2011 has been one of the most exciting years for networking I can remember. Innovations made popular this year will prove to bring more value to enterprise networking consumers than anything in recent memory. Not only are we transitioning to cloud architectures and highly automated systems management, but new trends promise to do for the networking industry what x86 and the PC clone did for the compute market. In contrast to the slim margins of the x86 market, leading network vendors make as much as 70 to 80% margin. And even with these tremendous margins, innovation in the network industry has paled in comparison to the x86 market.
I say it's time for things to change.
It wasn't that long ago that we lived in the mainframe era. Computers were highly expensive resources that few had access to. Hardware, OS and in many cases applications were vertically integrated, IBM controlled the pace of innovation in the industry, and the saying of the day was "You don't get fired for buying IBM." Then Phoenix figured out a legal way to copy IBM's BIOS, Microsoft found a loophole in its contract with IBM, and the PC clone was born. As a result, today virtually all Americans have access to compute resources, and this proliferation of compute has changed all of our lives forever.
Today, the networking industry looks a lot like the mainframe industry did back then. Platforms are predominantly closed and completely vertically integrated. Hardware, OS and applications are defined and controlled by the vendor, and the lack of diversity in the market creates high barriers to entry for startups and innovators. OpenFlow was actually created because it was virtually impossible for researchers to do meaningful research and experimentation, to bring innovation in networking. The following slide borrowed form a great presentation from Stanford's Nick Mckeown summarizes the difficulty of innovating on top of traditional networking platforms:
I recently sat down with Nicira CTO Martin Casado who had some very interesting commentary about innovation in the networking industry:
"It really is about innovation because, if you couple layers, it's very difficult to innovate on one layer without dragging the rest, which normally will hold you back. So if you look at a server, you've got a very well defined interface, your Intel CPU, then you have an operating system on top of that, which you have very well defined interface on top of that for writing applications, and often applications have very well defined interfaces on top of them. But if you look at networking gear, the hardware is proprietary, the operating system is built by the same guy that built the hardware, and the applications that sit on top of there are written by the same guy. So you can't innovate one without modifying the other two; they're much more tightly coupled than you normally see. I think going to work with horizontally integrated model, where you've got open interfaces on every layer, it's not just good for competition , but I also think it's the right methodology for building a system, where you want to evolve each layer independently. "
I also spoke with Big Switch Networks co-founder Kyle Forster discussing the coming innovation in the industry, and he commented:
"They're trying to evolve the networking industry from the mainframe type of model to something that looks more like the service model. This is not a new concept. This has been tried before. But there are three things that are different. The first is the maturity of the set of technologies. For networking, that's the rise of broadly available networking silicon. Third party available silicon fundamentally changes the industry landscape. The emergence of OpenFlow as a technology standard changes things a bit. And these things are happening in an area where requirements for networks are actually changing ... We think it's really going to change the dynamics of a networking conversation. It allows the emergence for really new, really big technologies." The entire conversation can be viewed here.
One of the things Kyle mentioned is the rise of broadly available network silicon. This is highlighted by Intel's recent purchase of Fulcrum Micro, about which Analyst Rob Enderle reported:
"Intel has reiterated that it is fully capable of doing to high end switches what it did for high-end servers: turn the low and mid-range offerings into something far more affordable"
Perhaps the biggest news this year on merchant silicon was the announcement that Cisco's new Nexus 3000 series is based on merchant silicon. It is also rumored that Cisco's highly anticipated Jawbreaker platform will be based on merchant silicon. In terms of what this means to innovation, I have some friends that worked in Cisco's business unit while the 3000 series was under development, and they commented that the 3000 moved from inception to launch significantly faster than any products they had seen developed within Cisco. And given that they provide Cisco with their highest density, best performing and lowest cost per bandwidth devices in their entire portfolio, I'd say that merchant silicon is a winner. Network World editor Jim Duffy recently reported that several industry analysts including Jon Oltsik and Zeus Kerravala believe Cisco will continue to focus on merchant silicon. This begs the question, which strategy will Cisco move forward with; the proprietary Nuova asics of the Nexus 5k and 7k platforms, or the merchant silicon based Nexus 3000 line, which appears to have a different strategic focus from the older Nexus lines? My take is that it seems Cisco is trying to play on both sides of the fence rather than putting a stake in the ground. I think most manufacturers would ideally prefer a solution that gave them a lock-in advantage, which is why it is important for consumers to vote with their wallets and choose solutions built around an open ecosystem.
