What is edge computing?

Edge computing is one of those terms that produce a reluctant nod representing some vague recognition of the term but leaves a big disconnect between the term and real-world examples. So — let’s sort things out.

What is edge computing?

The term edge computing refers to the concept of placing computers hosting applications and data geographically close to the end-users and machines that interact with them – at the edge of the network.

Examples of applications that are fit to run at the edge include journal management systems in hospitals, video analytics applications in retail stores, and production line control systems on factory floors. Or any applications that benefit from running close to the environment in which they are expected to contribute value.

The two histories of edge computing

To understand the impact of edge computing in modern infrastructure, it’s helpful to look at where it’s coming from. The history of modern computers starts with the mainframe and time-sharing era of the 1950s and 1960s, where large computers ran batched calculations in labs. Through micro-computers for individual use in homes and workplaces in the 1970s. The microcomputers eventually led to the Personal Computer revolution in the 1980s. And with the massive sea change brought about by the internet, we eventually ended up with cloud computing starting in the 90s, where the computing and storage resources of massive amounts of computers in centralized locations are made available on demand without active management by the user.

As the 2010s rolled around there was a gradual increase in usage of the term “edge compute” to capture the increased focus on the value of running applications for data processing closer to the users and machines.

Stepping back, we can think of the history of what we now call edge computing from two angles.

• The edge that has always been around One is about the fact that we have done various types of computations across locations for a long time, and by connecting those computational machines to the network they became edge computing.
• Edge as the answer to the increase in internet traffic The other is to think about edge computing as a way of making applications that were born on the internet more scalable and therefore more valuable also at the edge.

The edge that has always been around

Throughout history, there has always been the need to do computations across many locations. An example is the cash register. Patented in 1883 it came about to help with the addition of items and to produce a printed record of sales transactions in the form of a receipt to avoid embezzlement. And 90 years later, in 1973, the first Electronic Cash Register (ECR) was installed with networking capabilities. This meant that the computational machine that had been plodding along disconnected all those years had now found a way to phone home and share data with a central location.

Keep reading: What differentiates modern edge computing from legacy on-premises applications?

Edge as the answer to the increase in internet traffic

The specific label “edge computing” itself came about to describe a technical solution in the 1990s to what was then a problem based on the explosive growth of internet traffic and the web.

As the web became mainstream, the people tasked with operating websites experienced what they called the “flash crowd” problem. The traffic load from the ever-increasing amount of website visitors exceeded the capacity provided by commercially available servers and resulted in sites crashing or serving the web content very slowly.

A startup called Akamai came up with a solution. Their idea was to place the web copies of the content on many servers closer to users and serve a subset of the user population from each location. By putting the content close to the “edge” of the internet, they could provide faster service for users by serving the content from nearby servers.

Benefits of edge computing

The main benefits of edge computing come out of the inherent properties of an architecture that places computers close to users and sources of data. With the snowballing growth of decentralized data sources, many follow-on challenges around privacy, autonomy, performance, and economics are addressed by such an architecture.

• Local requirements on data privacy and residency can be hard to meet using data processing located in central locations. Placing the data collection and processing in the appropriate locales meets these requirements and allows for more precise rules on exactly which data to eventually export for additional processing and decisions.
• Many transactions done at the edge of the network must be designed to survive prolonged infrastructure outages. The risk of business-impacting outages is significantly reduced by putting the business-critical features on the premises and close to the transaction.
• Insights and predictive analysis are for many businesses needed in near-real time on data collected at the edge. Placing compute resources close to the data source brings low latency, high bandwidth and data offload as well as trusted computing and storage.
• The cost of transporting data collected at the edge for central processing can be significantly reduced, not the least following the explosive increase in such data. The cost of transporting high-volume data, like high-definition video, actually comes with a cost model that is not aligned with the value of the result. By processing the data locally, and generating valuable results without the cost of transport, the costs are brought in line with expectations

Example use cases of edge computing

Retail	ML/AI-based in-store applications e.g. video analytics, inventory management, and digital wayfinding. Integrated omnichannel approach. Improved customer experience with AR solutions and self-checkouts.
Industry and manufacturing	Deep insights and forecasting analysis from production lines in near-real time. Improved efficiency with ML/AI video analysis.
Healthcare	Activity tracking to ensure sufficient staffing and supply levels. Autonomous operations without vulnerability for connectivity disruption.
Energy	Deep insights and predictive analysis from on-site operations in near-real time. Personnel allocation with ML/AI video analysis.
Telco	Improved data privacy to meet enterprise, government, and telecom industry-specific compliance requirements.
Public sector	Easier accomplishment of GDPR compliance and delicate data management.

Different layers of edge computing

Perhaps the most confusing aspect of the term edge computing is that it is a broad enough concept that it is used to describe a very wide variety of real-world scenarios. We can make a simple classification based on the size, location, and purpose of the compute locations.

Keep reading: Defining the edge in edge computing

Regional and local edge

A regional edge location is an extension of the central cloud in that it is managed and offered using a cloud operating model. We can think of these locations as smaller versions of the very large hyper-scaler data centers that offer the same kind of infrastructure services, just with less compute resources and in more places. They are located before the last mile network to your compute edge (see below) location.

This is sometimes also known as cloud edge, referencing the network edge of the cloud provider.

On-site compute edge

The on-site compute edge can be said to be as far as general computer platforms go, i.e. computers running mainstream operating systems, and are configured to be able to run a multitude of applications at the same time. These types of locations are small enough that they can’t offer all the services provided by the central and regional clouds, and usually have hard limits on the available compute and storage. The type of compute hosts range from single small embedded devices to racks of servers. The essential characteristic is that it runs within the enterprise premise and uses the local network.

The compute infrastructure is usually owned and managed by the enterprise, but there are hybrid models where cloud providers put managed compute hosts within an organization’s premises.

Device edge

The device edge can be said to comprise all the smart devices connected to the physical world and sending and receiving large amounts of data to and from devices for processing and analysis. They are typically not based on general compute hosts but have limited compute resources and are in many cases updated with firmware-style software, including both the operating system and specific applications.

Another common term for these kinds of systems of physical devices is the Internet of Things (IoT) devices.