Since the term 'edge computing' first made its debut, it has been inseparably linked with the Internet of Things (IoT). An emphasis on the importance of 5G to the success of edge computing has solidified that association and subsequently de-emphasized the contribution of applications to edge and its explosive growth.

For every connected thing that enters the market, there is at least one application simultaneously launched. This is due to the need to collect the data devices generate and control devices remotely. We hear the term "connected" and think "network", but the reality is that things connect to applications that monitor, control, and manage them. The network is how they connect, but an application is what they connect to. Some of those applications are hosted in the cloud. Others reside on mobile phones. Still others are embedded in the thing. Regardless of where they reside, they exist and are essential to the adoption and subsequent success of the thing.

The extant relationship of things to applications has given rise to new application patterns that we call edge-native.

Different kinds of data

Edge-native application patterns are unique because they focus not only on interaction and logic flow patterns but flow of control and data.

Traditionally, the term data within application architectures refers to business-related data such as customer information, product catalogues, and order/billing histories. But as technology expands to things at the edge, operational data becomes a component of the architecture. Configuration and policies are unique data constructs that must now be factored into architectural decisions.

In this sense, edge-native applications are the first to bring together both IT (Information Technology) and OT (Operational Technology) requirements.

Edge-native application patterns aren't just about where traditional data is generated or stored. Most edge-native applications will not exist in a vacuum, separate from applications and services in the public cloud and private/data center cloud. Billing, account management, support, and other traditional digital business services needed for edge devices and applications will not see significant benefits from deployment at the edge and are therefore likely to remain in the public/private cloud.

Edge-native patterns are increasingly focused on where operational data is processed, and subsequently where decisions are made.

Data-driven decisions at speed

Rapid advancements in chip technology are offsetting the end of significant increases in computing power as described by Moore's Law. The ability to focus computing power on a specific type of task – such as cryptography – is not new. What is new is the application of this principle to other computing tasks such as the complex mathematical processing required for machine learning and analytics. This usage is growing explosively in the form of optimized compute: GPUs (graphics processing unit) and DPUs (data processing unit).

Optimize compute focuses on performing complex calculations. But more importantly, they focus on performing those calculations at speed.

The ability to rapidly ingest, process, analyze, and produce a result means a decision can be made in near real-time. This capability is not unique to the edge. After all, any optimized compute can do this task irrespective of where it may physically reside. What is unique to the edge is coupling optimized compute with proximity to the end user. The result is data-driven decisions at speed.

The truth is that we have not yet managed to find a way around the laws of physics, and the cost in terms of time to transfer data across long distances is still high enough to have a negative impact. While we may be able to leverage optimized compute anywhere to quickly come to a data-driven decision, we can't account for latency that may delay relaying that decision.

The ability to deliver data-driven decisions at speed is a critical one for users – whether human, machine, or software –across many industries.

• Manufacturing has long required data-driven decisions at speed to identify imminent failures that could endanger human lives or cause outages that cost the business millions.
• Healthcare needs data-driven decisions at speed to provide preventative or emergency care that could mean the difference between life and death.
• Companies managing remote outposts responsible for natural resource processing rely on data-driven decisions at speed to prevent accidents that can adversely impact the environment.

One of the difficulties organizations have had in adopting optimized compute has always the need to employ a team of experts in system-level and hardware engineering. F5 believes the answer is to provide a platform and framework that enables the rapid development of services and systems capable of tapping into the benefits of optimized compute at the edge.

This belief is driving F5 to develop solutions with partners for our edge-as-a-service platform that enable any organization to harness the power of data-driven decisions at speed.

The role of data processing in edge-native applications

The ability to take advantage of data-driven decisions at speed will disrupt existing application architectures.

While data has always been important, architectures have traditionally focused on its storage. Storage will remain a factor in architectural decisions, but edge and the availability of optimized compute introduces a new design factor: data processing.

This will result in new application patterns – edge-native patterns – that distribute data processing based on requirements that factor in speed of decision delivery. Edge platforms will need to support these patterns by embedding optimized compute capabilities along with policies that govern privacy and security of the data to be processed.

F5 aspires to deliver that platform – we call it Edge 2.0 – as we clearly see how data and edge computing will inevitably disrupt enterprise architectures and require a new approach to development, distribution, and delivery of applications. Data – traditional and operational – will be a key component of that approach.