The embedded computing market is demanding more computing power across many application areas: Industry 4.0 applications require synchronization of multiple machines and systems; machine vision in collaborative and cooperative robotics requires real-time processing of videos and images as well as other environmental data. The same applies to autonomous logistics vehicles. Many of the edge computing tasks that arise around the development of 5G networks require server class performance by default. Not to mention the growing need to deploy virtualized on-premise equipment (vOPE) in harsh environments to perform functions such as industrial routing, firewall security, and VPN technologies executed via software on generic embedded hardware. It isn’t only in this context that the ability to manage embedded systems remotely – both on the fly and out-of-band – is growing in importance. New fields of application such as inference systems for artificial intelligence (AI) also demand a lot of computing performance. System consolidation by virtualization with hypervisor technologies increases the demand for higher embedded system performance even more.

A new performance class at the embedded server level

All of these application fields demand a new performance class for the embedded server level, which classic embedded computer technology was unable to deliver until now. The number of processing cores as well as the bandwidth and number of high-performance interfaces were simply not sufficient and comprehensive remote management features were missing. However, processor manufacturers have made their server technologies increasingly efficient in recent years and now also offer multiprocessor platforms for the design of embedded edge servers, such as the Intel Xeon D1500 and the Intel Atom C3000 processor, or the AMD EPYC Embedded processor families. They convince with multiple 10+ GbE performance, massive latest gen PCIe lanes and comprehensive RASM (reliability, availability, serviceability and maintainability) features known from the classic server segment. Since such processors with a 30W TDP are soldered, it is even possible to develop completely passively cooled systems in robust designs.

New embedded form factor needed

Embedded computing vendors have made these new embedded server processors available on standardized Server-on-Modules on the basis of the COM Express Type 7 specification. This standard has been developed by the vendor independent PCI Industrial Computer Manufacturers Group (PICMG), under the editorship of congatec. Officially adopted in April 2017, this Server-on-Module standard is specifically developed for the design of flexible carrier-grade micro servers and rugged industrial servers. They function as building blocks for custom system designs, which are often demanded in the embedded server area since regular off-the-shelf server boards cannot be used in the majority of embedded applications. In many cases, such boards don’t fit the available space nor do they meet customers’ interface demands, which almost always differ individually in terms of number, configuration and location.

COM Express based COM

Image 02: COM Express based Computer-on-Modules integrate all core components on a single, standardized and application-ready off-the-shelf building block. The customized executes all application specific external interfaces and controllers.

Application-ready platforms

COM Express Type 7 Server-on-Modules integrate all core components, including CPU, RAM and the platform controller hub (formerly called south bridge) for peripheral interfaces such as Gigabit Ethernet, on a single, standardized and application-ready off-the-shelf building block. They utilize a standardized connector to the carrier board, which is customized to the individual application. On the carrier board, the dedicated interfaces are executed exactly where they are needed. To simplify the carrier board design, comprehensive design guides as well as free circuit diagrams are available. With a simple module change, a module/carrier board combination can easily be scaled in terms of performance. It can even be upgraded across processor, socket and memory generations, offering highest scalability. This scalability translates into a high degree of investment security that results in cost advantages and a maximum payback of NRE costs over decades. An extensive community of embedded vendors ensures competitive pricing and also provides a rich ecosystem of off-the-shelf accessories. In addition, as an independent standardization body for COM Express, the PICMG ensures continued vendor-neutral development of this Computer-on-Module standard.

Cost-saving performance upgrades

In the edge server segment, the easy upgrade capabilities for the installed base are most convincing for operators, as earnings per gigabit bandwidth are constantly decreasing. Operators of high-availability server farms are therefore looking for ways to optimize their investments in new rack performance. With Server-on-Modules, they don’t need to expand data centers in order to meet rising performance requirements. They only need to change the module. This is also of great importance for cloudlet server installations on cell phone towers, as there is normally no space left for further expansions. OEM vendors such as Christmann estimate that costs for migrating to a second generation amount to only about 50%, including all services that must be provided for new configurations, qualification and installation. When applying these costs, for example, to three innovation cycles, total investment is reduced to about two-thirds (200%/300%). This makes Server-on-Modules the essential enabler of fastest upgrades at lowest cost at the various edges. In addition, they contribute to a better environmental footprint of these systems over the entire lifecycle as most of the components can be re-used in the second and third phase of the installation.

All you need from a single source

congatec offers COM Express Type 7 Server-on-Modules with all important AMD and Intel processors. The company also provides cooling solutions tailored to each single processor that can be adapted to customers’ housings to leverage the maximum processor performance in their designs, as performance mainly depends on the die temperature. Carrier boards and system designs are available as well from congatec and its partners such as Connectech or iesy. Real-time hypervisor implementations are possible with the RTS-Hypervisor from Real-Time Systems. And in cooperation with leading vision providers such as Basler, congatec even enables the development of ultra-flexible, completely passively cooled systems for autonomous robotic vehicles with artificial intelligence (AI) by combining Basler camera technology with COM Express Type 7 Server-on-Modules and multiple GPGPU cards. So OEMs have many options to accelerate their time to market with these application-ready platforms from one of the world’s leading embedded computing vendors.