Using Edge Intelligence to boost Cloud performance

Software architects have a tough job today deciding on the best approach to a particular problem, because of the wide range of proven technologies available to them. Mainframe computers with terminals gave way to client-server architectures in the last century; these clients reduced to thin-clients and, in this century, we have seen these reduced to zero-clients. Computer engineers have virtualised mainframes into datacentres, which serve their clients in the same way logically, but in very different ways physically.

Enterprises outsource their datacentres to the Cloud, buying in infrastructure, platforms and software as a service. These “as a service” models involve paying monthly by users, data and performance, rather than paying outright for hardware and software licenses. Employees of these enterprises access their applications from their own computers, usually through internet protocols. Because all these technologies still exist, software architects have many options to navigate.

The common factor in all these architectures is that the data and the computing resources sit together. In the very different world of industrial computing, data comes from remote machines, sensors and analysis at the edge. 

The supply of data from industrial remote devices introduces technical challenges, and software architects need to discuss the different options with operational (OT) and telecommunications engineers to get the right solution.

Simple communication using existing networks, such as wired Ethernet LAN or wireless Wi-Fi already connected directly to the central database or Cloud through a permanent WAN, enables a straightforward communication.

In industrial environments, however, straightforward communication is rare. More commonly, the owner has commercial, security or technical reasons why the IoT data cannot use their network. Designers must select alternatives, usually cellular or other wireless technology, which introduces another choice, whether to use a private or public network.

The problems faced when using these communication systems include:

• Intermittent and unpredictable connectivity
• Lower transmission speeds
• Higher latency
• Jitter
• Higher transmission costs, usually directly proportional to data volumes
• Data security, especially in public clouds and networks, and in Internet-connected devices
• Device and user authentication
• Common protocols

System architects need to manage these financial and technical constraints to come up with the best solution. They can arrange to process data at the Edge rather than in the Cloud.

Edge computing introduces its own challenges, many of which may be alien to those brought up in the unlimited power of a datacentre.

·        Much lower performance due to fanless design of edge computers and devices

·        Environmental constraints such as space, ambient temperatures, shock, vibration and humidity.

·        Cost – there may be hundreds, or thousands of remote locations requiring compute resource. As well as the cost of the hardware, there is the cost of installation and maintenance.

·        Power supplies.

Edge computers can process large amounts of data from remote machines and sensors, and then transmit only the results. For example, they can send only the changes on an image, or the car number plate in text instead of a picture of the entire car.

Designers may want to use their Cloud analytical engines on the data. Edge computers can run agents that enable the same processing at the Edge, without having to transmit large quantities of data. This enables real-time decision-making without worrying about jitter and latency, and minimises transmission expenses. When the data is image or video, edge computers carry a dedicated graphics or video processing unit (GPU or VPU) to handle trillions of image processes in real-time.

How intelligence is distributed between Cloud and Edge computing depends on the application, the practicalities of installing Edge computers, their maintenance and operation, and the quantities and costs involved.

Advantech system architects deal with these problems every day, advising partners and customers worldwide, and delivering the computers at the Edge, with Fog or on the Cloud for industrial IoT applications. Send me a message if you are interested in talking more on this topic.