Computing is quickly moving from massive data centers to a distributed model. While there has been some transformation of where the back-end data centers reside, new 5G technologies and emerging business and consumer application requirements enable the movement of computing power toward the "edge" and closer to the data generation location. The "edge" can include desktops, laptops, small servers, sensors, and even applications that don't run on a central hub. Today, what constitutes the "edge" itself is still being defined.
According to ResearchAndMarkets, the edge computing market is poised to grow by $7.29 billion during 2021-2025, progressing at a CAGR of 26.47% during the forecast period - demonstrating the next generation of innovation that is expected.
How it Works
One of the primary enablers of edge computing and this transformation process is the computing hardware itself. With newer systems being developed to require less space, use less power, and have more processing power than ever before, remote installations have made decisions close to where the data is generated.
The latest generation of CPUs performs more work per watt of energy than previous iterations of products, enabling increased decision-making at a constant power level. This advancement is in tandem with the addition of small form factor GPUs that fit into even smaller servers. This means that computing systems are smaller and more power efficient than ever before – and with enough performance to execute artificial intelligence (AI) algorithms in homes, light poles, stores, and neighborhoods.
It's no longer required that data be sent back to a distant centralized facility to be processed and computed. Instead, the computation needed to power business and consumer applications and use cases is now available at the very edge of the network.
Where it Works
There are several real-life applications where edge computing proves to be beneficial. Many building environments benefit as most consumer appliances are outfitted with numerous intelligent sensors. Home electronics now require a surprising amount of data from controlling the lights, HVAC, sound systems, and external security. Some require a small server to filter the incoming and outgoing data packets. Office buildings have even more intensive requirements with embedded systems that constantly monitor the surrounding environment, including temperature and humidity, lighting, airflow, power, and security systems.
Even factories can now monitor every aspect of manufacturing using AI to modify and improve processes without stopping assembly lines. Such factories, where numerous robots assemble products, may require low latency communication between servers and robots. Still, there may not be a need to send performance data from the factory to a centralized database on an ongoing basis. Having localized computing allows for the assembly line to adapt and improve at a faster pace.
As companies start betting on the augmented reality "metaverse," where real-world viewing is overlayed with additional information, powerful computing at the edge will be essential. These applications will need to respond within milliseconds to head or eye movements, or the viewer's experience might be tarnished with notable delays or incongruent overlays. The response of a headset movement and displaying the new data requires very low latencies, which is accomplished by having local computing power and the more recent data close to the viewer. Transmitting the headset movement to a cloud data center and waiting for the response with new images will have intolerable latencies that make it an unacceptable option.
Furthermore, communications can be drastically accelerated between machine-to-machine and connected IoT devices thanks to localized compute power. Data and connectivity speeds that previously required the use of cables and thus were not configurable at a moment's notice can now be accomplished wirelessly. As a result, 5G enables new performance levels of communication between machines and lowers latency through a wireless network.
What's Next
With a continuum of computing from the edge to the cloud, data can now be processed at each step. This leads to a hierarchy of decision-making, with some data handled at the edge and others sent back to a centralized data center to be processed.
In the home example mentioned earlier, a small local server can coordinate appliances and environmental conditions without escalating the computing. Most of the data generated are of little or no use beyond that single home. However, in a neighborhood of hundreds of homes, power spikes, instability, or security breaches become valuable to the wider network. For instance, the details of a home break-in could be transmitted to the neighborhood server and then communicated to those living close to the incident or currently present on the scene and the police. This continuum of computing infrastructure, with levels and decisions able to be made at various levels - such as a home, neighborhood, city, and regional network – makes the entire system more efficient and responsive at every level. In addition, it allows for more rapid computing for basic tasks at the edge, freeing up centralized systems to only handle decisions that need to be escalated.
As the 5G era continues to emerge and the growth of data-driven workloads across embedded applications increases, we envision edge computing will play a critical role in addressing key network bottlenecks and fuel seamless connectivity for next-generation applications.
