Fields of Fiber

    Is SDN the Ethernet of the 80’s or ATM of the 90’s?

    Posted by Bill Miller on Feb 29, 2016 1:00:00 PM
    Author Image

    Everyone knows what Ethernet is, but how many people who started their careers in the past fifteen years have heard of ATM? No, not Automatic Teller Machine – Asynchronous Transfer Mode, a technology that dominated the hype cycle in the 90’s and then disappeared from public awareness, despite the fact that many applications still make use of it today.

    I have had the pleasure of being deeply involved in the industry through many changes. In 1980, I was in the field with Digital Equipment Corporation when the Ethernet DIX standard was launched to market. Later, I became a product manager of Ethernet at Prime Computer. In the 90’s, I was heavily involved in ATM (Asynchronous Transfer Mode) with General DataComm, heading up ATM product management and marketing. Moving on, I led cloud operations for a VoIP provider. Most recently I became deeply immersed in the SDN (Software Defined Networking) world with Fiber Mountain.

    The way I see it, the past four decades were dominated by four main networking technologies:


    The 1980's: Ethernet - The 1990's: ATM - The 2000's: Cloud - The 2010's: SDN

    I lived the hype cycles and tornados of these technologies, constantly facing new challenges and having fun doing it! Today, I am thoroughly enjoying the emerging SDN scenarios as people debate the definition of the technology.

    What does ATM have to do with SDN?

    In a recent debate with an industry expert who is a very close friend of mine, we took opposite sides on the future of the “SDN” acronym – will it survive, as did Ethernet, or will it be diminished like “ATM,” which became invisible as it was consumed by the application? ATM technology is still in use as legacy today, but no one sells it as “ATM.” It is an invisible technology that no one talks about. Ethernet, however, became the foundational technology that survived and is prominent in our homes and business networks. 

    The Software Undefined Network

    One argument that the term “SDN” could disappear is the fact that it is more conceptual than specific. This video from Moving Packets is a perfect illustration – ask a room full of industry experts what SDN is, and you will get at least as many definitions as there are people answering. Everyone may be talking about SDN these days, but despite that fact, few people agree on what the technology actually is and does!

    Both Ethernet and ATM were specifications-based rather than a largely undefined concept, which limits their usefulness in predicting the future of “SDN.” “Cloud” may provide a better comparison, but it can also be debated whether that term will survive, or eventually be consumed by the explosion of everything-as-a-service offerings. Both Cloud and SDN have similar challenges in that they are conceptual, and both have multiple technologies under their hoods. Ethernet and ATM are technologies that have been driven and narrowly defined by standards bodies, while Cloud and SDN each have multiple special interest groups, open source platforms and open compute platforms –which is not a bad thing, since innovation appears to be at its highest levels in history. 

    Will the term “SDN” endure? Personally, I am a firm believer that it will be around for some time, as new applications emerge and new network architectures challenge the leaders and incumbents. Innovation will continue, expanding the concept of SDN to include software controlled physical layer solutions such as Fiber Mountain’s Glass Core, which are transforming network infrastructures.

    Looking Forward

    SDN today is whatever each vendor is delivering. Industry analysts talk about “SDN capable” products vs. native SDN products. Anyone with VMWare’s NSX may think they are SDNs. Cisco ACI users say they are SDNs. Fiber Mountain’s multiple-award-winning Glass Core leverages SDN to separate the data plane from the management plane for layer 1 as well as layers 2 and 3, enabling programmability and dynamic management of both network traffic and the actual physical connections.

    Every week or two, I see another article debating the definition of SDN. The recent “SDN? I still don’t know what it is…” piece reminds me of ISDN’s “I still don’t know” rants. (ISDN, or Integrated Services Digital Network, was the circuit switched precursor to ATM). SDN, however, is being adopted for many reasons including automation and efficiency as well as programmability. 

    The conversation is in constant—and rapid—development, and part of the fun is participating as it unfolds.

    Here’s food for thought: will the real SDN please stand up?

    What do you think?

    Learn More about the Glass Core

    Topics: Glass Core, Network Architecture, SDN