Unboxing: Advantech Edge Intelligence System, EI-52
No matter what kind of IoT system you are building, there are three “must-have” elements: sensor nodes, gateways, and cloud networks. The gateway is the transition system between sensor nodes and the cloud and is responsible for transmitting information and commands bidirectionally. However, due to the limitations of distance, gateway functionalities have been improved and now share partial processing workloads with the cloud, which is nowadays considered an edge computing resource.
Many different types of devices can be considered as gateways within different IoT systems. For example, in a home appliance IoT application, a smart speaker may play the role of a gateway. Each node is connected to the speaker by Wi-Fi or Bluetooth, and the speaker transmits information and commands. Another example might be an IoT application for the public. Cell towers act as gateways, nodes connect to them via 4G LTE or 5G and information and commands are transmitted by cell towers.
Caption: EI-52 Edge Computer
Gateways are also used industrial IoT (IIoT) applications, which require industrial computers (IPC). In IIoT, there are many examples of gateways. Some applications might simply use Raspberry Pi single-board computers as gateways, others full sized server rooms. However, many gateways used in IIoT can only perform basic functions and lack comprehensive supporting software and hardware.
Recently, I received an Advantech EI-52 (EI-52-S6A1) Edge Intelligence System for review. I unboxed it, tested the system and found that it was simple to use. In this article I will use EI-52 to illustrate key points and characteristics that should be considered when setting up gateways for IIoT applications.
Three Applications for Industrial Grade Computers
IPCs are very different from ordinary home and office use computers. They may be installed next to high-temperature boilers or in areas with high chance of collisions in visitor hotspot zones. Therefore, the mechanical design of IPCs needs to be more about reliability. And the mechanical design of EI-52 is not only built for extreme environments, but it also has industrial grade robustness designed in and can operate in -10 to 50 °C temperatures. There are three major applications for EI-52; first, indoor installations such as shops, malls, and department stores; second, production line automation in factories; third, self-service kiosks in retail and convenience stores.
Mainstream Windows OS with Long-Term Service and Support
Having investigated the server market in the past, I noticed that tower servers were the most widely used servers in the retail industry. And even with the growing market share of Linux, Windows OS is still the mainstream OS used in the retail industry, simply because store managers and staff are more familiar with the OS and they can change settings or do basic trouble shooting under Windows much more easily. EI-52 uses Windows 10 Enterprise Long-Term Servicing Channel (LTSC), which not only offers OS familiarity, but also guarantees long-term after-sale service and support.
EdgeX Foundry and a Strong Technology Ecosystem
EdgeX Foundry is a highly flexible and scalable open source software framework that facilitates interoperability between devices and applications at the IoT edge. In practice, gateways need to use Docker container technology to execute EdgeX Foundry. Due to a major upgrade with Go Lang of EdgeX Foundry in 2018, its security infrastructure was enhanced, and computing resource requirements were lowered. In addition, it now uses Docker container technology instead of Hypervisor (VM, virtual machine), which is more lightweight and retains plenty of hardware resources for other use and future expansion. In terms of weight and lower resource requirements, the old version of EdgeX, including container required 519 MB of storage, but the new version only needs 68 MB. When simulating the operation of 100 nodes, the old version required 1302 MB RAM, but only 27 MB was needed for the new version. Apart from this significant saving, the initialization of the gateway service program was faster, from about 35 seconds for each service to less than one second. I checked the storage space used by all the pre-installed software of EI-52. It was only about 38 GB, including the operating system, EdgeX Foundry, and all related software programs.
EdgeX Foundry has a strong technology ecosystem due to the support of a variety of hardware and software vendors. Its offerings include various built-in communication protocols, drivers for multiple sensor nodes (such as cameras and barcode readers), and sample code. The related documentation, historical issues, and discussion records are also very comprehensive. EdgeX Foundry really simplifies and speeds up the development of IIoT applications.
