Edge and Cloud in Smart Manufacturing: Strategic Guidance for Optimizing the Balance

Much like how the pendulum that swung from local production to outsourcing seems to have started to turn back slightly, the pendulum that swung from centralized computing with the original computers to distributed with Personal Computers (PCs) back to centralized with cloud seems to have swung back to distributed computing with edge intelligence and fog. This, of course, does not mean that edge intelligence will replace the cloud, but rather, it will augment it. Balancing edge and cloud computing holds the potential to fulfill the business case promised by the fourth industrial revolution and to optimize smart manufacturing solutions.

Cloud computing gained momentum in the first place because it provides greater value, more computing power for a lower cost than entirely on-premise computing. Edge computing further lowers the Total Cost of Ownership (TCO) by normalizing data for app integration and workload consolidation onto gateways or edge servers that can process streams of data from multiple devices from multiple manufacturers. This can ultimately increase the capacity for data processing, while decreasing the volume of data published to the cloud, further lowering costs.

Smart manufacturing vendors with a cloud focus have already started investing in edge solutions. Amazon announced AWS Greengrass back in November 2016, SAP launched its Edge Services in July 2017, Siemens introduced its Industrial Edge solution in April 2018, and Microsoft made Azure IoT Edge generally available in June 2018. These rollouts, like the cloud platforms before them, moved slowly at first in the manufacturing sector. Unlike the cloud, which came as a push from the vendors, the end users will pull for edge solutions, which feel more comfortable to many as they more closely resemble older, on-premises solutions.

The above solutions and others from companies like FogHorn Systems, Telit, Eurotech, ADLINK, Dell, HPE, PTC, and Altizon can include several different tools, each of which adds separate value:

• Protocol Translation: This involves the ability to extract data from multiple proprietary protocols and deliver them to common apps or software solutions.
• Edge Intelligence: An intelligent edge provides Complex Event Processing (CEP), edge-to-cloud closed-loop machine learning capabilities, stream processing, and actionable analytics on-premise, and it runs on industrial devices with small compute footprints and a wide range of Operating Systems (OSs).
• Hyper-Converged Edge Systems: These systems bring together software-defined storage in Virtual Storage Area Networks (VSANs) and networking with virtualized computing over a hypervisor on edge gateways or servers. A hypervisor creates and runs Virtual Machines (VMs) to execute multiple OSs on the same physical gateway or edge server with a single administrative platform.
• Edge Versions of Cloud-Based Platforms: To meet the needs of managers on the factory floor and deliver low-latency use case-specific actionable analytics for immediate, real-time decisions, cloud-based platform providers have developed industrial edge versions of their platforms to run over gateways and edge servers. The edge versions of the platforms make it far easier for enterprises to sync up their apps in the cloud with real-world operations and equipment on the factory floor.

To flexibly adapt to clients’ needs and help them scale, smart manufacturing vendors must:

• Implement environments where apps can deliver immediate results at the edge with stream processing. Because stream processing at the edge can catch more anomalies in data than batch processing with its blind spots and jitters, it proves more effective at predictive maintenance, quality control, and continuous improvement. Not all use cases require the low latency delivered by edge intelligence, but stream processing with apps that run at the edge can lead to earlier failure predictions and defect detection, and it will help find the root causes of slow-downs or malfunctions.
• Offer code-free app development. Code-free app development will significantly speed up deployment and scalability of customer apps. Customized code comes with an immense cost for the end user, often requiring full-time data scientists and developers. A code-free environment encourages innovation from more people, eliminates many of the costs, shortens new project timelines, and makes it easier for vendors to onboard new customers to the edge platform.
• Give the end user as much control as possible over what data will be published to the cloud. Almost all factory floors have at least some unique requirements, and enterprises need freedom in their Information Technology (IT) strategy. Edge solutions should empower the end user to decide what data are published to the cloud for Artificial Intelligence (AI) model training and further analytics and what stays at the edge for the models to evaluate and sync the results later.
• Promote your solutions’ abilities to extract data on the factory floor, specifically how your solutions merge the data flows coming from machines with different manufacturers. Before manufacturers can send any data to the cloud or execute any AI models, they must figure out how to extract the data from their machines and systems. Some have already implemented standardized architectures, such as OPC Unified Architecture (UA), or use only one protocol from one manufacturer, but most use a collection of different protocols from different manufacturers and have no pre-existing networking. Smart manufacturing vendors must adapt to and leverage existing infrastructure. This often means providing protocol translation, ingestion, and normalization capabilities. All vendors should promote any capabilities they have along these lines because their clients will struggle to network machines otherwise. With them, solutions will scale far easier.
• Consolidate workloads at the edge. Off-premise cloud implementations cost significantly less than on-premise, but smart manufacturing vendors can help lower their clients’ TCO even further by consolidating workloads at the edge. Consolidating workloads requires normalizing, filtering, and enriching data for seamless app integration. They can do this any number of ways, but it can involve the use of hyper-converged edge systems on gateways, edge servers, or industrial PCs. This means the edge system can process multiple streams of data on a single piece of hardware, lowering TCO and reducing the volume of data that needs to be published to the cloud.

Following this strategic guidance should help smart manufacturing vendors and their clients implement scalable solutions.

For more insights and perspectives on manufacturing and the industrial Internet, please check out ABI Research’s Smart Manufacturing service.