The age of edge computing has finally dawned. The rapid developments in digital and mobile technologies have made edge computing increasingly more prevalent and more critical to the success of businesses across a wide range of industries.
What is edge computing?
Edge computing essentially takes memory and computing out of the traditional data center to bring them as close as possible to the location where they are needed — often in the form of handheld or local devices, appliances, or point-of-sale or physical units that are distributed across different locations. Edge means different things to different industries. For automotive, for example, it may mean the growing importance of compute capacity in smart cars or in handheld devices used by technicians and service centers. For retail, it might mean new kinds of compute capacity available at point-of-sale systems and new experiences being delivered to customers in storefronts. Even in the fast food industry, Chick-fil-A shared it was running edge devices with container-based applications in every restaurant.
Edge applications that interact closely with local devices in the field are getting more sophisticated and intelligent with every passing quarter. There’s a lot of opportunity and promise in edge computing for both end consumers and for businesses. These applications can offer customers seamless and personalized experiences, help improve business processes and more.
Let’s first examine some of the key promising use cases for edge computing. We’ll then discuss the challenges with edge computing and some of the key questions business leaders should ask themselves around supporting the intelligent edge.
Five key edge use cases
Field and IIoT
Various sensors and other field devices across verticals, like manufacturing, transportation and power, are prime candidates for edge computing. These devices can be HVAC systems, energy meters, aircraft engines, oil rigs, scanners in retail, wind turbines, connected cars, RFID in supply chains, robotics, augmented reality and much more. These are often characterized by applications that collect data from edge devices and analyze it for different business use cases, including security management, predictive maintenance, performance or usage tracking, demand forecasting and so forth.
Smart cities and architecture
Many cities across the globe are vying for the tag of a “smart city.” IoT devices will make living in such cities easier for citizens. The use cases here range from municipalities providing faster urban services, such as repair of equipment, traffic management to reduce gridlock, public safety and green energy provisioning.
Customer experience in retail and hospitality
Customer sentiment data and social media data is collected and analyzed to improve customer experience. Data here is being captured by a kiosk or a point-of-sale system or terminal.
Telematics data can be used for navigation or to influence dynamic pricing for auto insurance, predict required maintenance and more.
Facial and image recognition
These technologies can be used to identify customers and reduce fraud in verticals such as retail, banking and entertainment.
The edge represents a unique computing challenge
Edge computing is very different from traditional data center environments. Here are some of the reasons why:
- Compute and hardware constraints. Many edge environments are constrained from the standpoint of a technical computing footprint. For example, in the case of embedded devices, you can’t fit as much hardware as in a full-scale data center.
- Accessibility and operations constraints. Edge applications often pose logistical difficulties in deploying and managing human IT resources and do not allow for high operator cost. Companies cannot have a dedicated admin monitor and service each and every edge location. This is true in the case of wind turbines spread across thousands of miles, sensors located in the depth of oil wells or mining sites, payment processing devices at every checkout line at a department store or thermostats located in peoples’ private domains. These operator limitations — either due to distance, the volume of devices, geographical accessibility or other cost and ROI considerations — mandate that edge applications not only have a very low computing footprint, but also a low technical IT overhead. They have to be plug-and-play from the point of installation and beyond.
- Remote management. In many environments, skilled personnel are not available to deploy and manage the system on a regular basis. An unskilled operator may need to perform simple plug-and-play deployments. This includes delivering secure edge application updates, debuggability in the instance of problems and deployment of additional devices. Edge applications need to be highly sophisticated and should be able to provide a range of features, including data caching in case of lost connections; raw data stream processing to filter and analyze relevant data; message brokering for event-based applications; device management; fault tolerance and so forth. Saving bandwidth costs of constrained networks is another important consideration.
- Connectivity. The ability of the technology provider to work with all sorts of latency and jitter issues is key.
- Support for air-gapped deployments. The ability to manage remote air-gapped devices in compute-constrained locations without resorting to manual intervention is a key requirement in edge computing. High latency to the central cloud can cause delays and interfere with the workings of the application. This also means that assumptions that originate in normal operations mode of data center networking often do not hold true in edge environments.
- Security is a foundational consideration. This includes secure communication from the data center to the edge and ensuring the privacy of data both at rest and in motion, anonymizing sensitive customer data stored at the edge. Other security requirements include establishing mutual trust between the central data center and edge devices, the ability to find and stop rogue devices in the event of an attack and secure communication over the WAN.
- Unified architecture and release processes that span both edge deployment targets, as well as traditional data centers. This is a major challenge since many edge applications also need to be deployed across other environments or data centers, creating a complex and practically unmanageable matrix of code bases, pipelines, deployment processes and operational practices. These architecture silos are as much a cause of technical debt as the data and processes silos.
Digital transformation via the intelligent edge
In light of the challenges above, the following are some key questions technology leaders should ask themselves around edge applications.
- How can we impact the customer experience in a connected world? What ecosystem partnerships can be beneficial for specific use cases? For instance, if you are a retailer, how can real-time customer traffic affect dynamic promotions to drive sales? Can you partner with banks or players in complementary verticals to gain customers?
- How will new edge experiences integrate with existing channels and processes to increase customer engagement? Based on the above example, can we suggest other products to customers based on their previous purchases?
- How can existing business workflows be augmented using edge insights? Going back to our retail example, how do popular items affect just-in-time manufacturing, procurement and supply chain workflows? Can these be augmented with this fresh data?
- What is the right architecture stack that enables the business to accomplish these capabilities? What does that look from a cloud, data and middleware design standpoint? Do we need a combination of virtual machines and containers running on the edge?
- What does this mean in terms of enabling self-service across the technology stack and various stakeholders? Should business transformation be about autonomous capabilities? IT should not become a bottleneck. For example, can certain busier stores receive more compute capacity on the fly without a long manual provisioning cycle?
- How can we do so while eliminating most of the grunt work around building, operating and managing clouds? In short, how can our IT department avoid becoming “cloud janitors?”
- How can edge systems continuously learn and improve? Can robots or drones deployed in distribution centers learn how to assemble boxes and stock them in the right areas? Can a mathematical model be deployed that enables a robot to understand metrics, such as mean time to restock, error rate, accuracy in business processes and more?
- How can the data being collected across edge devices help reduce unnecessary inventory, damage and other quality issues?
- From a compute cloud standpoint, the key requirements are to support low latency, a high degree of workload parallelism and fault tolerance — is this viable?
All IoT Agenda network contributors are responsible for the content and accuracy of their posts. Opinions are of the writers and do not necessarily convey the thoughts of IoT Agenda.