Early this year, I wrote a post I called "What is all the hype about OpenFlow", and this really is the crux of the matter: is the excitement around Software Defined Networking (SDN) real, or is it being over-hyped. All significant technology movements must address real, serious problems. The main theme I took away from today is that OpenFlow and SDN are not simply great ideas to improve networking technology; these technologies address truly critical problems. Another important question is that, perhaps SDN is more than hype, but that doesn't necessarily mean it will have a lasting impact. There have been many 'almost' movements in the past based on sound principles, that have fizzled often do to political squabbling. But as I said in my day 2 recap, I am convinced that SDN is different and will fundamentally reshape the way that networking solutions are built and how networks are deployed and operated. The #1 reason: money. The single biggest thing that was said that really demonstrated why SDN will grow, and grow quickly was from Verizon's Chief technologist for Digital Media, Stuart Elby who presented the following slide (sorry for the poor picture quality):
If you look at the upper right hand corner of the chart, you will see a black line intersecting with a double red line. The double red line represents Verizon's revenue, and the black line represents the current trajectory of network operating costs. As you can see, with current generation technologies, rising costs will very quickly and dramatically surpass revenues if nothing changes. This leads me to the #2 reason why SDN will become a carrier-grade reality very quickly: consensus. SDN may not be the only way to address these challenges, but according to Elby, SDN is where Verizon is looking to solve this challenge: "We see tremendous reason to go ahead and do this quickly, because of cost savings" noted Elby who also stated that Verizon expects much more than incremental cost savings from SDN, which at the same time will deliver substantial increases in business agility. Beyond Verizon, a quick look at the roster of presenters at the summit shows the leaders of the tech industry are not just attending, they are advocating, and each of them appear to be developing concrete plans around SDN. The promise of SDN seems to hit the sweet spot of what businesses are looking for, a significant increase in agility and innovation capabilities with a significant decrease in operational cost savings, if only this had a human element, then it could have a trifecta ... enter reason #3: The human element.
What do I mean by this? I mean that in order for SDN to gain momentum, it must result in products being sold, and one must consider the people who influence the purchasing decision. In sales, the overriding philosophy is to try to make your sales pitch to the highest possible title in the company, but this doesn't quite work in the field of networking. Networking is quite complex, which has resulted in engineers having more influence on purchasing decisions than engineers do in many other domains of computing. The engineers who must support decisions to make changes in network products and processes don't have to pull the check out of their personal piggy bank. I am not saying these engineers don't care about saving their business money, they do; however one also has to consider when a new vendor tries to sell their product, that the engineer is the one left holding the bag. They are the ones that have to potentially relearn aspects of their job, they are the ones that get called in the middle of the night when something fails and they are the ones whose necks are on the line when outages occur. When visiting the product demos, I was able to provision a national-scale MPLS VPN network in about 4 mouse-clicks; it was then I saw firsthand that the human aspect of SDN was very real.
The summit was filled with technologists and engineers that were excited about SDN, and for technologists I don't think this excitement has to do with money, the excitement is found in the significant, meaningful advancement in technology that will make the lives of network engineers better and their jobs much more enjoyable. Google principle engineer Dr. Amin Vahdat really drove this point home, continuing a line of thought suggested by Dr. Scott Shenker suggesting that many engineers enjoy or find value in being masters of complexity. Vahdat posed that if network engineers like taking the risks necessary to operate networks today they must be 'adrenaline junkies', noting that common network operations often require downtime and can be quite risky. For example it is well established that a large percent of network outages are due to operator error. Dr. Vahdat presented the following slide (Note the last bullet point says "I introduce a product from a new vendor")
Vahdat stated that the resultant overriding philosophy is in effect, "If we don't touch the network, maybe it won't break", which results in "a network that gets in my way rather than enables computation and data storage". Vahdat noted that while servers and storage are incrementally updated frequently, the network upgrade path is out of sync with the rest of infrastructure which he says results in "millions of dollars of compute and storage systems under-utilized". He noted that the typical upgrade path for the network is to "amortize the network over 6 to 8 years and then tear it down and start over".
Vahdat made another interesting point, which is that the complexity of the network leads to "fundamental under-investment in network". Now this sounds a little strange, I haven't run into many enterprises thinking 'gee, I'm just not spending enough on my network". But what I think Vahdat meant by this was that there are many network problems that applications try to solve through complicated and convoluted techniques in the application layer because they can't simply talk to the network. This point was really driven home by Juniper's David Ward, who gave several examples of multimedia applications that utilize various methods in the application layer to try to gain a derivative, poor understanding of the network. Ward presented the following slide:
So I don't think the point is that we need to spend more on the network, rather that application developers have to jump through a lot of hoops at the application layer rather than involving the network directly in the application development. To see some excellent applications of this concept, I recommend checking out some of the demos presented by Stanford's grad students from the conference which will be available here. Many of the projects demoed at the summit and others can also be found on the OpenFlow site here.
