ARTIFICIAL INTELLIGENCE WILL REVOLUTIONIZE WI-FI

I’ve never hidden my love for wireless networking.  Wi-Fi has moved from a nascent technology to one that is widely accepted and become so commonplace that we wonder how we ever functioned without it.

It started from autonomous access points and was followed up by controller based architecture (with a centralized controller and thin access points).  And, as we learnt from the challenges in deployment of Wi-Fi and the ability of environment to impact user experience, companies have constantly tried to innovate.  Some focused on building dynamic channel or power planning, some controller-less  and others tried to make it work in single channel (don’t deploy single channel until you have read the challenges here).

Of course, the discussion around evolution of Wi-Fi architectures is not complete without talking about the adaptive beam forming antenna technologies.  This technology automatically adjusts to changes in the Radio Frequency environment providing stronger signals to the client.  Standards of Wi-Fi have been evolving constantly also to enable great user experience (read about 802.11ax).

A lot of  development has also happened around improving computing real time location of Wi-Fi clients as well as business applications using the indoor location.  And, of course, with everything moving into cloud, Wi-Fi controllers or management moved into cloud too.

And, yet, even after years of evolution and innovation, the vendors avoid the conversations that are centered around the guarantee of quality of experience for wireless users.  Wireless networks today still requires hands and heads.  The effort and time that goes into its support is significant.  There is a need of wireless management autonomic system that “watches” the network, “understands” normal functioning, “analyzes” real-time performance against norms, then “acts” to automatically solve known problems.

The problem is that  the data source in wireless network is very big.  The data varies at every transmission level.  There is a “data rate” of each message transmitted.  There are “retries” for each message transmitted.  The reason for not able to “construct” the received message are specific for each message.  The manual classification and analysis of this data is infeasible and uneconomic.  And, hence, all data available by different vendors is plagued by averages.   And, this is where, I believe Artificial Intelligence has a role to play.

Only use of AI can change the center of focus from evolution of wireless or adding value to wireless network to automating ensuring the experience.

Deep neural nets can automate the analysis  and make it possible to analyze every trend of wireless.   Machine learning and algorithms can ensure the end user experience.  SDN will play a key role to enable the programmability of the Wi-Fi network, however, SDN is not the end goal.  The end goal of the programmability (or sdnization) of wireless network is still the enablement of  self managed wireless network.  A network that proactively ensures the end user’s mobility experience.

Of late, AI is graduating from science fiction to reality.  The cheap computer processing power has made it real.  I am convinced that by 2020, enterprise Wi-Fi will be significantly artificial intelligence based.

I invite wi-fi engineers to supplement their domain expertise with artificial intelligence tools so that we can free up systems administrators from sitting at a console or waiting for alerts.  That is the only way to drive down IT support and administration costs over time.  The fully autonomic systems that I envision will take on more of the characteristics of human intelligence for managing Wi-Fi.  Until we see that product, we can use this as a yardstick for measuring the evolution of Wi-Fi management products.

Tags: , , ,