The Industrial Internet of Things is generating more and more data that needs to be managed. This data complexity needs to be mastered. What role does edge computing play in this?

Bernd Hantsche: It is often not very expedient to send all sensor data on a long journey to the cloud. In this sense, edge computing plays a decisive role: for example, when local monitoring is sufficient or data can be processed into information on site. However, if additional external data is required for data processing or swarm intelligence is to be established, cloud computing is definitely worth a closer look. In other words, it is ultimately important to define what is needed and what the goal is when making a decision.

When should industrial users rely on edge and when on cloud computing?

Hantsche : If long-term scaling is difficult or impossible to estimate and fluctuates by several orders of magnitude in terms of computing power or storage space, then cloud computing is easier to implement. The background to this is that the resources can be adapted more easily. Pay-per-use is the buzzword here and scaling is just a mouse click away. In contrast, with edge computing you often have to plan an acquisition, finance it and you yourself are liable.

A cloud can also be very attractive if you want to access external databases, remain flexible with regard to your software and processing tools or want to exchange data with third parties. The same applies to computing power and storage space: an on-site computer is subject to maximum values. In the cloud, 'computers' can become virtually infinitely fast at the click of a mouse and store infinite amounts of data in a practically fail-safe manner.

What needs to be considered when 'linking' edge and cloud computing? What are the biggest challenges?

Hantsche : In principle, data should be compressed, filtered and prioritized at the edge if possible. This keeps the amount of data and therefore the data traffic to a minimum. Surveillance videos can serve as an example here: Avoid sending a 4k UHD video stream to the AWS server. Instead, it may make sense for video analysis to take place at the edge and for suspicious persons or actions to be identified, for example. This means that only the data marked or relevant in this way is then transferred to the cloud if necessary.

The challenge lies in balancing the interface between the edge and the cloud: on the one hand, DSPs, GPUs and storage space are getting better, faster and cheaper; on the other hand, internet connections are getting faster, cheaper and, not least, more responsive. The large hyperscalers can also buy innovative algorithms and tools more quickly, integrate them into their offering and thus make them accessible to the market.

How should standardization and open source be seen in the context of edge and cloud computing?

Bernd Hantsche, Vice President Product Marketing Embedded & Wireless at Rutronik

© Rutronik

Hantsche: Standardization is definitely to be welcomed in the medium term, as this is the only way to improve a system in a modular way with manageable effort and risk. However, there is currently still a lack of binding standards. It is therefore currently advisable to use common 'quasi-standards' that are accepted on the market, such as Message Queuing Telemetry Transport (MQTT) for telemetry messages.

The advantages of open source in terms of security, update speed and innovation generally apply to software - including edge and cloud. The financial interests of companies will also ensure that some solutions are only available to buy as proprietary software. After all, this makes possible liability issues easier and there is usually binding support.