The Innovator’s Dilemma and NFV
If you haven’t yet read the Innovator’s Dilemma, you should. This seminal book by Clayton Christensen is now approaching its 20th anniversary and is as relevant as ever. It analyzes the hair pulling difficulties that incumbents face in embracing disruptive technologies that can reshape industries and shift the balance of power between newcomers and themselves.
Microsoft is central to three classic examples. They grabbed the PC software market from IBM, very nearly ceded the Internet browser market to Netscape, and then by ignoring search engines until it was too late, ignominiously lost the Internet advertising market to Google.
The definitive telecommunications proof point is Cisco, who under the noses of Bell Labs and a host of mammoth telecom equipment companies, emerged as the dominant IP router supplier worldwide. I recall in the mid-90’s sitting at an industry dinner with the CEO of a major equipment company who remarked that he expected that IP routing would be a passing phase because the applications of statistical communications were highly limited.
Two Major Dilemmas That Exist Today
Industry incumbents in fact face two dilemmas. The first is that they are taught to listen to their customers, and what customers almost always prioritize are current solutions, only faster and cheaper. Moreover, customers rarely seek new approaches from their incumbent suppliers.
The second dilemma is that in their initial stages, disruptive technologies have performance limitations that severely restricts their scope of applications. They are not sufficiently refined or have the economies of scale to compete against established technologies. So they aim either at niches, or where “good enough” plus “cheap enough” are the major acceptance criteria. This makes it very difficult for incumbents to invest in these technologies because it diverts money from proven technologies that address larger mainstream markets.
However, and this is the hair pulling part, over time successful disruptive technologies steadily improve their performance (it took nine years to evolve from childlike MS-DOS to the highly acclaimed Windows 3.0), while simultaneously widening their spheres of application and gaining economies of scale. Before you know it the newcomers have their foot firmly in the door and it is too late for incumbents to react. This is sort of like placing a lobster in a pot and slowly bringing the water to boil.
NFV Is a Classic Case In The Making
This brings us to network function virtualization. NFV – which migrates packet networking functions from dedicated hardware that rely on customized processors, to run as software on commercial off-the-shelf (COTS) servers – has all the ingredients to be a classic innovator’s dilemma case study in the making.
Network equipment makers are facing tough decisions. For example, if you manufacture high end packet processing equipment today using specialized hardware, when do you start investing to cannibalize your own equipment? Or if you are a transport equipment provider, and you want to take advantage of NFV to compete in new markets, perhaps try bundling functionality in new ways. What is your optimal functionality rollout?
Invest too much, too soon, or in the wrong areas, and you may find that you are diverting critical resources away from your current core markets. Invest too little, too late, or in the wrong areas, and you can find yourself locked out. Many suppliers are now facing their particular NFV investor’s dilemma. The only certainty is that disruption creates winners and losers, and that the market for networking and packet processing equipment will look quite different in a few years.
Complex and Existentialist Decisions
Indeed, there are a lot of packet processing functions that can be “NVFized”: address translation, packer forwarding, switching, broadband gateways, VPN gateways, authentication, firewalls, deep packet inspection, DDOS mitigation, load balancing, SLA monitoring, mobile NodeB processing, caching, and WAN optimization – to name some important ones. The promised benefits are huge including lower costs, easy updates, and an ability to mix and match multiple functions on the same platform.
But where do you start and how quickly do you invest? As things stand today, price-performance still favors dedicated equipment and appliances for most applications. While NFV has started to implement certain functions, like customer premises equipment that delivers basic routers and firewalls, it doesn’t yet cut it for processing-intensive functions. For example, to implement core routers, enterprise intrusion protection systems, or deep packet inspection, it can take dozens of COTS servers to fill the shoes of a single piece of specialized hardware.
Yet, and this is the heart of the innovator’s dilemma, over time COTS platforms will become more powerful (sometimes aided by accelerators) inexorably enabling NFV to address broader swathes of complex packet processing functionality. So if you are in the packet processing business, or want to use NFV as a point of entry, you have some serious thinking to do.