How to Leverage Edge Computing to Get the Most out of Enterprise IoT Solutions

Summary

The Internet of Things has significantly stirred up interest in the topic of edge computing in recent years. The fact is that edge computing has high hopes for unleashing the potential of ever-increasing volume of data that is produced by IoT devices. By 2025, this volume is expected to reach a whopping 73.1 ZB, but to extract value from it, data flows respectively need to be competently allocated, managed and analyzed. Edge devices might be helpful here.

As statistics show, edge computing is already used in most IoT applications in one way or another. Going beyond routers and firewalls, edge computing allows you to perform smart automation and predictive maintenance, optimize the data lifecycle and, therefore, reduce costs. Although the intelligent edge is a new application, it is already possible to take a closer look at how it manifested itself within industrial, transportation, healthcare and other common Enterprise IoT solutions to get the most out of it. In this article, PSA explains which environments require edge computing, which cases of  IoT + edge tandem are the most promising, and how to distribute working loads competently over IoT ecosystem.
 

The place of edge devices within enterprise IoT solutions

The traditional architecture of Enterprise IoT solutions assumes that data is collected from things, such as products on a conveyor belt or equipment on a shop floor; then it’s transferred to the cloud for further processing and analysis. In such a system, edge devices are used in the traditional way–to distribute data over the network and control network traffic. Routers, firewalls, multiplexers and switches are among such devices where the latter allows the enterprise to connect to industrial devices. In general, they allow you to establish IoT enterprise-wide, regardless of its location.
 
For IoT automation solutions, AI-generated ideas in the cloud are sent back to the field to induce some action. Surely, this approach suffers from a lack of speed and throughput, which becomes critical for widely distributed architecture and a large amount of data that needs to be processed. This is where intelligent edge devices come into play, taking on processing, analytics, and action. Their benefit is revealed through their proximity to the data source allowing them to reduce latency through the delegation of some basic decision-making credentials to them.

• End-devices. Smart Sensors & Actuators, wearables, cameras, and other sensing-enabled devices are located as close to things as possible. Their computing capabilities are limited through compact sizes.
• IoT Gateways allow only potentially valuable information to reach the cloud. This is a bridge between things and the cloud: performing aggregation, pre-processing, filtering data, verifying its authenticity, and cleaning the raw data. They also provide management for the field devices, ensuring automated distributed architectures.
• Edge Servers is a general name for environments where extensive computing at the edge is taking place. A physical server, a laptop, an embedded system, or a system-on-a-chip can take over this function. This category has attracted special attention in recent times since it allows for the redistribution of computing loads throughout the system, and the introduction of Artificial Intelligence (AI).

 Thus, when placing certain loads on the edge, the system saves time on communication with the cloud, which allows for a quick response to the situation in the field. In addition, higher availability and reliability are provided, as well as security since the data does not go beyond the local storage. To make this configuration cost-effective, it’s crucial not to overload the cloud.

The most promising edge computing applications for enterprise IoT solutions

Cloud hybrid + edge
Since the main value of a hybrid cloud reveals the opportunity to shift workloads between various environments, the extension of such infrastructure to edge solutions looks as natural as possible. By connecting edge devices, more options appear on how to optimize workloads. This allows for cost-effective scalability and increased resistance in case the system malfunctions. In hybrid environments, the edge component provides additional flexibility to the whole system, while the cloud component increases coherence between the distributed assets.
 
When it comes to an excessive amount of data that circulates around the IoT ecosystem, you gain more value by putting real-time computing to the edge, leaving behind cloud computing deep analytics. At the same time, the cloud can be used as a management center providing transparency for the whole system. Since you won’t appeal to a specialist for every node if something goes wrong, Hybrid Cloud + Edge tandem allows increasing controllability. Open-source projects like Micro Shift help to extend mission-critical platforms like Kubernetes to the edge and keep them consistent.

Edge computing + AI/ML
Bringing AI to the Edge is gaining more and more popularity, as Edge can overcome the speed limits imposed on the cloud. Sometimes it’s the only way of implementing the AI-enabled IoT ecosystem, since an internet connection may be unavailable, or in case of increased data security. Anyway, it expands the possibilities for processing raw data in the field, that’s indispensable for time-critical automation solutions and for highly distributed systems. Including the Cloud in this chain becomes not only time-consuming but costly, since it requires significant resources in the form of Internet channel bandwidth, traffic, additional energy, etc. For example, processing information from sensors at an oil refinery generates more than 1 TB of raw data per day, which requires excessive resources to process it.
 
AI algorithms put to the edge devices bring new use cases, such as remote monitoring, and predictive maintenance, and advanced automation. Machine vision deployed at the edge also continues to gain momentum. But anyway, implementing AI in the IoT ecosystem requires cooperation between the cloud and edge. Training of AI still takes place in the cloud, since we need excessive computing power for this operation, while deploying is set on the edge. The increasing success of this model allows us to speak about AIoT (Artificial Internet of Things).

Edge Computing + 5G
The combination of edge and 5G promises to enhance the main advantage of edge computing–performing operations maximally close to real-time. So far, we can’t speak about the desired 1-millisecond response, but 5G is already 16 times faster than LTE1. This speed allows several applications to be easily deployed in the cloud, such as real-time monitoring of robots, drone control, automated vehicles, and even remote MRP services and surgery. The peak data transfer rate of 20 Gbps allows you to create augmented reality applications and operate with heavy data, such as 4k video.
 
In general, 5G allows the enterprise to deploy lower power edge devices while providing higher computing capabilities. The bandwidth of such a network permits you to connect as many as 100 times more devices than via 4G or LTE technologies! Such capabilities open the way for the most incredible applications of the Internet of Things, such as dark factories.

Energy harvesting on the edge computing
As Enterprise IoT solutions expand, the problem of powering devices arises, as sensors and other edge devices can operate outside the coverage area of wireless or mesh networks to cover the entire monitored surface. Fortunately, energy harvesting technologies are advancing, offering different energy harvesting technologies for different IoT applications, which will significantly extend the battery life of low-power devices. For instance, solar- or vibration-based energy harvesting systems are applied for in-built car devices for car-to-infrastructure communications, while energy harvesting sources of light or thermal energy sensors are successfully utilized for workplace automation. Among the latest developments, we can distinguish an autonomous NB-IoT module that uses the energy of ambient light. The solution is based on solar cells and PMIC with MPPT function.

Recommendations for introduction of edge computing to enterprise IoT solutions

• Edge infrastructure is already mature enough to create complex IoT applications with a balanced distribution of computing power between cloud and edge.
• Extending cloud infrastructures to the edge makes sense for real-time operations as it minimizes latency. Cloud servers in such an infrastructure increase transparency and increase asset control.
• When creating Enterprise IoT solutions, it is important to foresee edge components from the very beginning, to then be able to increase its benefits, including with the help of AI.
• Training models require heavy-duty hardware, so it's better to train AI in the cloud and deploy the finished model on the edge.
• Expanding IoT infrastructures requires the optimization of edge resources. The introduction of energy harvesting technologies might be beneficial here.