When billions of IoT devices communicate via 5G networks, all this data needs to be stored and analyzed somewhere. Depending on the criticality and computing power required, processing can take place either at the edge of the network - i.e. directly on the device - or in the cloud, or somewhere in between. Self-driving cars, for example, evaluate the majority of the collected data directly in the vehicle, as they have no tolerance for time delays. On the other hand, measuring devices that do not require real-time analysis can easily send their data to a server in the cloud. Certain application scenarios in an industrial context, on the other hand, are perfect candidates for fog computing: oil and gas companies, for example, have to process mountains of data somewhere that they cannot directly evaluate in their sensors due to the computing power required. Important tasks are therefore carried out in mini data centers in the immediate vicinity. With this solution, which is something between edge and cloud computing, the path of the data from the end devices to the data processing location is also shortened, resulting in a significant minimization of the time delay compared to conventional IT infrastructures.

Edge computing is gaining in importance

However, self-driving cars and many other application examples are dependent on edge computing. Communication in near real time is only possible with data processing directly in the end device or sensor - keyword SoC (system on a chip). The information collected at the edge of the network is mostly "disposable data" that has no long-term relevance; smart traffic concepts are no longer interested in whether the traffic lights were red five minutes ago. However, the IoT device in the vehicle needs such information for immediate decision-making. However, extremely large volumes of data and low latency times pose an insurmountable challenge for conventional data centers. From a geographical perspective, centralized data processing is always too far away: in order to communicate with each other, end devices and sensors must first radio the server in the data center and wait for its response. This makes it impossible to react quickly. In areas with limited bandwidths in particular, it is also difficult to adequately move large volumes of data to the cloud. Edge computing therefore forms a layer between the data center and the IoT sensors of the end devices at the edge of the network. By using analysis algorithms and applying pre-trained AI models, data processing can take place locally, thus ensuring faster decisions. The data, in turn, is aggregated so that only the derived results are sent to a central location for further processing or long-term storage.

However, edge computing is only one aspect of this new world, the current 5G mobile communications standard is the other. Due to the response times in the millisecond range that many IoT scenarios require, only wired networks have been an option to date. However, these have severely restricted the use of self-driving robots or other mobile devices, particularly in industrial production. The latest mobile communications standard provides a remedy. After all, 5G allows an immense number of devices and other data sources to be connected at a transmission rate of up to 10 Gbps. And latency in particular is greatly minimized thanks to 5G, which is an indispensable prerequisite for the automated coordination of robots and other machines.

Almost unlimited possibilities

IoT, edge computing and 5G therefore combine wireless real-time communication with real-time analysis capabilities. Companies in the industrial environment can receive live information about production, so to speak, thus avoiding unnecessary delays and fundamentally optimizing operations.

Monitoring critical areas is a classic application scenario: in addition to conventional sensors for temperature, humidity, pressure, voltage and the status of controllers and control valves, this also includes audio, video or lidar systems, which deliver significantly higher data rates and still need to be evaluated in real time. With the help of high-speed cameras and software based on artificial intelligence, irregularities on the production line that are no longer perceptible to the human eye can be detected. However, if only milliseconds are available for a decision, the loss of time due to data transport and processing in the cloud is already too great. Data analysis directly on site, on the other hand, prevents unacceptable delays - otherwise defective products would have already moved on down the conveyor belt before a robot arm could even access them and reject the corresponding part. Storing an image data stream locally on the edge device and analyzing and evaluating it there is fundamentally simpler, safer and also cheaper. Only the results of the analyses, including anomalies and critical incidents, are then transferred to the cloud. The models can be optimized there, which in turn are made available to all devices as an update.

The combination of edge computing and 5G also provides the urgently needed buffer in the event of unexpected events: classic robot tasks such as controlling different paint spray guns in automotive production could certainly be controlled via the cloud, but only a plant's own network - i.e. private 5G - provides additional flexibility for the production facilities. This is because in a largely automated three-shift operation, the requirements in terms of constant transmission speed, latency and reliability are extremely high. If there are disruptions to the network connection to the cloud or the data transfer does not take place at the required speed, the manufacturer faces a real problem. With an additional edge computing infrastructure, however, they can secure the necessary computing power directly on site in an emergency - after all, the entire frequency band is available for exclusive use with Private 5G - and can thus ensure stability in production with a data buffer while protecting sensitive information by defining their own security guidelines.

Security of sensitive data

With all these possibilities, however, the issue of data security must not be forgotten. In principle, information at the edge of the network is less protected than in the central data center. This is especially true when data processing involves various devices that are individually much less secure. This aspect is particularly relevant in light of the fact that production facilities are historically designed for availability and not security. The majority of data traffic is often not encrypted, which means that sensitive information is openly accessible in the network. This poses a high risk for remote access, maintenance and diagnostics, for example: sensors and actuators operate via two-way communication, which can be infiltrated by cyber criminals. This is exacerbated by the long service life of production systems of 20 years or more, which makes it much more difficult to update firmware, operating systems and APIs and to use anti-virus software. This means that vulnerabilities can often no longer be closed due to a lack of updates.

A truly secure IT environment therefore requires a comprehensive strategy. The first goal is visibility and constant real-time monitoring at all network levels in order to detect security threats in good time. The next step is to automatically prevent attacks and detect vulnerabilities identified by monitoring. In the case of unknown zero-day threats, machine learning can help to intelligently stop threats. An orchestrated platform ensures consistent, network-wide enforcement of policies. Micro-segmentation and access control of the various networks and devices also make sense. Finally, a private 5G network as a closed system already increases security because the data on the campus can be optimally protected by targeted, individually configurable measures. If the sensitive data is also processed on site and does not move to the cloud, further attack vectors can be ruled out.

The implementation - no witchcraft

The step from the laboratory to the production environment always requires a stable operating model. This is why a holistic analysis of the possibilities of 5G in combination with edge computing is unavoidable in order to identify individual, profitable areas of application for the company. Based on an economically viable use case, those responsible should not lose sight of the issue of sustainability. If, for example, the reject rate in chip or monitor production can be significantly reduced through the use of intelligent solutions, the environment benefits - because fewer valuable raw materials are lost. Rejects are a thorn in the side of every production manager. Good preparation is also essential. An edge cloud structure presents IT managers with a number of challenges, which is why a comprehensive platform is all the more important. Modern solutions support AIOps (Artificial Intelligence for IT Operations) to provide the necessary transparency and control. With the help of artificial intelligence, the management of workflows can be significantly simplified and problem solving in increasingly complex IT environments can be automated and thus accelerated. However, as companies often have neither the necessary human resources nor the required expertise to master the tasks associated with edge computing and 5G, an as-a-service model may be the answer. With a managed edge computing platform, companies can implement applications more quickly and use them productively. At the same time - depending on requirements - the system integrator acts as a central point of contact for the end-to-end operation of the solution. This enables companies to achieve their goals: greater operational efficiency and more security at lower costs.

The author: Marcus Giehrl is Practice Director Innovations and Smart Technologies Germany at NTT.

© NTT

The fact is: Whenever the amount of data on site is too large to be moved to the cloud or there are high requirements for low latency times for processing, edge computing helps. Even complex deployment scenarios can be realized that are centrally controlled despite their local distribution. However, edge computing alone is not the solution to all the problems associated with conventional IT infrastructures. Only in combination with 5G, the cloud and technologies such as artificial intelligence can processing at the edge of the network deliver on its promises. hap