Edge computing is becoming the extended arm of the cloud: According to a recent forecast by market analyst firm IDC, in three years’ time 40 percent of cloud installations will already have edge computing capabilities. It’s a trend that Telekom is also currently observing: "Data is the new oil. That's why it's only logical to bring refineries as close to the source as possible," says Sascha Smets, Senior Product Manager at T-Systems. "But that doesn't mean that edge computing will replace cloud models. Edge nodes always need an instance to control them. If edge computing resources and the public cloud are intelligently combined to form an edge cloud, then the best of both worlds can be used."
From summer 2019, Deutsche Telekom will be offering the opportunity to combine lean edge devices with the scalable infrastructure of the Open Telekom Cloud in the form of the Open Telekom Cloud Edge. But why exactly do companies need cloud services if they can already process data locally with computing resources? "Edge units usually consist of small, agile and lean hardware that takes on predefined tasks at the point of origin of the data, which often involve low latencies," says Smets. "In this tandem, the public cloud can take on more complex downstream tasks for which scalable computing resources are essential.”
The most common examples include Internet of Things (IoT) applications, such as the control of automated industrial trucks on a company’s premises. There's a division of labor here: The vehicles transport pallets, for example between the warehouse and the production line. And an appropriately programmed edge node clears the way: it can open roller shutters, for example, if it detects that an industrial truck is approaching the position and has to drive through it. Or prevent industrial trucks from colliding. This is done in real time to ensure a smooth operation and maximum safety. An edge computing system therefore serves as a control unit directly on site.
However, the truck receives its route from the public cloud, which can be connected via the Internet. This is where the planning platform is located, where companies define routes for their entire fleet of industrial trucks. And that is done centrally for all the locations where automated transport and forklift trucks are deployed.
By exchanging data with the public cloud, companies not only keep track of the position of their industrial trucks, but can also plan further steps with the data that is generated. For example, in maintenance: It’s possible to identify when the vehicle has to go to the charging station more frequently – a strong indication of wear or damage to the battery. Worn parts can be replaced or repaired to prevent breakdowns. In this way, the public cloud and edge nodes become a well-coordinated team.
Providers of automated industrial trucks, however, can benefit from thousands and thousands of anonymized pieces of device data – subject to user consent – which they can use for the further development of their products and analyze in the public cloud.
Digitalized quality control is another application scenario that many companies are already implementing: Edge nodes monitor the data stream from sensors in production and ensure that predefined threshold values are adhered to. If necessary, the edge system stops production in the same fraction of a second that a fault is detected. It’s a mechanism that prevents defective parts from circulating and potentially causing consequential damage. This process must take place in real time – production cannot wait for feedback from a cloud that may come with a few seconds delay.
However, the public cloud is ideal for the analysis of large amounts of data and predictive maintenance. For this reason, the data stream also migrates continuously to the public cloud, where it can be stored inexpensively and analyzed as needed. For example, for machine learning processes: "The combination of edge computing and the public cloud in the edge cloud offers the ideal conditions for all applications that require fast response coupled with high scalability and a wide range of functions," says T-Systems expert Smets, "and at a fraction of the cost that would be incurred for on-premises instances. Small edge computing units are already available at 2 percent of the cost of a private cloud instance. And by combining them with the public cloud, they also offer the same benefits."
In addition, edge computing can significantly accelerate artificial intelligence processes. In machine learning, for example, pre-filtered data streams make machine learning much faster. Relevant information is filtered out before it is sent for analysis. For example, edge nodes can be programmed to analyze video data from surveillance cameras while ignoring redundant images and transferring only those that show changes to the cloud. Then in a second step, the almost infinitely scalable cloud resources then take over the actual machine learning process – which, however, is much more efficient because pre-structured and filtered data is already available.
It’s a principle that can also be transferred to other areas: "For example, an edge device on a gas turbine calculates the audio signal of a microphone in real time into the frequency spectrum in order to transport only these small amounts of data to the backend," was the description in Crisp Research in a recent article on edge computing.
Another advantage of edge computing is the ability to centrally distribute new software and tasks to all the affected sites. For example, edge nodes on site and the public cloud in the data center share the work: Edge nodes collect data from sensors on machines, pre-filter them, sort out unimportant or redundant data and send the relevant data to the public cloud. There, the data from all locations is analyzed centrally. On the basis of the knowledge gained, companies can use the public cloud to adjust the parameters for controlling their production lines anywhere in the world in real time.
Similarly, companies can provide their edge nodes with software updates centrally from the public cloud. Or give them completely new tasks: So, for example, an edge node that analyzes images from surveillance cameras and forwards irregularities to the cloud today could take over the control of industrial trucks tomorrow.
Even today, lean hardware resources on site help companies to carry out latency-critical processes reliably and quickly. The combination with the public cloud provides companies with many more opportunities for deriving added value from data generated in the value chain. No in-depth expertise is really required: With the Open Telekom Cloud Edge, companies not only receive the necessary hardware resources, but also expert advice from the relevant experts. Interested companies can contact the Telekom Cloud team directly.