When we started the company back in 2010, the vision was to use SDN controllers along with physical and hypervisor virtual switches to build a range of new networking products that the world had never seen before. Steve Jobs’ “Think Different” campaign was a mantra around our (very small) office at the time.
As I look back on the last few months, a few points make me particularly proud of the journey we’ve traveled.
First, Verizon’s recent announcement to use Big Switch software for the SDN switching fabric underpinning their massive scale NFV deployment is a validation that the open hardware plus SDN software approach is ready for prime time. As the single largest customer of networking equipment in the world today, the Verizon team gets to see every product on the market (and many not yet on the market). They are one of the most demanding engineering teams in the industry, and are known for exhaustive lab testing before any design goes to production. A video interview with the planning director leading the project is here.
However, the more subtle part of this announcement was the technical achievement -- an SDN switching fabric that can achieve both ‘chaos-monkey’ level resiliency and optimizations for the unique traffic patterns of NFV that are simply out of reach of traditional L2/L3 network designs. Compared to traditional switching, this is just different.
A second recent milestone that made me think back to our early days was last week’s Software Gone Wild podcast. Two of the key leaders in our technical staff - Sandip Shah and Andy Shaw - spent an hour on the show talking about how we use one of our SDN software loads with the same open networking hardware to build a fabric for chains of security services (nicknamed “Big Chain”). This isn’t re-purposed L2/L3 forwarding, but rather a new and unique approach to chaining, load balancing and health checking service chains. We are particularly proud of this Ivan Pepijnack “ah-ha” moment.
The last milestone I wanted to point out with this blog was the accelerating pace of innovation in monitoring fabrics (also known at small scale as Network Packet Brokers, or NPBs) that is going on right now due to SDN software and open networking hardware. Ten years ago, monitoring fabrics were basically groups of matrix switches that carried traffic from taps to various analysis tools. Over the last decade, they have integrated increasingly sophisticated traffic filtering and grooming functions. Going forward, our goal is to provide something different. Even a relatively small (4 switch) monitoring fabric has the hardware horsepower to aggregate traffic from hundreds of taps and more than a terabit per second of traffic. Are there analyses that are particularly well suited to this type of layout? We’ve launched the first few, and this is just the beginning.