I could guarantee that it would involve elements of software defined networking (SDN), as Brocade has quietly been assembling a platoon on SDN-focused luminaries. What I came away with surprised even me.
Mike Schiff takes up the reigns from Lisa Caywood for the title of Mercifully Short Introductions. I’m glad that Brocade assumes that we just need a short overview for both ourselves and the people watching online. At this point, if you are unsure of who Brocade is you won’t get a feel for it in eight short minutes.
Curt Beckman started off with fifteen minutes of discussion about where the Open Networking Foundation (ONF) is concentrating on development. Because Curt is the chairman of the ONF, we kind of unloaded on him a bit about how the ONF should really be called the “Open-to-those-with-$30,000-to-spare Networking Foundation”. That barrier to entry really makes it difficult for non-vendors to have any say in the matters of OpenFlow. Indeed, the entry fee was put in place specifically to deter those not materially interested in creating OpenFlow based products from discussing the protocol. Instead, you have the same incumbent vendors that make non-OpenFlow devices today steering the future of the standard. Unlike the IETF, you can’t just sign up for the mailing list or show up to the meetings and say your peace. You have to have buy in, both literally and figuratively. I proposed the hare-brained idea of creating a Kickstarter project to raise the necessary $30,000 for the purpose of putting a representative of “the people” in the ONF. In discussions that I’ve had before with IETF folks they all told me you tend to see the same thing over and over again. Real people don’t sit on committees. The IETF is full of academics that argue of the purity of an OAM design and have never actually implemented something like that in reality. Conversely, the ONF is now filled with deep pocketed people that are more concerned with how they can use OpenFlow to sell a few more switches rather than now best to implement the protocol in reality. If you’d like to donate to an ONF Kickstarter project, just let me know and I’ll fire it up. Be warned – I’m planning on putting Greg Ferro (@etherealmind) and Brent Salisbury (@networkstatic) on the board. I figure that should solve all my OpenFlow problems.
The long presentation of this hour was all about OpenFlow and hybrid switching. I’ve seen some of the aspects of this in my day job. One of the ISPs in my area is trying to bring a 100G circuit into the state for Internet2 SDN-enabled links. The demo that I saw in their office was pretty spiffy. You could slice off any section of the network and automatically build a path between two nodes with a few simple clicks. Brocade expanded my horizons of where these super fast circuits were being deployed with discussions of QUILT and GENI as well as talking about projects across the ocean in Australia and Japan. I also loved the discussions around “phasing” SDN into your existing network. Brocade realizes that no one is going to drop everything they currently have and put up an full SDN network all at once. Instead, most people are going to put in a few SDN-enabled devices and move some flows to them at first both as a test and as a way to begin new architecture. Just like remodeling a house, you have to start somewhere and shore up a few areas before you can really being to change the way everything is laid out. That is where the network will eventually lead to being fully software defined down the road. Just realize that it will take time to get there.
Next up was a short update from Vyatta. They couldn’t really go into a lot of detail about what they were doing, as they were still busy getting digested by Brocade after being acquired. I don’t have a lot to say about them specifically, but there is one thing I thought about as I mulled over their presentation. I’m not sure how much Vyatta plays into the greater SDN story when you think about things like full API programmability, orchestration, and even OpenFlow. Rather than being SDN, I think products like Vyatta and even Cisco’s Nexus 1000v should instead be called NDS – Networking Done (by) Software. If you’re doing Network Function Virtualization (NFV), how much of that is really software definition versus doing your old stuff in a new way? I’ve got some more, deeper thoughts on this subject down the road. I just wanted to put something out there about making sure that what you’re doing really is SDN instead of NDS, which is a really difficult moving target to hit because the definition of what SDN really does changes from day to day.
Up next is David Meyer talking about Macro Trends in Networking. Ho-ly crap. This is by far my favorite video from NFD5. I can say that with comfort because I’ve watched it five times already. David Meyer is a lot like Victor Shtrom from Ruckus at WFD2. He broke my brain after this presentation. He’s just a guy with some ideas that he wants to talk about. Except those ideas are radical and cut right to the core of things going on in the industry today. Let me try to form some thoughts out of the video above, which I highly recommend you watch in its entirety with no distractions. Also, have a pen and paper handy – it helps.
