Edge Computing
Computer manufacturers show a clear edge
Is the cloud hype over? This year, providers are increasingly offering powerful systems at the "edge" of the network - so-called edge devices.
For years, the industry's mantra was: data in the cloud, where storage space is unlimited. This is not a problem for a location in a well-developed industrial area with fiber optic infrastructure, but with the triumph of the Internet of Things, there are more and more applications that rely on narrowband and unreliable wireless connections. Or the latency times for a data connection to the cloud are simply so long that the entire application is not technically or economically viable. In these cases, the data must be processed or at least pre-processed on site. The range of high-performance equipment available at the edge is growing all the time.
Fujitsu, the largest manufacturer of industrial x86 mainboards in Germany in terms of volume, has opted for a new board format: Mini-STX. At 140 × 147 mm2 , this board format offers around twice as much space as the compact NUC boards from Intel. Above all, however, the systems are easier to cool and the board is large enough to fit enough connectors.
Iesy and Congatec are also presenting an innovation with Mini-STX: a modular edge server with a base board in Mini-STX format. Instead of a permanently soldered processor, the board has a module slot for COM Express Type 7, allowing the system to "grow with the tasks" or be scaled in terms of performance. The system integration was done by iesy, the modules with Intel Atom or Xeon-D come from congatec
Adlink takes a more traditional approach and scales not only the computing power but also the housing size: from a compact gateway the size of a notebook to a full-blown but fanless embedded PC with PCI and PCI Express slots.
The industrial PC provider Plug-in Electronic is also presenting an interesting approach: a fanless embedded computing system with Intel C236 chipset and Nvidia Pascal graphics card. Thanks to the Pascal GPU, this system can be used for demanding image processing tasks or for parallel calculations such as AI-oriented applications, robot control and machine learning.










