You see, wire telegraph is like a very, very long cat. You pull on its tail in New York and it meows in Los Angeles. Do you understand this? And radio operates exactly the same way: you send signals here, they receive them there. However, there is no cat. – Albert Einstein
I read that quote the other day and it got me thinking about how you never really understand something unless you can explain it simply. We have so many technology choices and specification efforts we are working on these days, it is hard to keep track at times. Still, we are continually discussing them, trying to find the most effective ways to explain them to our peers both inside our respective companies and in the industry.
At times, we slip into our comfy place and drop into techno-babble (no real babelfish actually exist outside the confines of Hitchhikers Guide to the Galaxy). But we need to learn how to speak about technology in plain, simple terms. It is quite challenging at times.
One current concept I am working on is related to how we will scale network components to the point where we need them for potential future requirements. We have been very successful at making things work, as they are now based on technologies that have served us well for many years. One challenge I see is that we are trying to apply these same concepts to future needs, but we recognize that we are dealing with a complex network ecosystem today.
Yet, as that is our comfy place, we enjoy sitting in the plush chair of technology we built for ourselves and attempt to use those same technologies to construct the network of the future. My concern is we are challenged using these technologies today and yet we insist on using them to build new gadgets. My belief is we cannot solve today's challenges using the same ideas that created them in the first place. We need a new way of seeing things.
As I thought about the problems of scale, I realized that the way to meet future growth needs is to think about new ways of viewing things. We definitely will use soon-to-be released products to grow portions of our network. However, we cannot use them to solve all of the predicted growth needs in all of our networks.
The more I think about this, the more I realize that the most efficient and effective way to scale network components is to decompose those devices into their functional components. I divide devices into two separate functions; autonomous and cognitive.
Autonomous functions are those that require no outside control, e.g. moving data from interface A to interface B. Cognitive functions are those that require a conscious intellectual activity, e.g. route computation or scheduling. Those autonomous functions work well with custom ASICs to optimize their performance. Cognitive functions can be decomposed into software, which allows them to be easily scaled using virtual machine technologies.
Stay tuned for my next blog for more discussion about how breaking network components into functional blocks can let us achieve the scale I am talking about while simplifying the network…
— Jeff Finkelstein, Executive Director of Strategic Architecture, Cox Communications