At the recent Open Networking Summit, nearly every network vendor including Cisco sang the praises of Software Defined Networking (SDN), noting that SDN methodologies have the potential to bring substantial innovations in networking while significantly reducing equipment costs and streamlining operations. If every major networking vendor, the largest network operators and service providers, and all of the top academics and researchers believe that SDN and open networking will bring tremendous benefits to consumers, maybe it's time to stop buying products built on proprietary asics to deploy proprietary features in a completely proprietary vision. Now is the time to invest in an open ecosystem to share in the multitude of benefits that an open, competitive and healthy market can bring to consumers.
It is interesting to consider what the Ethernet switching market may look like in a few years, and I had a recent experience that gave me a glimpse. While I was looking at the new Cisco 3064, I couldn't help but notice that it was a mirror image of several other products on the market. Here is what I mean:
I could keep going down the feature list to continue to stress the similarities of these platforms. My point here is that network forwarding hardware is already starting to look like the x86 market. And with most vendors supporting an "industry-standard" CLI, increasing commoditization of networking hardware seems inevitable. For this reason, I find that merchant silicon should be core to enterprise networking strategies moving forward.
Not everyone is as bullish as I am on this topic. As reported here, Rob Soderbery, senior vice president and general manager of Cisco's Unified Access business unit recently stated: "We have competitive ASIC development. We always evaluate a make/buy decision. ASIC development is a core part of our strategy", causing the blogs author to note that Cisco appears to be sending "mixed signals".
EtherealMind Blogger Greg Ferro weighed in on the topic with an excellent post stating:
"I have the view that Merchant Silicon will dominate eventually, and physical networking products will become commodities that differentiate by software features and accessories - not unlike the "Intel server" industry ... One interesting feature is that John Chambers continues to publicly state that custom silicon is their future. There are parallels with Sun Microsystems who continued to make their own processors in the face of an entire market shift, and that doesn't appear to have worked out very well."
As for OpenFlow, even with as new as it is, it is already starting to reveal the tremendous innovation that can occur in an open ecosystem.
BACKGROUND: Instant Expert: All you need to know about OpenFlow
One example is simple routing. After all these years, routers still have very little intelligence in terms of packet forwarding, which is their main function. They still forward based on primitive, static metrics. But with OpenFlow, Stanford researchers built highly intelligent dynamic load balancing, with only 500 lines of code, which they propose as a standard feature of a network operating system. Like Nikhil Handigol says in this presentation, "Load balancing is just smart routing ... our network itself should be able to do load balancing for us." Yet today, we still have to invest in expensive dedicated hardware, and highly complex configurations for something that should just be simple. The picture below from Handigol's demo shows the average response time for http requests from a pool of servers. The left side of the diagram shows the response with standard packet forwarding, and the right side of the graph shows the results of Handigol's simple load balancing code. No expensive dedicated appliances, just simple code proposed as a primitive feature of a network operating system.
Some may say that the current industry has done innovation perfectly well. I would argue the same could be said for the mainframe industry. There was no shortage of innovation. But when contrasted with the open x86 ecosystem, open innovation always wins. Networking has had no Moore's law and Hitachi's law driving innovation, but given the trends, I suspect we will.
While I am enthusiastic about current trends, the reality is that for them to have the effect that I hope for, these trends are reliant on the market and reliant on each of you. While I think the industry has suffered due to a lack of diversity, I don't blame the current market leader; the market allowed the current model to propagate. And once upon a time, there was good reason for that, but no longer. Many of us in networking don' t take different network architectures into consideration; we trust one vendor, and follow whatever they tell us is best. But the market is changing. There is an opportunity to bring innovation and differentiation to your business, and to your own career, if you leverage these market trends to your advantage. I think we are getting to the point where our decisions are much like they were with the PC clone ... some continued to pay the premiums of the leading vendor, others took advantage of the open ecosystem, and we all know who won in the end.
The opinions presented in this blog are those of the author and do not necessarily reflect those of his current or past employers.