Embedded server modules for edge data centres

The growing need to lower latency and reduce energy-hungry data traffic over long distances is leading to rapidly rising deployment of server technology and data centres at the edge. Since harsh environmental conditions can often prevail in outdoor settings and investments in these infrastructures are usually made for a longer term, robust and long-term available server technologies are required. Server-on-modules based on the COM-HPC and COM Express standards offer a highly efficient design basis for this.

In a study from February 2020, IDC predicted that half of all enterprise critical infrastructures will be deployed at the edge in the future. As a consequence, the global edge computing market, estimated at $3.5 billion in 2019, will grow at a projected compound annual growth rate of over 23% between 2020 and 2026, according to Grand View Research.

The vertical markets in this segment are diverse: one of the largest submarkets and a trend enabler is telecommunications, with 5G technologies in the lead. Here, high bandwidths lead to a need for more performance at the local level. The number of required base stations is also going up since fewer devices can be managed at full bandwidth utilisation and the radio ranges of these base stations are shorter than in previous mobile communications standards.

The IT and telecom market for edge data centres is outstripped only by the market for colocation applications, which involves data centres that are located in the cloud to ultimately be shared by companies. These two segments are by far the largest, accounting for 54%, or more than half the total market.

Edge computing demand is also expected to rise in the banking and finance sector, which comes as no surprise since the number of digital transactions increases with the availability of faster payment terminals and the proliferation of contactless technologies. Demand for edge server technology is also growing in the public sector, spurred by greater security requirements and the smart city trend.

Other segments include applications with higher and higher bandwidth demands and lowest possible latency in the energy and healthcare sectors, as well as the market for autonomous driving infrastructures and industrial manufacturing. The latter accounted for about 10% of edge data centre revenue across Europe in 2019 and is expected to grow by 20% by 2026.

What needs bearing in mind is that conditions at the edge are different from those in air-conditioned server rooms. Especially the smaller edge data centres, sometimes also called mini, micro or even pico data centres, are often installed in the open air, housed in protective rugged outdoor cabinets or containers. Exposed to wind and weather, they must provide optimum protection against heat, cold and climatic shocks as well as moisture, dust and, if used offshore, sometimes also from salt water.

ASHRAE, the US-based professional association for heating, refrigerating, ventilation and air-conditioning engineers, whose standards and guidelines many data centres follow when it comes to server room management, has issued a bulletin with guidelines from the Technical Committee 9.9 on how to prevent outages and problems. Apart from a provision to compare the environmental conditions with the component specifications, the bulletin essentially only points out that doors should not be opened unnecessarily for service and maintenance purposes as this risks letting in very hot, cold, damp or dusty air too quickly.

A maximum permissible temperature fluctuation of 20°C within one hour and a maximum of 5°C in 15 minutes are specified, with the added note that any temperature change should be gradual rather than sudden. This almost sounds like the even-tempered conditions of server rooms are to be transferred to the desert or to offshore drilling platforms and wind farms.

Would it not make sense to develop edge servers in such a way that they do not have to be treated like raw eggs? Then there would be no need to spend quite so much money and energy on air conditioning because the hardware is specifically designed for an extended temperature range and can also withstand temperature shocks.

This applies especially to scenarios where no multi-racks are deployed, which consist of just a few servers operating redundantly as a fog server farm to control real-time processes reliably with the lowest possible latency. Processor manufacturers like AMD and Intel have already recognised this need.

This is also why more and more entry-class server processors are coming onto the market, with specifications similar to embedded processors. They are designed for use in harsh environments as soldered processors with BGA sockets and cover wider temperature ranges than standard IT. They are also supported for seven years and more, which makes it easier for OEMs in this segment to launch their own products and maintain them over the long term.

To address the needs of rugged fog and edge servers, the embedded computing community has created two module specifications: COM Express Type 7 and the new COM-HPC Server standard; specifically designed for these new processors. They are a perfect basis for the development of the many small, distributed edge and fog server installations, required for instance in the automotive sector for all the small server racks along roads and highways.

Thanks to real-time hypervisor technologies, they also enable perfect performance balancing. If this was not already enough, they also significantly reduce the total cost of ownership by making it easy to scale the performance when the next fog generation comes along. At that point, instead of having to replace an entire rackmount system, only the processor module needs to be swapped.

Yet there are other challenges beyond ruggedness when it comes to real-time edge computing applications; for instance, finding the best set-up for the fog services as well as the edge devices that are connected via time-sensitive networks. Given the many tasks to be handled at the edges, OEM customers and professional end users need a customised mix of computing capacities. Modularity is therefore key, both at the hardware and software levels to deliver perfectly tailored application-ready platforms from a single source.

Modularity at the hardware level has long been Congatec’s core competence. The company’s proprietary hypervisor software for real-time capable virtual machines complements the platform offering for fog servers at the software level and creates the basis for OEMs to develop their own tailored rugged fogs. Congatec plans to expand its offering with future support of partner solutions in the areas of vision, AI, VR, AR, big data analytics and dedicated edge computing services. The company is also evaluating virtual machine configurations to be able to offer IoT gateway and security functions for the detection of vulnerabilities, attacks and anomalies, or cryptographic functions. This might even go as far as supporting FIPS 140-2 Level 3, or BSI Common Criteria EAL5 standards used in high-security applications.

Andreas Bergbauer, senior product line manager, congatec AG