Satisfying the Industrial IoT’s Demand for High Performance at the Edge

As the IoT and increasing network bandwidths significantly change how applications are designed, edge and fog computing are powering server technology at the edge of the IIoT, and Server-on-Modules with Intel® Xeon® processor technology are stepping into a key role.

Edge servers perform important tasks for the network as well as for local applications, and in this role they are— in addition to IoT gateways— the most important building block for the IIoT. Fog computers fulfil very similar functions. But what exactly is fog computing? The term was coined in analogy to cloud servers: In contrast to clouds, fog settles in the valley. Clouds are high up in the network, fog is low down at the local IIoT level. Unlike edge servers, fog servers add real-time virtualization across multiple systems to the picture. So, fog servers provide a highly reliable and—depending on the design—real-time capable local infrastructure for the Internet of Things.

Figure 1: In addition to the bridge functions of gateways, edge and fog servers also provide complete control and management of applications locally.

Figure 1: In addition to the bridge functions of gateways, edge and fog servers also provide complete control and management of applications locally.

Client/server Computing in the IoT Era

Unlike IoT gateways, edge/fog servers don’t just function as bridges, but provide complete control and management functionality for the local applications, which are built by the servers and the slim clients, i.e. smart sensors and all types of devices (Figure 1). The borders are often fluid. Why would one need such a local IIoT infrastructure with edge and fog servers, which is basically a new way of client/server computing with complementary gateway functions? There are several reasons:

  • Continuously Increasing Data Volumes In the commercial segment, an example is on-demand streaming of videos that need to be decoded in accordance with the client. This requires transcoding capacity at the edge to optimize video resolution and compression in line with the respective bandwidth and resolution of the end device. If one were to handle all traffic via a centralized server, the network would currently be hopelessly overloaded. If, on the other hand, all computations would have to be executed by the smart device itself, many new applications would be impossible to implement.
  • Higher Latency Latency is higher when data and apps are retrieved from a central cloud instead of locally. Applications that need to react quickly require local performance. Compared to cloud servers, local servers can therefore increase quality of service.
  • Real-time Capability is only feasible for deterministic network designs, and also requires that all application data and responses are delivered within a defined time limit. A centralized cloud hierarchy, sometimes with transoceanic communication paths, limited WAN bandwidth and singular servers, cannot guarantee this. Locally distributed fog servers also offer increased availability through redundancy.
  • More Data Traffic Via Centralized Servers If one were to transmit all data from the sensors, sometimes delivered at high frequencies, to a centralized big data platform, the network would very quickly be overloaded by the number of data packets. Pre-processing of data therefore makes a lot of sense.
  • Putting It All Together Last but not least, if local applications consist only of smart sensors and actors— from RFID tag to motion controller—a local server that resides at the edge of the IIoT is required for all these clients in order to configure the many individual parts into a whole and to form an application.

IoT Servers See Big Demand

Many applications can benefit from high-performance edge and fog servers. In the industrial sector, Industry 4.0 installations are prime examples. For the Energy Sector, they help smart grids monitor wind farms or control micro grids with various renewable local sources. Facility management is a matter of managing complex sites. The transportation sector uses server technology in smart railcars, where it is used to improve inventory management or to operate infotainment systems. And the latter are of course also found on airplanes and coaches.

Autonomous driving also requires powerful on-board computers. The same applies to all other autonomous vehicle applications, such as drone surveillance. Other target markets are found, for example, in video surveillance and the DOOH/digital signage segment. Yet another large application field is the medical sector with IoT connected devices in the hospital as well as in home care and diagnostics. Lastly, there are Smart City applications and carrier-grade infrastructures—for all conceivable types of IoT services.

Defying Harsh Environments

With so many different IIoT applications requiring edge or fog computers, demand is for a robust, highly flexible server technology with a long-term availability. Most of the new edge, fog and embedded servers are an integral part of the devices or machinery, so they underlie space constraints. It would therefore be advantageous to implement all necessary interfaces directly on the board level. The most efficient way to achieve this is with standardized computer modules and application-specific carrier boards. For this purpose, congatec defined a new class of server-grade computer modules, Server-on-Modules. They make it easy to utilize the latest highly integrated server technology as application-ready components and to implement the necessary customizations with limited effort and cost.

Figure 2: Measuring 125mm x 95mm, edge/fog server designs with Server-on-Modules have an extremely compact footprint.

Figure 2: Measuring 125mm x 95mm, edge/fog server designs with Server-on-Modules have an extremely compact footprint.

