Sify-Banner1
V Shekhar Avasthy
Chief Data Scientist and Principal Consultant
FactsnData

Next-Generation Networks: Intelligence 2.0

Reliance of enterprise on information technology increased over last three decades, leading to a higher reliance on networks. The shift from standalone desktops to client–server to cloud-based computing has increased this reliance manifolds. And just like traffic on roads can be managed intelligently, so can the network-traffic be managed well with proper analytics. Many vendors offer analytics tools with proprietary algorithms and varying degrees of complexity, some even claiming their tools to offer big data analytics, or even artificial intelligence (AI) and machine learning capabilities! So, has network analytics truly evolved? The answer is a clear NO. A deeper dive and running some simple experiments with synthetic data on networks with such claimed capabilities shall prove that such claims are very often an intelligent marketing tactic of packaging a simple rule-based engine as an analytics or AI engine!

The network analytics of date can at best be classified as sophisticated network intelligence with some degree of self-repair capabilities. What this means is that, as many network administrators complain, the analytics module by itself consumes so much of resources and offers so little that it becomes a worthless exercise to utilize them.

So, what does the womb of time hold for the networks of future?

The answer is not very straight forward, but such networks shall have some of the following capabilities:

  • True AI. AI of today is often over-hyped and elementary rule-based modules are often passed off as AI networks. One school of thought defines AI as the capability of taking a decision in a completely unknown situation based on past experiences. Using that definition, the networks of future shall have the capability of managing unknown (or hitherto unseen) situations.
  • True advanced analytics. Network analytics of today, like analytics in most other fields, is limited to, at best, predicting. In reality, just like new technologies such as stimulus-response modelling are changing the landscape of text mining, by Making Future Happen, TRUE advanced network analytics shall not only have the capability of predicting but also attaining the network-efficiency goals by taking decisions with little human interventions. While some vendors may claim their tools to possess such capabilities already, reality is that any tool of today has very limited or superficial capabilities toward that end. For example, capabilities such as A/B testing, RCA (root cause analysis), emergency traffic prioritization are either inexistent or at nascent stages as on date. When it comes to network reports of very large scale networks, the resource requirement for report generation defeats the basic objective as in most cases, reports are still not generated using sample data but on complete datasets!
  • Recommender engine with scenario building capabilities. Similarly, most network analytics tools of date do not have capability of giving recommendations whenever a new network (or part of it) is designed or restructured. Using past data, a good recommender engine can rely on best practices library, historical data etc. to throw useful recommendations to the designers. Such recommendations shall be both, pro-active (meaning that admin shall be prompted automatically to consider some suggestions based on network behavior) as well as reactive (implying that when the admin is designing network, or asks the engine about a proposed scenario, recommendations pop up).

Well, this is just tip of the iceberg – networks of future will have many more capabilities.

Share this:

Stay Updated on Enterprise Network and Carriers Industry.
Receive our Daily Newsletter.