So all of these cool applications sound great, but when will they become a reality for enterprise IT? The two largest North American vendors that gave product demos were HP and Dell. I am a Dell employee, I am not a Dell spokesperson, and I should note I do not work in the product group or marketing, and I don't have any exposure to roadmap and strategy so please note any observations I make here are speculation based on their presentation and demo. I am trying to be unbiased so I have included videos of the Dell & HP presentations and demonstrations so feel free to draw your own conclusions. It was exciting to see both of these tech titans taking in active role in developing and promoting SDN, both delivered excellent presentations. I feel that Dell Networking CTO Geng Lin's presentation was particularly notable because he gave direct insight into Dell's comprehensive product strategy around SDN. Dr. Lin's presentation indicated that Dell has plans for Data Center, Campus and Branch office that will leverage OpenFlow and SDN, though product specifics were not included. The demo area offered some insight into part of what Dell's application strategy may be around SDN. Dell's SDN demonstration showed their 'Advanced Infrastructure Manager' orchestration platform with API integration into a OpenFlow controller from Big Switch Networks. The demonstration showed the orchestration platform automatically provisioning local and remote or hybrid cloud overlay network topologies over an OpenFlow network using Force10 switches.
If you have been following my posts on SDN, you may be thinking that it sure smells a lot like wine and roses; however as with any movement in technology, there will be bumps in the road that will need to be ironed out. Internet Research Group analyst Peter Christy gave an excellent presentation highlighting many of the very real challenges that OpenFlow and SDN will face to gain market adoption, which I think offered a much needed perspective to the conference. Christy presented the following slide:
After his presentation however, an audience member reflected on the critical problems that the network operators had shared throughout the day and posed a question to the entire panel of industry experts on stage: how else can we resolve the challenges described? I felt a bit of a stunned silence across the audience, and as optimistic as much of the audience was about SDN and OpenFlow, I felt this question provided something of an ah-ha moment. There were no answers offered. This did not lesson Christy's points in any way, he raised many of the same concerns I see among my enterprise customers every time I talk about SDN, which are basically that SDN is new, still needs to mature, and particularly for enterprise customers we still haven't seen many products hit the market (yet) that are addressing significant pain points.
One area I think Christy is rightly skeptical about was his concern about how fast commercial-grade technologies could be developed. My observation though after listening to Google, Facebook and Yahoo speak, was that each of these companies faced never-before-seen scaling challenges. Today Facebook for example has more traffic than the entire global internet had only a few years ago. Former Facebook VP Jonathan Heiliger noted that as Facebook faced massive growth, they were simply trying to keep the site afloat. But in this process something amazing happened, the web portals were able to not only meet the challenge, they made orders of magnitude improvements, defining new systems architectures that are redefining the way the entire world consumes technology. Christy was right to be skeptical, I am convinced that there is no technology manufacturer that could have come close to addressing the scaling and overhead challenges that the web portals have faced as effectively as they were able to. Historical wisdom would have said that if you have a performance or density problem, use specialized hardware. But it was through software that that the innovators in cloud architectures were able to exceed the limitations of specialized hardware in every measure of performance and scale. Now the industry is seeking to build upon this mantra, only time will tell how effectively technology manufacturers will be able to respond to these innovations.
Another important point that Christy made was observing that standards bodies have become notoriously political, and have significant challenges themselves. And there is a lot of truth to this. In his series on innovation, Harvard professor Clayton Christenson observed that proprietary tech fills an important role when new and immature disruptive technologies are first launched. During this period new technologies experience rapid innovation which may be better served being unhampered by efforts to gain political consensus. While I can't disagree with Dr. Christensen on this point, what he described here is a problem not unlike any other problem that mankind has solved. And while there is still much to be proven, Open Networking Foundation (ONF) Director Dan Pitt gave an excellent presentation that demonstrated that the ONF put a lot of thought into how they designed and structured the organization to incorporate lessons learned from older standards bodies, software communities and from the devops and open source movements. He noted that the ONF's charter would not allow technology vendors to serve on the board of directors, but rather it should be governed by the network operators who have to live with the results. Working group chairs are assigned by the board, and a system of checks and balances has been put into place to try to prevent the problems that some standards organizations have become notorious for. This is not an end to problems, but I believe the structure of the ONF and the promise of SDN will help lead the posture of the industry into a new era that will be better at standardization, open interfaces and hopefully take steps away from vendor lock-in, towards more open innovation that will benefit the industry and the larger community.