David is talking about networks from a systems analysis perspective. As we add controls and rules and interaction to a fragile system, we increase the robustness of that system. Past a certain point, though, all those extra features end up harming the system. While we can cut down on rules and oversight, ultimately we can’t create a truly robust system until we can remove a large portion of the human element. That’s what SDN is trying to do. By allowing humans to interact with the rules and not the network itself you can increase the survivability of the system. When we talk about complex systems, we really talk about increasing their robustness while at the same time adding features and flexibility. That’s where things like SDN come into the discussion in the networking system. SDN allows us to constrain the fragility of a system by creating a rigid framework to reduce the complexity. That’s the “bow tie” diagram about halfway in. We have lots of rules and very little interaction from agents that can cause fragility. When the outputs come out of SDN, the are flexible and unconstrained again but very unlikely to contribute to fragility in the system. That’s just one of the things I took away from this presentation. There are several more that I’d love to discuss down the road once I’ve finished cooking them in my brain. For now, just know that I plan on watching this presentation several more times in the coming weeks. There’s so much good stuff in such a short time frame. I wish I could have two hours with David Meyer to just chat about all this crazy goodness.
If you’d like to learn more about Brocade, you can check out their website at http://www.brocade.com. You can also follow them on Twitter as @BRCDcomm
Brocade gets it. They’ve consistently been running in the front of the pack in the whole SDN race. They understand things like OpenFlow. They see where the applications are and how to implement them in their products. They engage with the builders of what will eventually become the new SDN world. The discussions that we have with Curt Beckman and David Meyer show that there are some deep thinkers that are genuinely invested in the future of SDN and not just looking to productize it. Mark my words – Brocade is poised to leverage their prowess in SDN to move up the ladder when it comes to market share in the networking world. I’m not saying this lightly either. There’s an adage attributed to Wayne Gretskey – “Don’t skate where the puck is. Skate where the puck is going.” I think Brocade is one of the few networking companies that’s figured out where the puck is going.
Tech Field Day Disclaimer
Brocade was a sponsor of Network Field Day 5. As such, they were responsible for covering a portion of my travel and lodging expenses while attending Network Field Day 5. In addition, Brocade provided a USB drive of marketing material and two notepads styled after RFC 2460. At no time did they ask for, nor where they promised any kind of consideration in the writing of this review. The opinions and analysis provided within are my own and any errors or omissions are mine and mine alone.
Blog comments are rarely the best place for longer discussions, so I’ll keep this brief and we can continue in Orlando, if not before.
In my mind, SDN is more evolutionary than revolutionary, and I’m not sure yet how I view it all. So far I tend to view the whole thing as a retread of other ideas we’ve seen in the industry for years and that haven’t necessarily achieved any particular levels of anything…
Let me explain. Think about the big Tier-1 ERP systems, like an Oracle Financials for instance. The front end *simplifies* the interaction with the backend *dangerous* part (the database) and let’s you hook in with APIs or whatnot; you don’t interact with the database. Except that you do. You hire Oracle DBAs to run your database… oh, and then you hire front-end developers (more closely related to software engineers than IT) to work the front end systems.
At the end of the day, what do you have? A hell of a lot of complexity, interactions, APIs, interfaces, and more places where things break. I see the same thing long term from SDN. Eventually you’ll have your hard-core engineers running the back-end and getting in when the front-end fails or can’t do a thing. Then you’ll have a team of front end developers maintaining the rules-bases, etc. On a large scale, you’ll probably lower your headcount a bit, but not by as much as I suspect a lot of people think.
In the end, will any of this lead to more stability and less human error? On the surface, perhaps. But you’re simply shifting your complexities and failure points somewhere else. It’s like collapsing layer-2 so you can kill spanning-tree… that’s great, but now you’ve moved your cheese to layer-3 where it may or may not be more stable.
In fact, I think I just stumbled on my new motto for SDN: Moving the cheese.
By the way, I feel bad that I didn’t point out that I love reading your take on all of this and your writing is excellent, as always. 🙂
Pingback: Brocade’s Pragmatically Defined Network | The Networking Nerd