Architecting a Cloud-to-Edge Computing Strategy
After a decade dominated by the growing reliance on centralized third-party public cloud services, the deployment of edge computing systems and Internet of Things (IoT) devices represents a partial pendulum swing back to decentralized data management. Large cloud service providers, however, are eager to extend services from the cloud to the edge. As a result, organizations face decisions on how tightly to tie evolving architecture to a single service provider.
Cloud computing services such as Amazon Web Services (AWS) and Microsoft Azure allowed large corporations to ease dependence on large, costly private data centers. They also enabled emerging companies to rapidly scale up to compete with larger competitors even if they lacked the time or financial resources to build their own computing infrastructure.
But as organizations seek to capitalize on a growing torrent of data from edge devices and servers, they can’t afford to be constrained by the latency and bandwidth limitations inherent in transmitting data up to a cloud service. In critical situations, operational control requires real-time monitoring and analytics capabilities.
“Now people realize that data should be close to where it’s absorbed from or consumed,” said David Linthicum, chief cloud strategy officer at Deloitte, the consulting and advisory services firm. “If you’re flying an airplane, you don’t want to send data down to a cloud server to see if the engines are on fire.”
Still, cloud resources are essential, for cost-effectively analyzing large data volumes over time that can feed models for implementation on devices operating at the edge—whether in a plane, controlling autonomous vehicles, re-calibrating factory machines, or deployed throughout a so-called smart city.
“The problem everybody is trying to solve is how to balance data between edge-based devices and cloud systems,” Linthicum said.
New York-based Oden Technologies developed an industrial automation and analytics platform that used a cloud-to-edge architecture to provide manufacturers with an AI-powered production recommendation system to optimize production and hit peak factory performance. An edge gateway server connects to the company’s Oden Cloud to enable complex data processing and real-time machine learning in edge applications.
“Our Golden Run platform uses a machine learning model that is periodically updated. Whenever there is enough new data, we run it to see if there might be a new optimization to implement,” said Oden co-founder and CEO Willem Sunblad. “We might do that a week at a time so it doesn’t have to run at the edge, but predictive quality should run in real time and for that you want latency to be low and don’t want the dependency of the internet.”
How to Navigate Cloud-to-Edge Models?
The Linux Foundation-supported State of the Edge 2020 report noted that infrastructure to support edge computing is “nascent” and enterprises may have to implement their own until the technology matures. “Until there are public ‘edge clouds’ and ubiquitous high-speed networks (e.g., 5G connectivity), custom edge installations may be the only way to reliably implement an edge application,” the report asserted. “However, as the demand for edge applications grow[s], the cloud will drift to the edge. Edge computing will become part of the standard internet topology.”