WISE-DeviceOn, Time-saving Backend Management Software
In terms of operation, EI-52 uses EdgeX Foundry, which uses microservices for managing nodes. In addition, the gateway also needs to be managed so Advantech developed a solution for device management called WISE-DeviceOn. WISE-DeviceOn is divided into two parts, one is WISE-Agent, which is installed on the device to be managed (in this case EI-52), and the other is WISE-DeviceOn cloud service, which is established and operated by Advantech as the backend. WISE-Agent is responsible for collecting the information of the managed device and transmitting it to the backend, and at the same time, receiving commands from the cloud for controlling the device.
Through the backend, the administrator can check the real-time status of many gateways in the field, such as network connection status and abnormal events. If abnormal events are detected, then further diagnosis can be carried out remotely. What’s more, device data can be further analyzed for predictive maintenance. Administrator can update the firmware on all gateways over the air and remotely control power switches, as well as configure notification conditions. More importantly, WISE-DeviceOn allows grouping of multiple gateways (device group), so that batch operations can be performed. This saves tremendous time and effort.
Public Cloud Support
After EI-52 collects all the information in the field, it needs to upload it to the cloud. In theory, any cloud service can be used if the data format and protocol are correctly mapped to the desired cloud service standard. However, most people choose well-known cloud services, such as Microsoft Azure, and use services focusing on IoT applications, such as Azure IoT Hub. As mentioned above, the cloud is an indispensable part of the overall IoT operation. Through data aggregation at the cloud, the administrator can be aware of the overall changes to field devices, discover subtle patterns from analysis of collected data and make follow-up predictions and responses.
Two Optional Functions: Inference Acceleration and Backup
EI-52 can perform AI inference locally as it shares inference workloads with the cloud, which makes it more timely and power-saving, and reduces data transmission cost (public cloud services are usually charged by data usage). In this regard, the 11th Gen Intel® Core™ i5/i3/Celeron processor in EI-52 uses the built-in AVX-512 VNNI instruction set, which can accelerate INT8 inference (the benchmark of AI inference in recent years) by utilizing hardware circuits. You can also choose to add Advantech's VEGA-330 AI module to EI-52, which is in a mini-PCIe form factor and includes two Intel® Movidius™ Myriad™ X VPUs, which significantly speeds up the inference of video streams.
Apart from the Edge AI module, the EI-52 can also be equipped with Advantech’s other 5G and Wi-Fi modules to meet various communication requirements for the cloud and sensors. The modules have all been tested for compatibility, performance, and thermal reliability to ensure stable operation. Gateways require higher standards for information security than sensor nodes and cloud networks, because cloud servers are mostly located in data centers with strictly monitored systems and several layers of protection, whereas gateways and sensor nodes are mostly placed in open environments and public spaces, which makes it difficult to have full protection. Gateways can be hacked not only through the internet, but also by using physical methods. When a single sensor is attacked, quite often it only creates small local damage, and workloads can be easily covered by neighboring nodes. But, once a gateway is attacked, the impact extends to the entire area and cannot be overlooked.
Therefore, optional anti-virus and data back-up and recovery software utilities should be installed in EI-52. Advantech has cooperated with software providers to test and verify that McAfee anti-virus software and Acronis data back-up and recovery software can be used to provide EI-52 with all the necessary protection and security measures.
In addition to software utilities, EI-52 also has a hardware chip (Infineon SLB9670XQ2.0) to provide TPM 2.0 security-related functions to avoid data tampering. At the same time, data is saved in a scattered manner, so that even if a hacker accesses the entire EI-52 device, the data inside the it cannot be correctly decoded without the verification key.
From this unboxing experience, I can conclude that although general simple and basic gateway solutions provide enough functionalities to get by, for long-term stable and reliable operation, a comprehensive edge intelligence solution is a must. Such solution should have robust mechanical design, good after-sale services, a strong software developer ecosystem, remote diagnosis/firmware update/batch operation, inference acceleration, and data security and protection. EI-52 focuses on these key points and is the ideal gateway to play a proactive role in IIoT applications.