NLP for IBN (intent-based networking)

If you have proficiency in the English language, chances are you understand the following phrases as being fundamentally the same, even if you would tend to use one over the rest:

  • There are none
  • There, none are.
  • None are there.
  • None there are.

Some of the phrases are clunky, but nonetheless, you understand the meaning of each, you understand, for lack of a better word, the INTENT (though we might tend to normally associate the word ‘intent’ with a work request or an action).

We live in a time when many different technologies are converging, and many different technology transitions are occurring: the shift to cloud computing, the emergence of SD-WAN, the use of machine learning in security, microservices, and an increased focus on automation, analytics, and operational efficiency being just a few. This makes crystal balling the future, challenging, at best.

One of the research areas in the AI/ML universe that has made real progress is natural language processing (NLP). I am not saying it is perfect, but it does appear to be the kind of application amenable to improvement by current technology and techniques. With the rise of Amazon Alexa, Apple Siri, Google Home, and other interfaces, it is just a matter of time before we are all talking to our networks in natural language(s), and what a great thing that will be when the natural language processing in our network software interfaces can understand our INTENT – doesn’t that seem like a good fit for a time in the evolution of networking where there is so much conversation about INTENT based networking (IBN)**?

I can’t predict with any precision what will be the year of NLP in networking, but the fact that there will be one, seems likely. If it is more than five years off I would be very surprised. There is a learning curve, and trade offs for sure. For example Amazon Alexa, in its developer toolkit, makes an explicit and well understood trade off in favor of higher precision, at the cost of more definition work. Maybe that kind of trade off will not be needed in the future.

Because the age of interfacing with networks via NLP seems probable, the only question that remains is what will the protocol be between centralized components and distributed components, as well as between the distributed components themselves? Will the industry stick to tried and true approaches of either one abstraction that aims to work for all vendors or one industry-standard format that every vendor has to snap to? Or instead, will we enter an age of flexible sentence creation using subjects, predicates, objects, and complements; a flexible age where intelligent natural language processors can create, and understand intent, as required? I hope it is the latter, because I believe that is ultimately a much more powerful paradigm, but it is not a trivial change, so it is hard to know the timing, though every networking vendor, IMO, should have an opinion on what that timing is.

Footnotes:

** While the strict focus of intent-based networking is a shift from imperative statements to declarative statements, the current industry conversation tends to expand IBN to include AI, ML, intelligent-decision making, and the ability to retain focus on, and achieve, a desired intent, in the face of changes in the network, which is a significantly different operational model than that of an imperative statement driven network, which is one of the reasons there is potential for it be a powerful paradigm.

 

This site uses Akismet to reduce spam. Learn how your comment data is processed.