From Cloud to Edge
Over the years, the cloud has provided all the processing power to the IoT endpoints. Not anymore. The focus has shifted from collecting massive data to meaningful use of analytics. It is a waste of network resources, for example, when a smart thermostat persists in sending its temperature readings every five seconds when the readings remain unchanged. Not only does the succession of readings gobble network bandwidth, the cloud server has to find storage for the data, and it wastes processing time. On the other hand, a sudden change of temperature from ambient to 400 ºF may indicate a fire or explosion nearby. Knowing when such a drastic change occurs constitutes “meaningful use.” So, what is edge computing? Instead of having cloud servers doing all the processing, at the edge a network’s endpoints/network nodes or smart sensors will perform some of the processing. Edge computing can potentially increase overall network efficiency and reduce costs. With edge computing, a smart thermostat will only send data to the cloud server if the readings exceed the temperature profile.
Modern Edge Computing Emphasizes Relevant Data
To monitor the performance and process of oil drilling, a modern offshore oil rig may be outfitted with up to 30,000 sensors. Imagine how much data would be generated, yet only one to two percent of that data would be useful. Other modern marvels, such as a smart, connected factory, can easily generate 20TB of data on a daily basis. Instead of processing all that data at the cloud, an edge device will process data locally and make decisions to send only the relevant data to the cloud server.
Figure 1: To monitor the performance and process of oil drilling, a modern offshore oil rig may be outfitted with up to 30,000 sensors.
How Edge Computing is Changing
In the interest of making more meaningful decisions, developers are incorporating artificial intelligence (AI) into more and more edge devices. For example, GPU processors and/or modules are speeding performance and making stronger inference decisions in mining, retail stores, smart parking, smart cities, manufacturing, and more.
What is the role of smart sensors in edge computing? Sensors are getting smaller and smarter. In the past, most sensors simply performed the basic function of converting physical parameters such as motion, stretch, heat, and moisture to analog electrical signals or digital data. Smart sensors can do much more. For example, smart leak detectors can monitor and detect gas or liquid leaks. When a leak is detected, it can wirelessly alert the central server that actions need to be taken. This real-time alert can do a much better job than humans can to avoid potential disasters. While edge computing offloads cloud servers, smart sensors can offload the edge devices. Depending on your perspective, some would classify smart sensors under the edge computing domain. Note that smart is not limited to processing power and decision making capability. Scott Nelson, Chief Product Officer and VP of Product at Digi International, pointed out that smartness includes multiple elements, as shown in Figure 2.
Details on the “Sensor + Edge Compute + Cellular = Recipe for IoT Retrofit Success” session presented by Scott Nelson at Sensors Expo & Conference 2019 on June 25 can be found here.
Figure 2: Smartness is multi-dimensional which include availability, security, programmability, machine learning capability and its physical property. Image credit: Digi International, Inc.
The Future is Bright for Edge Computing
Edge computing will continue to evolve as more powerful processors become available to include AI functions. Coupled with smart sensors, the edge domain will take over local decision making. This is particularly useful in building smart cities, performing cybersecurity duties, and carrying out many of the automated functions. As processors are becoming more powerful, lower in cost and consuming less energy, more functions will be performed at the edge. This, in turn, will open up new opportunities for developers of gateways, network hubs, edge devices, and sensors.
You can learn about these new opportunities from a multitude of sessions dedicated to edge computing at Embedded Technologies Expo & Conference this June 25-27 in San Jose, California.
Keynote | What Does 5G Mean for IoT
Gerardo Giaretta, Qualcomm Technologies, Inc.
Embedded Live Theater Session | Wednesday, June 26 at 1:35pm | Bringing AI to the Very Edge
Martin Croome, GreenWaves Technologies
Session | Unleashing the Innovative Power of Data from Core to Edge
Zvonimir Bandic, Western Digital
Session | What It Takes to Build an AI System
Michael Clay, Intel
Session | Edge AI: How to Do Real-Time Sensor Analytics with Inexpensive, Commodity Hardware
Stuart Feffer, Reality AI
Session | The Journey to Digitization: Bringing the Industrial IoT to Life with Edge-Cloud Infrastructure
Dhawal Tyagi, ioTium
Session| Affordable Computing at the IIoT Edge Node
Edel Griffifth, Adesto Technologies