Edge computing is everywhere these days, popping up as a regular fixture in IIoT-related events and IIoT vendors' press releases. But given the term’s seeming ubiquity, edge computing is often defined, vaguely, as a technique that takes advantage of data processing in the farthest reaches of a network, close to where that data originates.
To launch a successful edge computing project, however, requires a more substantial technical underpinning, as a recent Industrial Internet Consortium white paper explains. It requires identifying where exactly the edge is, what defines edge computing and what reasons should guide its implementations for IIoT applications.
Given the relative dearth of information on the subject increases the risk organizations deploy an edge computing architecture that is ill-suited for its targeted application. “Part of the reason that we put this white paper together is that we see people that are doing edge compute applications where the’ve stranded some sort of processing power out in the middle of nowhere, without any orchestration, sharing of data or figuring out of what kind of architecture they need,” said Todd Edmunds, senior solutions architect at Cisco who coauthored the IIC Introduction to Edge Computing in IIoT white paper.
The white paper is more focused on technical considerations than high-level executive matters concerning edge computing. It includes detailed information on edge architectures and use cases.
The document takes the position that edge computing represents a logical function-based approach to data processing and, as such, refrains from assigning a rigid definition of what an edge computing architecture looks like. Instead, there is a continuum of edge computing applications and the question of where the “edge” in edge computing is depends on an individual’s point of view and the specific business focus driving an edge computing project.
One end of the edge computing spectrum could be a constrained application where the edge is defined relative to a type of device. Temperature monitoring and control are an example of such an application. Edge devices in such an application could be thermocouples and a heating or cooling unit. The edge computing device would be the temperature controller responsible for running a related algorithm that makes adjustments to the temperature.
At the other end of the spectrum of edge computing architecture could be a string of globally distributed manufacturing plants, where a cloud-based analysis platform processes data related to the output of those facilities. “In that case, the plants themselves become the edge,” Edmunds said.
When considering such a potentially diverse set of applications, it is invaluable to have input from experts from a range of disciplines, said Lalit Canaran, vice president of emerging technologies and IoT at SAP. “We have a cross cutting advantage. When we want to talk about security, there's already this Industrial Internet Security Framework that has been published by IIC,” Canaran said. “We can draw on the diverse membership of the IIC, which spans networking, software and industry sources.” Individuals from a range of companies consulted on the guide, including ABB, Rockwell Automation, Huawei, Fraunhofer FOKUS, Moxa, PwC, SAS and sensor-maker SICK.
The coauthors listed on the paper include Edmunds and Canaran, as well as Mitch Tseng of Huawei. “I would say, we were the brain power only very early on,” Canaran said. “At some point, we became the guides and eventually we became the editors.”