What Is Edge Computing?

What Is Edge Computing?

Edge computing refers to processing, analyzing, and storing data closer to where it is generated—at the edge of the network. This enables rapid access to data insights and real- or near-real-time responsiveness by time-sensitive applications.

Enterprises can use edge computing to automate processes, improve reliability and efficiency, and drive innovation. Processing data at the edge also helps to reduce data transmission and storage costs.

How Does Edge Computing Work?

Many edge computing use cases—such as those in industrial automation, self-driving cars, and smart hospitals—are time sensitive and require split-second analytics to ensure that actions are carried out safely and reliably. They often require real- or near-real-time data processing and have strict latency requirements.

For example, vehicles with autonomous emergency braking (AEB) systems depend on real-time analysis of data from cameras and sensors to identify obstacles and automatically slow or stop a vehicle to avoid a collision. Artificial intelligence (AI) analysis of that image and sensor data happens at the edge in the car’s central computing system, rather than being sent to and from the cloud. This enables rapid action to prevent an accident.

By integrating edge computing into processes and workflows, enterprises can unlock many transformative use cases and drive efficiencies.

Edge and Cloud Computing Differences

The exponential growth in data collected by billions of distributed devices is driving a shift from sending data to the cloud for processing and storage to a distributed model where some computing occurs at the edge of the network, closer to where the data is created.

Doing business-critical computations in the cloud, such as real-time fraud analysis of stock trades, can be a nonstarter. Sending data to the cloud for processing can result in slower response times due to increased latency as the data travels across the network. Additionally, it leaves operations vulnerable to service disruptions that impede the timely transmission of data and can take critical workloads offline. This is a major setback to implementing time-sensitive applications that use analytics to make critical decisions.

Edge computing technology, by contrast, provides the responsiveness that time-dependent processes need. By doing most of the computation on devices and hardware at the edge of the network, latency and bandwidth consumption are kept at bay.

Edge to Cloud

While edge computing unlocks many opportunities to extract value from data, the cloud remains essential as a central data repository and processing center. Intelligent edge networking solutions can be used to automatically determine whether data remains at the edge or gets sent to the cloud for storage or deeper analysis, helping to optimize the flow of data.

Benefits of Edge Computing

Moving vital data functions from the cloud to the edge helps drive efficiencies and innovation. Here’s why:

• Low latency: Cloud computing can struggle to deliver fast, low-latency performance, resulting in inefficiency, lag time, poor customer experiences, and placing critical operations in jeopardy. Moving data processing and analysis to the edge helps speed up system response time, which is vital in near-real-time applications.
• Reduced bandwidth usage: Internet of Things (IoT) technologies generate large amounts of data. The cost of sending this data to the cloud can quickly become impractical. Processing and storing data at the edge minimizes the amount sent over the network, reducing bandwidth usage and fees from transmitting and storing large volumes of data in the cloud.
• Improved reliability: Service disruptions and inconsistent internet connectivity can hinder the flow of IoT data, dampening productivity. Being able to store and process data at the edge improves operational reliability. It also allows intelligence to be distributed to remote locations with poor connectivity.
• More privacy and data sovereignty: Edge computing helps to keep proprietary, sensitive, and personal information safe by limiting the amount of data sent over the internet to data centers. It can also help organizations generate value from data that is required by law to stay within a specific geographic region.

Challenges of Edge Computing

Implementation complexity is a primary challenge of edge computing. Edge applications must overcome any number of technical challenges, including legacy infrastructure, siloed data, and resource provisioning throughout a heterogeneous system with many variables.

Converging these disparate workflows onto a single unifying platform can help to break down these barriers, simplifying the integration of new capabilities. Converged workloads eliminate or minimize data silos, simplify management by improving interoperability, and reduce latency between data generation and decision-making. This helps businesses extract the greatest value from their infrastructure and investments.

Edge Computing Security

Securing the edge requires an approach that curtails complexity. Edge solutions introduce many components and devices to the local network, which increases the potential number of attack surfaces and heightens security risk. It can also complicate security management.

As such, fortifying the edge with hardware- and software-based security technologies is essential. Edge devices with built-in, silicon-based security technologies and a hardened platform can help support continuous protection against intrusions and malicious code. Meanwhile, a software-defined edge can help to consolidate security management, simplifying the deployment of security updates and allowing for easier integration with heterogeneous edge devices.

