Edge Computing and Manufacturing Applications
Published on : Monday 07-03-2022
Sijesh Manohar on how edge computing is driving a new class of manufacturing applications.
Edge computing is evolving to be a key technology component in the digital transformation of the manufacturing industry and the Industrial Internet of Things landscape. As edge computing becomes a mainstream technology, manufacturers wanting to maintain a competitive differentiation should have a plan in place. The edge computing market is expected to grow ten times by 2028.
Edge computing brings in the ability to process large volumes of machine data and data from sensors at the physical location to provide immediate responses to the control and automation systems. The data processing at the edge could be in the form of simple rules or complete machine learning (ML) inferences. The results are accessible by end-users from the applications in the cloud. Light custom applications such as analytical real-time monitoring dashboards for users at the edge can be made available on this infrastructure. Figure 1, shows a typical edge infrastructure at a factory shop floor that is close to the machines and managed by applications in the cloud.
Manufacturers are realising the potential and slowly and steadily pioneering an edge infrastructure into their landscape. The edge computing infrastructure can spread across multiple layers like a local or regional data centre. Implementing an edge compute infrastructure that consists primarily of a network of smaller devices on the shop floor allows data processing near the machines.
While this is important, how does it impact existing and future cloud applications for the manufacturing industry? Today new cloud applications follow a container-based microservices architecture. Such cloud applications are portable to be deployed in various underlying infrastructure platforms. The edge infrastructure could itself contain such a container orchestration platform making it easy for applications to be deployable at the edge. One might argue this is synonymous with on-premise software but is a new class of emerging applications that are essentially cloud applications that can extend to the edge.
In recent years, with the advent of SaaS applications, organisations are transitioning from traditional on-premise software to the cloud. ERP’s key processes, such as order management, procurement, production scheduling, warehousing, etc., were relatively easy to move to the cloud, but for manufacturing execution (ME) applications.
Having ME applications in the cloud makes it a centralised system with a view of multiple factories geographically distributed in different locations, which is highly beneficial in the longer run. Few of the advantages are listed here:
1. Order scheduling will have much more flexibility across factories. For example, an order previously scheduled to a factory can flexibly be rescheduled to another one if there have been any unforeseen events at the factory. Integration with order management applications will enable re-scheduling to consider the customer order priority. For example, direct orders are given priority compared to ‘made to stock’ orders.
2. It becomes easy to compare the overall efficiency, costs, profitability across factories in near real-time.
3. The advantage of near real-time integration to other ERP functional areas. For example, customers can now make last-minute changes to their orders since the order management application has a real-time view of manufacturing operations. For instance, customers can change the colour of the car they ordered until the order has not yet reached the paint shop. Or to adjust transportation schedules (an inbound or an outbound order) based on the instantaneous operational status on the shop floor, making the supply chain more responsive.
4. The benefits of being on a SaaS application, the software vendor is responsible for the overall maintenance, upgrades, and operations of the software.
On the contrary, the resistance to move ME applications to the cloud were primarily due to:
1. Security concerns – manufacturers are hesitant to have their manufacturing operations data in the cloud.
2. Execution systems need to have low latency interaction with the control and automation systems.
3. Manufacturers want the ability to schedule the upgrades and maintenance of these applications to minimise disruptions to factory operation.
With better security mechanisms and stricter regulations, more organisations trust cloud applications even with their financial and accounting data. I expect, in due course, these concerns on security will be less of an issue.
Barring the security concern, the other two are real and need a solution. Due to physical limitations, the total time taken for a round trip from the shop floor control system to the cloud application, cannot be reduced to less than ten milliseconds. A probable approach is to have the application deployed on the cloud and edge to leverage the advantages of both. This new class of cloud applications is architected slightly differently to become a distributed application. Figure 2 represents a ME application distributed between the cloud and edge.
Let me give a sense of what such an application setup could be. The application in the cloud will also have additional deployments at various physical locations called satellite edges. These edge applications are essentially the subset of the cloud application microservices also deployed on the edge infrastructure. The end-users of the application are oblivious of whether they are accessing the edge application or the cloud application. Transactional data recorded on the edge applications are temporarily buffered on its local persistence and later synchronised asynchronously to the cloud application in the background. It is important to note that the edge application is highly business aware of its location and only downloads a subset of the master data needed from the cloud application.
On the shop floor, a network of small devices on the manufacturer's local area network forms the edge infrastructure for that location. These edge devices communicate through a secure link to the cloud infrastructure that hosts the cloud application. Various deployment topologies can be considered, with edge devices connected in a tree structure with the root device or a limited set of devices having this secure connectivity. This setup depends on the way the application is architected. Each device can come with various compute and storage options that bring flexibility to the setup options. Additionally, a container orchestration platform on this edge infrastructure makes the application reliable and available. To manage the deployment of the application components to the different edge locations warrants the need for a separate central cloud application that provides this management cockpit.
The number of edge locations in such a hybrid setup must be configurable and depends on the business need. For example, a factory location may choose to access the application in the cloud and not have a satellite deployment.
The advantages of such a setup are:
1. The management of edge deployment provides flexibility for the manufacturer to decide the upgrades and maintenance windows for each factory. Cloud application maintenance and upgrades performed by the SaaS provider are unlikely to follow the manufacturer's schedules. The satellite edge applications could still function during those cloud outages to allow business continuity at the factories.
2. It allows the edge applications to have low latency integration with shop floor automation systems to provide much-needed immediate responses. At the same time, conceal likely intermittent connectivity issues with the cloud.
3. Edge applications of several ERP applications deployed at a location can interact to complete business processes locally. This capability ensures critical business process continuity when connectivity is temporarily lost to the corresponding cloud applications.
The third bullet point above emphasises the need for the edge to be self-sufficient to complete business transactions that involve multiple applications. Consider a situation when an assembly line gets stuck due to a missing component. The work-in-progress item needs to be put away in a warehouse to unblock the assembly line for the subsequent orders. To do this ME application will need to trigger a transaction in a warehouse management application that will automatically activate forklifts and robots to move the component to the warehouse. The edge applications of the ME application and the warehouse management application can interact locally to complete the business process without communicating with their respective cloud applications. The ability to ensure business continuity due to intermittent connectivity to the cloud is critical to the manufacturing industry. This addresses a real problem, as some factories might be in remote locations.
Such distributed applications combined with data processing at the edge make edge computing very powerful. In the coming years, edge computing will play a key role in transforming the technology landscape in the manufacturing industry.
(The views expressed in this article are those of the author and may not reflect those of SAP.)
Sijesh Manohar is a vice president in IoT product development at SAP Labs LLC Palo Alto, California – USA. Sijesh has successfully led global development teams in India and US for several innovative software products at SAP. Recently he has been extensively working in the research and development of new IoT and Edge products. He is the co-author of the book "Internet of Things with SAP" – SAP PRESS 2020. https://www.linkedin.com/in/sijesh-manohar-b080044/
