Is there a missing component in network evolution?

Posted on: 9th September 2015

In the latest Aria Networks White Paper – Evolution to a business driven hybrid network in the world of SDN, NFV and LSO – we argued that SDN and NFV are very much part of the solution to a network that can provide business value. Yet being “part of” implies there is still another critical component to take into consideration. That part is “Network Intelligence.”

In using the term Network Intelligence, I specifically mean Artificial Intelligence (AI). When networks finally have superior levels of intelligence embedded in them, to automate the role of a human network designer/operator to one that not only operates in a dynamic nature but also at the high speeds demanded of it, then a significant shift in network management can occur. It is probably important to note here that this does not mean that humans are removed from the equation; quite the opposite. A human is required after all to clearly define the objective to which a network, and its exposed services, must operate.

But even when looking at AI, the fundamental issue to consider is that the network’s requirements are not static, even over a period of months let alone hours or minutes so how can it be possible to bring evolving levels of intelligence in? What is required is an “evolutionary intelligence” – that can orchestrate networks in response to ever changing objectives.

Of course this represents a significant change in future thinking and planning and there will be many who will contest that this is a step too far but it is, in Aria Networks’ opinion, inevitable for these reasons:

  1. Value is increasingly becoming the services and not the transport. Therefore to remain competitive CAPEX (non-generic equipment) and OPEX (zero touch human intervention) need to be reduced.
  2. Feature and service velocity will become key competitive differentiators. If you can’t provide functions that in turn enable your customers to remain competitive they will go with an operator that can.
  3. An end user consumer or business server is seldom constrained to one network so contract mediation between network domains will also need to be automated.
  4. The “One Network” needs to be dynamic.

So, as networks can be, or will soon be, fully abstracted you can start to appreciate how the complications of introducing evolutionary computing techniques to SDN/NFV enabled networks will be significantly reduced.

Whilst, to some degree, this is a futuristic looking statement, it is also very much a concept that is not completely out of reach for today’s service providers. Aria Networks is already using evolutionary AI to do network planning in highly complex networks. In the future the most competitive networks will all be applying this intelligence.