Examples of Edge Computing Devices

An assortment of edge devices are used to collect and process data at the edge. These include:

• IoT sensors: The most basic IoT devices, such as RFID tags and vibration sensors, collect data and send it to an edge server for analysis. This type of setup is common in real-time monitoring scenarios where time series data is streamed to a central hub for monitoring and analysis, allowing industrials to better understand their entire operation.
• Edge servers: Designed to reduce latency, specialized servers are used for analyzing data, storage, networking, and security at the edge. They can be located on-site, at decentralized self-managed edge data centers, or on a private network provided by a service provider at the network edge.
• Intelligent edge devices: These devices have intelligence built into them, meaning they can run analytics at the data generation site without communicating with a server for instant responsiveness. Examples include smart sensors that monitor outpatient vitals, smart valves for automated fluid control, and smart city signage that keeps commuters updated with timely travel information.
• Computer vision-enabled devices: Equipping intelligent devices with computer vision capabilities allows them to perceive objects. This type of AI is useful in a range of environments, from retail smart stores, to smart hospitals, to the manufacturing assembly line.
• Edge network solutions: Network components, including controllers, Ethernet adapters, and gateways, comprise the backbone of edge infrastructure. They connect edge devices to an edge server and also to the cloud so data can flow throughout the network. Combined with 5G wireless connectivity, these technologies enable enterprises to build out a scalable, high-bandwidth, low-latency edge cloud.

Edge Computing Use Cases

Innovative edge use cases enable enterprises to solve complex problems and gain a competitive advantage.

Retail

Intelligent kiosks are helping to redefine the retail experience with innovative and frictionless services like cashierless checkout, smart vending machines, and drive-through menus capable of natural language processing and contextual up-sells. In addition, edge computing can be used to improve retail inventory accuracy and help avoid supply chain snarls.

Manufacturing

By enabling flexible and responsive manufacturing, edge computing provides a foundation for Industry 4.0. For example, Industrial IoT (IIoT) sensors on the assembly line keep track of equipment wear and provide predictive analytics to pinpoint when maintenance is needed, minimizing downtime and costs. Elsewhere on the factory floor, near-real-time machine vision can check for defects and streamline packaging processes.

Education

In the classroom, edge solutions support smoother interactive and immersive learning experiences, facilitate adaptive learning through rapid data analysis and feedback, and help teachers by providing insights on learning patterns and comprehension. Smart campus edge systems can monitor pedestrian and traffic flow to detect safety hazards or proactively alert staff to facility or technology issues before they become widespread.

Healthcare and Life Sciences

Edge computing is helping health professionals in myriad ways. Medical staff can monitor patient vitals collected by AI-enabled wearables for improved outcomes. Radiologists use machine and deep learning inference to rapidly evaluate medical imagery to accelerate diagnosis. Laboratories are using edge devices to automate instruments and critical workflows to get results faster.

Edge Computing Hardware Components

Edge solutions are built from various hardware components that enable intelligence to be distributed throughout operations. These include:

• IoT and embedded processors: IoT-specific processors are used to provide the onboard intelligence for smart edge devices. These context-specific processors are optimized for real-time processing and accelerating inference workloads at the edge. Because edge devices are deployed in unique environments, often under challenging conditions, edge processors may have unique features, including support for fanless and ruggedized industrial form factors.
• Field-programmable gate arrays (FPGAs): These versatile AI hardware solutions are reconfigurable and can be programmed to fit the needs of various edge use cases, enabling updates and modifications without having to replace hardware. They are especially well suited for real-time processing, computer vision tasks, and neural network inference at the edge, where devices and applications need to be adaptable and high performing.
• AI Accelerators: Depending on the edge use case, AI accelerators, both integrated and discrete, can provide additional required performance and processing power. Integrated AI accelerator engines built into CPUs provide optimized AI performance without requiring specialized hardware that can complicate system design. Discrete hardware AI accelerators, like GPUs, can be used to augment the capabilities of the CPU to handle the challenges of the most demanding Edge AI workloads.

The Future of Edge Computing

The use cases for edge computing are expected to continue to proliferate. Today’s dominant computer vision approach at the edge is beginning to expand to include multimodel applications involving generative AI, natural language, and robotics. Advances in AI and edge computing technology could allow for large-scale transformation. Looking ahead, edge computing solutions could dynamically adjust production in manufacturing, accelerate critical health diagnosis, and help drive the transition to clean energy by integrating renewables into the power supply. By bringing intelligence to the edge, businesses can evolve and optimize their workflows for the future.