Living on the Edge, with Thales
It’s a big day today, at least in my universe – Thales just finished its acquisition of Gemalto. This post has been hovering over my head for months now, and my hands itching to pen it down. Internet of Things, Big Data, Artificial Intelligence and Cybersecurity just went up a notch for me today. But the thing that excites me the most about this day is that we (team Gemalto) have stepped into a world of mission-critical edge computing, like never before.
Edge computing is not a new thing. In fact, one could contest that it dates back to computing itself. But the term edge computing has surfaced ‘relatively’ recently. One feels that the emergence of IoT as one of the most hyped terms may have been an impetus for edge computing as a thing, even though neither can be called a proper subset of the other. A very common and non-pervasive example of edge computing is the use of CDN (content delivery networks), a set of distributed servers that are responsible for providing reliable and efficient Internet to billions of users worldwide. For simplicity’s sake, imagine a bunch of users in your office trying to access the same content from a website. A CDN would be tasked (among other things) with ensuring that this content is not downloaded dozens of times from its source.
The information exchange with BurgerJoint.com and its timely delivery is important for someone who is really hungry and, of course, for BurgerJoint’s business. But it can hardly be classified as mission-critical. Enter Internet of Things! The use of sensors and wireless connectivity combined with mobile computing has opened the world of computing at the edge, for both mission-/life-critical applications and otherwise. There’s hardly anything mission-critical about connected street lamps, for example. Connected (‘connected’, but not that ‘smart’) lamps are often cited as one of the easiest IoT investments that can be done by smart cities due to a relatively easily proof of ROI (how much energy you can save; how much cost can you reduce etc.). For most implementations, connected lamps require connectivity with a gateway device. The lamps could merely be responsible for sensing the light intensity, but it’s the gateway that would instruct a group of lamps in an area to get lit up. Together, the lamp and the gateway form the intelligent “edge”.
The lamps and the gateways are geographically dispersed, the distance ranging in kilometers. But they are geographically static and immobile. A similar example is a factory that uses connected robots/cobots. Where a lamp may use some cheaper long-range communication (e.g. LPWAN) with high latency, a factory could possibly use WiFi between multiple connected robots in the facility. The gateway device in this case could be something as simple as a WiFi router. But depending on how intelligent and autonomous the robots are, a smarter gateway device might be needed that has more compute power than a simple router. Functioning of these robots could often be in the mission-critical (defects can be expensive) or life-critical (if human operators are nearby) realm, so a lower latency communication like WiFi or Ethernet is mandatory. As opposed to connected lamps, the connected robots are likely to be in just a few meters’ range, but require much higher compute power and lower latency communication. In both cases though, the gateway and the connected devices are static.
A drone, depending on its application, is a rather complex edge application. When used for recreation, the light-weight and low-range versions pose little threat to human life. But a drone being used in a naval or military surveillance operation can be considered a mission- and life-critical application. Any autonomous or semi-autonomous vehicle already has a lot of real-time computing requirements to interact with the environment. But a surveillance drone may also be required to capture video footage and send back to a security operating center. So a shift from the factory application, such a drone would not only require low latency and high bandwidth, but would also be on the move. Clearly, a significant amount of compute power needs to be embedded in the drone itself, rather than relying on a gateway device. It’s important to remember that this still doesn’t eliminate the need for having a gateway or cloud infrastructure. It merely implies the need of more computing power right at the very edge to address the high demands of the use case.
Edge computing is merely an architectural paradigm, representing the processing (or computing) of data closer to its source.
Edge computing is merely an architectural paradigm, representing the processing (or computing) of data closer to its source. How close - depends on the need for lower latency or lower cost or merely an optimization of energy resources. It is a form of distributed computing, fed by the new impetus of Internet of Things. One of the most sort after applications of edge computing today is autonomous cars. Self-driving cars are actually even more complex systems than many aerial vehicles. The operating environment is a lot more uncertain with a much higher human life involvement. The intelligent edge in this case is heavily dependent on the device itself (the vehicle), but requires a standardized vehicle-to-infrastructure (V2I) communication mechanism that requires speaking to gateway-like fixed devices (smart traffic lights, road sensors etc.).
As more mission-critical and life-critical applications of edge computing emerge, more and more compute power will be pushed out closer to the source of the data. Subsequently, a lot of Big Data (due to a high volume and velocity of data) and machine learning processing will happen at the edge as well. It might sound like an antithesis of the cloud revolution…everything moving to decentralized servers. But it isn’t. Edge computing is not an alternative to the cloud; it complements the cloud. No wonder that the cloud giants AWS and Azure have already introduced AWS Greengrass and Azure IoT Edge respectively. A parallel movement that will keep edge computing in fad is 5G. In my view though, carriers/MNOs will only be augmenting edge computing, and not necessarily drive it. For example, while 5G might have some use in connected vehicles, it is never going to reach the reliability or low latency to move the compute from the vehicle to the cloud. Autonomous vehicles will always require on-board compute (including graphic and AI processing) in the vehicle that can make sense of the radio, imaging or LIDAR data to make instant decisions. Similarly, for cases like VR (virtual reality), the more likely bearer for the high bandwidth traffic is WiFi, since it’s not likely that people will be wearing VR goggles and walking on the streets. Gaming itself though, is likely to be a huge category of edge computing in the consumer world. It already is with so many game consoles! This illustration from the McKinsey report lists some of the sectors that are going to lead the edge computing revolution. The illustration below is from the same report.
I’ve written in the past, and I’ll say it again, the coming decade is about building experiences. We already have a lot of technology. Now it’s about putting the pieces together to deliver seamless experiences that make the use of technology frictionless. This, in a world that is marred by a huge trust deficit in the technology realm. This is where the world of cybersecurity merges with the world of IoT, edge computing, big data and artificial intelligence – delivering digital security by design. The huge challenge that follows is to securely integrate the physical world with the digital to deliver new experiences; like providing safe and reliable transportation through autonomous trains; delivering entertainment and connectivity to passengers tens of thousands of feet in the air; building nano-neurons that could revolutionize disparate fields; or pushing the edge way out into the space by developing autonomous rovers for a mission to Mars – all in a day’s work at Thales. It is the time for new experiences – the time for living on the edge, with Thales.
(Couple of slides from Thales's Capital Market Day presentation from last year)
DISCLAIMER: All the cool views presented in this post are my own, and do not necessarily reflect the views of my past or present employers.