What is edge computing?

This article originally appeared in the April '22 magazine issue of Electronic Specifier Design – see ES's Magazine Archives for more featured publications.

Edge computing has emerged in recent years as an important enabler of the increasingly powerful applications that are built upon embedded IoT devices. But what exactly is it and how is it enabling a more powerful IoT? Rob Oshana, Vice President Software Engineering, Research and Development, Edge Processing at NXP Semiconductors discusses.

Although it is closely associated with the Internet of Things (IoT), edge computing is actually a specific type of distributed IT architecture, defined as “delivering low latency nearer to the requests” or “as all computing outside the cloud, happening at the edge of the network”.

In the first generation of the cloud, data centres contained vast compute and storage resources, which enabled applications to be centrally hosted. These applications were shared across multiple users, who sent their requests from web and mobile applications.

Techniques such as virtualisation enabled further sharing and economies of scale.

Over time, the ‘cloud in a box’ concept emerged, where some user requests were processed outside of the cloud, reducing the latencies inherent within communication over the Internet. This computing on ‘the edge’ or on network elements, including routers and base stations, was first defined as ‘edge computing’ in 2009.

The IoT drives innovation in edge computing

For many IoT use cases (such as remote sensor monitoring), data rates are low and response times are not an issue. However, as the number and sophistication of modern IoT applications continues to increase, the traditional cloud computing model is coming under pressure. This is accelerating the growth in edge computing.

The connected automobile, for example, is packed with sensors, generating as much as 4 terabytes of data per day – and many onboard applications must act on sensor data in real time.

The ultra-low latencies demanded by these applications and others, such as industrial robotics and remote telesurgery, cannot be achieved across wide area networks. At the same time, the enormous and growing volumes of data generated by the IoT consumes significant amounts of costly network bandwidth and data centre storage.

Much of this data contains sensitive information and therefore security – from the point of collection all the way to the data centre – is a growing concern.

The core capabilities underpinning edge computing

The demands on the Internet and the cloud have driven significant innovation in edge computing, facilitated by the falling costs of power-efficient electronic devices, which are available in ever-shrinking form factors. It is now possible to embed much more processing power into the IoT devices which are collecting data, and therefore place decision making capability at ‘the edge’.

This increase in IoT edge computing is founded on a number of core capabilities, examples of which are discussed in the next four subsections.

Local processing power

At the heart of the said innovation is the microcontroller unit (MCU). Current-generation MCUs integrate powerful microprocessor units (or MPUs) with a rich set of connectivity, multimedia, and security capabilities, while providing real-time functionality such as interrupt processing and scheduling.

These sophisticated devices are powerful enough – once they’re in the domain of the data centre – to host software at the edge.

This capability enables local decision making, resulting in commands being sent to machines with very low latency. The amount of data flowing to and from the cloud is also vastly reduced. In parallel to these hardware developments, software advancements have also been key to the growth in edge computing and sophisticated machine learning.

Accordingly, algorithms can now be downloaded to the MCU, further increasing the intelligence located at the edge.

Power efficiency

This increased local processing power comes at the cost of increased power consumption. Many IoT devices, however, use batteries or alternative energy sources. In some applications, however, the devices may be located remotely, making it difficult or impossible to change the batteries.
Power consumption is therefore another important characteristic of an edge device and MCUs employ a range of techniques to minimise power consumption, using low-power silicon materials and partitioning functionality such that it can be switched off when not in use.

Consider, for example, a smartwatch application controlled by an NXP edge processor. Such a processor has two separate Arm Cortex core domains: one supporting the watch functions while the other is responsible for real-time processing. Power management and partitioning built into the processor enables each domain to be shut down into a deep sleep mode.

Security

The proliferation of IoT devices increases the ‘attack surface’ of the applications that depend upon them, and any data transmitted to the cloud is also vulnerable. A holistic approach to security is therefore essential: one that encompasses the edge application code, the data being processed, and all the data being communicated between the edge and the cloud.

Advanced security measures are therefore a key requirement of intelligent edge processing devices, and isolated secure subsystems and secure software enablement are fundamental to their design.

Connectivity

Wireless communication is at the heart of the IoT and applications can vary significantly in their networking requirements. Short range network protocols such as Zigbee, Thread, Bluetooth LE, and Bluetooth Mesh support low-power requirements and also enable mesh networking, while Wi-Fi is better suited where data volumes and speed are more important.

5G is rapidly replacing 4G/LTE for wide area communications, with its increased speeds and bandwidths and reduced latencies. Many applications, particularly where mobility is required, may use multiple protocols. Owing to this, multi-mode IoT devices, which implement several of these protocols simultaneously, offer distinct advantages.

Edge computing: a new hierarchy

The pressures placed on the cloud by the exponential rise of the Internet of Things have driven innovations in both hardware and software, leading to the growth of edge computing. The increased performance and options in communications networks have contributed to a new distribution of computing power between the data centre and the edge based upon the specific characteristics of the application.

Ultimately, by securely placing decision making and intelligence at the edge, this new hierarchy empowers a whole new generation of advanced applications.