From my early days in 2008 at Stanford University’s Clean Slate Lab to today’s seemingly unending announcements that big name networking vendors are dramatically changing their stripes (Dell allows choice of NOS on their hardware, HP has an open source NOS project, Juniper is selling JunOS for white boxes), one thing is clear: the traditional networking model of closed, vertically integrated devices with “box-by-box” management is going obsolete.

To help demonstrate where this trend has been and where it’s going, I’ve put together a timeline (below) of the significant events and announcements from the early days to now.  For context, in the late 2000’s, hyperscale data center companies, like Google, Amazon, Facebook, and Microsoft, were finding significant cost savings by designing their own servers in partnership with “white box” server companies from Taiwan.  So, similar to the servers, it was only logical that they started designing their own “white box” networking gear.  

In tandem, Stanford University created the Clean Slate Lab to evangelize and support the growing movement around SDN.  While many innovations came out of that group, including the first OpenFlow specifications,  in retrospect one of the most interesting was the Indigo project - the first completely free disaggregated NOS.  Started by Dan Talayco (early Big Switch Networks employee), Indigo was a modest open source, free to download Network Operating System (NOS) and OpenFlow agent for white box switches.  Indigo is in many ways an early precursor to the Open Network Linux project that Big Switch leads in Open Compute today.  Between the need from hyperscale data center companies and the software from the Clean Slate Lab (Indigo, SDN, etc.), the beginnings of the open networking movement were formed and the results of which are still being played out today.

Looking into the future, I claim we’re going to see much more disruption in this space.  The remaining traditional network vendors (Arista, Brocade, Cisco, Huawei, etc.) will be forced by the market to disaggregate, selling their operating systems separate from their hardware.  Correspondingly, as the barriers to entry dissolve (e.g., via Open Network Linux), more players will compete in the traditional NOS space, to the point where traditional vendors will likely give their software away for free to keep their hardware business.  Open source software will become a dominant theme for networking; I’m tracking an explosion of new open source routing protocol implementations (Quagga, GoBGP, ExaBGP, BGP4J, etc.) and it’s only a matter of time until these stacks mature and the relative value of a high-quality routing stack is marginalized.  The compounding trends of OCP-style open hardware (e.g., Facebook’s Wedge and 6Pack) and merchant silicon will further reshuffle the value chain in networking, forcing traditional networking vendors to reinvent their business strategies.

People often ask me what these announcements and movements mean for Big Switch as a company and the answer is simple: this is 100% validation of the bets we made in 2013.  This trend of network disaggregation, automation, and hardware commoditization is one we have been fostering for longer than anyone and around which have built our entire company.  By design, our product’s value proposition in not about building a better network operating system (NOS), but building better control, automation, and visualization systems between switches, routers, analytics tools, virtualized orchestration systems, and NFV-like appliances to simplify how networks are operated, deployed, and debugged.  In other words, while we track with interest the latest announcements in the NOS-layer (and contribute to it with Open Network Linux), it does not affect our core value because, at least architecturally speaking, we’re above it.

Open Source Networking Timeline:

For more SDN history, please check out : http://go.bigswitch.com/rs/974-WXR-561/images/The%20Path%20To%20SDN%20Final.pdf

 
Rob Sherwood,
CTO, Big Switch Networks