Author: By Jim White, CTO of IOTech
This past year, we’ve seen edge computing emerge from pilot programs to deployments. We believe 2022 will be the year that edge computing is fully integrated into the architecture of every major industrial IoT system.
Prediction 1: There will be pervasive adoption of AI/ML at the edge. The new status quo: edge systems will incorporate AI and machine learning. Simple rules engines and edge analytics are already at the edge. Today, organizations demand more intelligence there. The raw compute to run AI/ ML at the edge was a prohibiting factor, but this is no longer the case. While training ML systems will occur largely in the cloud or in the enterprise, ML models running on lighter AI runtime engines at the edge are more common place and will soon be the norm. Visual inference has been a leading use case, but other AI/ML solutions are soon to follow. Edge platform providers will play a key role in developing solutions that can easily integrate AI/ML technologies.
Prediction 2: Hybrid edge/ cloud architectures will be the norm. It’s not edge compute “or” cloud compute, it’s a case of “and.” Organizations are finding that processing edge data needs to be performed at the edge and in the cloud or enterprise. Although initially there was much excitement related to the cloud providers reaching down to the edge, the reality is that there are significant challenges in moving all edge data to the cloud and performing all the processing in the cloud. The cost of data transport, latency issues and security/data privacy concerns are among the chief challenges. Likewise, the raw processing power of the edge and the ability to do deeper exploration of the edge data over longer periods of time for better insights means edge computing alone is not a solution. Solutions must allow for the right processing at the right levels, and this calls for hybrid edgecloud architectures.
Prediction 3: The industrial sector emerges from edge/IoT research mode. The industrial sector is becoming focused and organized in its effort to offer new solutions at the IoT edge. Businesses in manufacturing, building automation, and smart energy are in full “build” or “buy” mode for IoT edge solutions. Many large industrial-sector businesses are fully committing to grow their edge/IoT products and strategy. Buy mode leads when companies need to accelerate digital transformation (see demand solutions rather than pieces/parts below).
Prediction 4: Customers will demand solutions rather than pieces/parts. Companies looking to benefit from edge/IoT technology are looking for more fully integrated solutions. They want immediate tangible business outcomes and are not interested in receiving a bucket of technology parts that they have to pull together themselves. For system integrators, it means developing the right technology partnerships to pre-assemble and deliver complete solutions to customers. Integrators will naturally gravitate to edge products that are inherently more open and flexible, as these will be easier to integrate and adapt to more use cases.
Prediction 5: Realization that K8s is not enough edge management. Organizations deploying and orchestrating IoT/edge applications are discovering that Kubernetes is not always a fit in resource-constrained edge environments. Furthermore, K8s only addresses part of the edge management need. There is more to edge management than managing/monitoring containers; an edge-management solution must deal with preparing, managing and monitoring the host edge nodes, allow for rapid configuration changes, and even assist in sensor/ device onboarding. K8s will be part of some edge-management solutions where more resources or smaller K8s solutions (K3s as an example) can be applied. However, the thin, resource-constrained, network-constrained, latency-concerned, sometimes non-containerized, OT-device-touching environments demand alternative and more complete edge management solutions.
Please note the button below links to the Smart Industry website.