Key to Success? Standards

A big advantage for the new Server-on-Modules is the availability of the open COM Express standard for Computer-on-Modules, which is hosted by the PCI Industrial Computer Manufacturers Group (PICMG). This specification provides everything needed to design industrial high-end embedded servers. The PCB layers are designed for high EMC compatibility in harsh industrial environments. The two double-row SMD connectors are robust and offer 440 pins for numerous I/Os. COM Express is optimized for standard high-performance computer interfaces; at the same time, it meets the highest robustness requirements thanks to a stable connection to the application-specific carrier board. High-end embedded designs therefore rely on COM Express, especially if the standard features of the motherboard do not meet the design requirements, or if the application has limited space (Figure 2). So why not use the same specification to develop fully server-class COM Express modules?

Server-on-Modules Bring High Reliability

Server-on-Modules differ from traditional Computer-on-Modules in processor type, feature set, power and performance class. Advanced Server-on-Modules provide powerful server-class tools to help manage distributed IoT, M2M and Industry 4.0 applications, making them the perfect fit for many devices connected via edge or fog servers. With server-class remote management and an integrated baseboard management controller with watchdog timer and power loss control, these modules handle remote monitoring, management and maintenance tasks, including out-of-band management—a must for industrial server technology, if it is to offer high reliability.

Figure 3: congatec COM Express Basic modules with Intel® Xeon® processors provide server-class computing performance and up to 32GB ECC DDR4 RAM.

Figure 3: congatec COM Express Basic modules with Intel® Xeon® processors provide server-class computing performance and up to 32GB ECC DDR4 RAM.

Performance Not Dependent on a Dedicated GPU

The world’s first Server-on-Module platforms presented by congatec are equipped with the latest 5th Gen Intel Xeon server processors and define a completely new performance class of COM-based embedded computing designs. The conga-TS170 module, for example, is available with different Intel Xeon processors: For applications that require particularly high parallel processing or graphics performance, the version with Intel Xeon E3-1515M v5 processor and Intel® Iris™ Pro graphics is the best fit. The GPU of the Server-on-Module provides 128 MB eDRAM and, thanks to 72 execution units, three times the parallel processing power of the Skylake architecture without Iris graphics. This gives developers of highly compact COM Express based server designs access to a new level of performance that would have previously required a dedicated graphics unit.

For applications demanding high media-processing capacity, the most suitable versions are those featuring the latest Intel Xeon processors E3-1578L and E3-1558L that Intel introduced at Computex and that congatec had integrated in its Server-on-Modules by the time of the launch. They also offer 128 MB eDRAM, and a maximum TDP of 45 watts. However, with up to 700 MHz the integrated Intel Iris Pro graphics unit provides double the base clock speed. This increases media processing power significantly since the hardware accelerators that are used, e.g. for fast video transcoding, invariably run at base clock and not in graphics turbo mode. This makes them Intel’s first real-time HEVC transcoding processors for up to two 4k HEVC output streams or up to 15 full HD HEVC (1080p) streams. And with up to 32 GB address space for video memory (compared to 1.7 GB for the Intel Xeon processor E3-1505M), reliable media performance is guaranteed even at high data loads. This benefits infrastructure applications for video conferencing, surveillance cameras and video-on-demand offerings.

For managing distributed IIoT edge and fog servers all conga-TS170 variants also provide powerful server-class tools. Thanks to Intel® vPro™ technology and congatec’s board management controller with watchdog timer and power loss control, the Server-on-Module is also fully equipped for remote monitoring, management and maintenance, including out-of-band management.

Server-on-Module Evolution Moves Forward

The next step in the evolution of Server-on-Modules will firmly embed this new module class from congatec in the upcoming COM Express specification 3.0. In fact, the Type 7 pinout—which is to be specified for the first time in the new revision—is specifically tailored to the needs of fog and edge server applications. Its main focus is the provision of powerful connectivity with up to four 10 Gigabit Ethernet ports on the module, using the pins of the three DDI graphic interfaces of the Type 6 pinout. This is a smart and largely compatible reclassification as server chips don’t generally offer graphics interfaces and server applications don’t require multiple high-resolution graphics output locally. As editor of the PICMG specification, congatec is actively involved in developing the COM Express 3.0 standard. Customers can therefore put their trust in congatec as a competent and reliable partner.
Dan-Demers-Pic_webDan Demers is the Director of Marketing – Americas at congatec, Inc. He holds a B.B.S degree in International Business from Grand Valley State University, Grand Rapids, Michigan and an M.B.A. from Ashford University, Clinton, Iowa. Mr. Demers has over 19 years of experience in embedded computing having worked with Fortune 500 companies in the Industrial, Medical, and Communications markets.

Share and Enjoy:
  • Digg
  • Sphinn
  • Facebook
  • Mixx
  • Google
  • TwitThis