Edge computing is not a fundamentally new concept and, in some ways, it’s “Back to the Future.” In the 1960s and 70s, business computing was characterized by mammoth, centralized mainframes. With the advent of the personal computer, we saw a move to distributed computing from 1980–2000 when workers utilized PCs to complete their basic and complex tasks. In the past decade, computing and data storage has increasingly moved to the cloud, making centralized computing “hot” again. But with a sharp increase in data generation and a need for faster processing at or close to the source, like with autonomous vehicles (AVs), we are seeing a shift back to periods where computing was done closer to the users and consumers.
To best understand edge computing it is important to differentiate between the device edge and the cloud edge. Device edge computing runs software on hardware that is local and owned by the end user. This takes place in instances like calculating a tip at a restaurant on the iPhone’s calculator app or a sensor on a train car calculating the average speed. Meanwhile, the cloud edge, sometimes referred to as cloudlets, fog nodes, or micro datacenters, sits on the periphery of the cloud, closer to the end user. For example, data accessed frequently by a device like a user’s favorite movie, can sit in a cloudlet, and push the data to the device far quicker than it would take to fetch it from a huge data center. In the case of video, edge computing could improve the viewing experience for the user and mitigate buffering. One might assume that edge computing will replace the cloud, but they work in tandem — edge computing simply takes certain tasks away from large network servers and moves them closer to the user.
“Why now?” you might ask. The answer is that in order to efficiently operate devices like drones, AVs, connected machines, and smart homes, we need to develop faster, more efficient computing systems. All of these innovative devices interact with their environments by taking in data from the external world, analyzing them, and then creating suitable responses. Many of these devices also have capabilities not seen on 1990s-era PCs. They include neural net accelerators, artificial intelligence, and machine learning — components that allow these devices to “see” and “think” on their own. Edge computing is needed in order to perform tasks that demand a real-time reaction time and cannot tolerate a possibility of a network system outage or high latency responses.
In the case of an autonomous vehicle, various sensors on the car measure temperature, speed, traction, etc., and communicate the information to the onboard computing system. This system analyzes the information and directs the engine, wheels, and brakes in an appropriate manner. In an emergency situation, for example if a deer were to hop in front of a vehicle, the system would need to respond immediately. There is no time for sensors to send data to the cloud and await a response. If Amazon can experience glitches during Prime Day, we can’t rely on the cloud for life and death situations.
Another advantage of pairing edge computing with cloud computing is that it can lead to more efficient storage. In the case of autonomous vehicles, unnecessary, repetitive data is pruned away, and the car sends only the most vital information to the cloud for learning and analytics purposes. After data from thousands of vehicles are analyzed, updates are then pushed from the cloud to the vehicles — creating a virtuous cycle. Not only is this process efficient for the device at the edge, but it prevents the cloud from storing an inordinate amount of superfluous data. This is important because although data centers can house massive amounts of data, that storage is energy and capital intensive and the communication of this data can clog bandwidth. And with the research group IDC predicting nearly a ten-fold increase in annual data generation by 2025 — to 163 zettabytes per year — it’s important that only essential vital information is stored in the cloud for long periods of time.
Going forward we will see more and more applications of edge computing, particularly in healthcare, virtual and augmented reality, drones, AVs, smart cities, and the remote monitoring of oil and gas. At Fusion Fund, we’re seeing more and more start-ups embrace edge computing. We firmly believe that advances in edge computing will drive the some of the most innovative consumer and industry products and applications going forward.