Fichier 287


The "New Edge": When Artificial Intelligence meets Networks.

The "New Edge": When Artificial Intelligence meets Networks.

By Georges nahon, corporate vice president north america - February 16, 2017

Photo Georges Nahon

More powerful and sophisticated than the edge networking, the new edge will promote the rise of IoT in intelligent cooperation with the 5G network and the cloud.

Up until now, the concept of edge computing/networking has been described as a smart node at the very extreme end of a network with storage and processing power. Its goal is to reduce latency. In this model, the end points have often been considered as "terminals" or sensors with little to no intelligence. Decisions that need to be taken close to the place where the terminal or sensor is located require very fast response time and intelligent processing at the edge network node. But not at the sensor itself.  

Things have changed with the development of Artificial Intelligence and the related need for high performance parallel computing. Specific processors initially designed for graphics boards and video games machines have found a new life in enabling the current explosion of machine learning: they are the GPUs or graphic processor units. They are now appearing in many places, in datacenters, cars, drones, robots, smart IoT devices. Their price will continue to go down and their performance will increase. 

A new actor has thus appeared in networking and computing : this new edge, which captures and processes data, makes decisions and also connects with other "extremities" of the same environment such as other smart sensors or embarked systems of the same environment. It allows autonomous vehicles, cars, vans and trucks to connect between each other and to drones for example.

The advent of 5G in mobile communications will play a major role in helping to deal with the need for lower latency -- but anything that can be processed locally will be. And the sheer amount of data generated by the giganormous smart Internet of Things population will require intelligent cooperation between the processing in the smart edge devices and the cloud (AI) servers, all tied together with ultra-fast 5G networks.

There is so much power and storage now in these smart devices that they are resembling small datacenters themselves where real time becomes a critical component. We can imagine this, for example, when a self-driving car has to make a decision when confronted with a critical situation, such as avoiding collision with unexpected pedestrians, cars and a host of other unforeseen "obstacles" requiring sudden stops or turns. Drones need to make super-fast decisions based on image recognition and it will be happening locally at the drones themselves. Connections with the cloud will still need to happen to feed and improve the "centralized" learning system. Which in turn will feed back the aggregated data and learnings of smart devices at the new edge and propagate them back to the central brain in the cloud – at 5G gigabit speeds.

An unprecedented amount of real world data is being captured by radars, lidars, accelerometers, sensors, cameras, localization devices. "A self-driving car generates about 10GB" of data per mile, a lytro-camera generates 300GB of data a second" says Peter Levine of Andreessen Horowitz, a well-known Venture Capitalist firm in Silicon Valley." …'For the first time in computing history, we are collecting real world data about our environment whether that is vision, location, acceleration, temperature, gravity information". The role of networks is profound – to deliver these learnings at gigabit speeds. 

With IoT, people have been talking about 26 to 30 billion or so devices in 2020 and we have probably about 6 to 10 billion of those already in service. In terms of value of the IoT in 2020, we've seen numbers ranging from 1.9 $ trillion to 14.4 $ trillion, so we are looking at a much larger market than smartphones. It is exemplified in a study by Cisco, which mentioned that "globally, the data created by IoT devices will be 269 times higher than the amount of data being transmitted to data centers from end-user devices, and 49 times higher than total data center traffic by 2019." There is a number of technical and operational challenges when dealing with such a large population of smart IoT devices that still need to be solved.

There is another important difference between the new edge and the current one: and that is not just required bandwidth, but scale. The amount of devices involved is in fact several orders of magnitude greater than the already very large mobile phone (about 4.7 billion) and smartphones populations (about 2 billion). Power and speed are important, but scale is king.The new edge of network will play an important role in making the new age of IoT happen and more importantly, impactful.

Georges Nahon,
CEO,Orange Silicon Valley