Edge computing in IoT: Increasing network efficiency to ease traffic
Edge computing in IoT is tied to IoT's growth in a number of vertical markets due to the benefits of processing data close to the data source, which relieves network overload.
Agricultural researchers estimate that, by 2050, the world will have more than 9 billion people, making it more challenging than ever to feed this ever-expanding population. But Patrick Smoker isn't waiting. Smoker, director of IT for Purdue University's College of Agriculture, and the college's 300-person faculty are already working on ways to feed a hungry world, relying on innovative networking technology to help them find the answers they need.
Time is of the essence: Food shortages have already hit the South Sudan, Somalia, Nigeria and Venezuela. Millions more people are at risk in other regions of the world, especially as weather conditions intensify as a result of climate change. While technology won't save the world in and of itself, its use can have a dramatic impact.
To that end, researchers at Purdue's College of Agriculture are using a series of IoT sensors, local circuit boards and edge computing devices from Aruba and Hewlett Packard Enterprise to track data on groundwater quality, plant health and emissions, so researchers can help farmers grow food more efficiently. The trick is to process the data at the edge of the network -- as close to the source as possible -- which aligns with the broader, global use of edge computing in IoT.
Enterprises are increasingly interested in edge computing, because, as more IoT devices are deployed, businesses need ways to quickly analyze and process the data generated by them. The goal is to eliminate the need to haul data back to the cloud or an in-house data center.
Today, only 20% of enterprise data is processed outside of data centers, according to Gartner. But, by 2022, the majority of enterprise data will be produced and processed outside data centers, Gartner said. By 2025, up to 90% of all data could be handled by edge devices. Purdue's researchers use edge computing for a number of reasons, according to Smoker. For example, edge computing is used to preprocess data, allowing staffers to extract only the valuable data from a raw image. This substantially decreases the size of data files sent over the network and optimizes the use of finite bandwidth.
In other cases, edge computing lets researchers analyze sensor data for quality assurance before adding it to a larger data repository for further study. Or, they can add appropriate metadata to sensor data, such as GPS coordinates and plot information.
"In still other ways, we can use edge compute to make decisions within the closed systems of farm implements, in the same way onboard computing is used in today's automobiles," Smoker explained.
This includes using sensors to track the performance of implements and using the data collected to take real-time actions based on the information received. That information could span a number of variables, among them the speed at which an autonomous farm implement travels across the field, the seed planting density or any number of actions associated with optimizations, like fuel consumption or fertilizer application rates.
"We use these tools to study crop genetics," Smoker said. "By 2050, we want to be able to feed the world's population, so we have to increase our capacity to produce food in a way that lowers the impact on the earth, but increases nutrition and makes it possible for crops to grow in new regions."
Focus on edge computing in IoT
Thomas Bittman, a Gartner vice president and analyst, labeled IoT adoption as one of the primary forces driving edge computing benefits. As a result, companies are racing to find ways to retool their IT infrastructure to stitch sensors and other remote devices with their existing operational technology.
It's important not to think of edge computing as a market, but as a computing topology that will be applicable to many unique markets: enterprise, consumer, industrial, things and mobile.
Thomas Bittmanvice president and analyst at Gartner
"It's important not to think of edge computing as a market, but as a computing topology that will be applicable to many unique markets: enterprise, consumer, industrial, things and mobile," Bittman said.
"I would say edge computing will emerge in all verticals, wherever there is an industrial plant, employees, customers or locations," he added. "The major effort will be enabling net-new use cases, going after smaller business moments, automating more broadly in plants, and increasing immersive experiences for shoppers and employees."
Data processing, compression and filtering are common tasks that will be performed on the edge," according to Mike Boudreaux, director of connected services at Emerson Automation Solutions, a manufacturing automation company in Round Rock, Texas.
Service providers are in a good position to deploy edge devices that have no application software on them. Then, cloud systems can push specific applications' software to the devices. This will let companies monitor applications they want to support and deliver software updates on a periodic basis, Boudreaux added.
Edge computing in IoT helps feed the world
Purdue's precision agriculture initiative relies on a variety of devices for edge computing in IoT, ranging from sensors, solar-powered wireless gear, servers and other devices -- all designed to help Smoker keep track of what's happening on the university's sprawling 1,408-acre farm, a few miles from the university's main campus in West Lafayette, Ind.
During any given research experiment, the computing devices calculate precisely what, where and how much treatment to apply. In most cases, once the implement has done its job, final data gets sent to the college's central data center on the main campus.
"It's just more cost-effective for us to use an in-house data center," Smoker said. "We looked at cloud computing services, but for the work we were doing, it was too expensive."
The agriculture college also collects data for research using a vehicle, dubbed the Phenomobile, (see photo), which collects data like light reflection and wavelengths. In the context of agricultural research, a plant's phenotype describes observable characteristics that result after the environment has influenced it, such as height, biomass, nitrogen and leaf shape.
Purdue University's Phenomobile collects and analyzes data in the field.
Purdue's IoT infrastructure gives it the ability to automate sensor data collection and stream it to the data center in real time. A trailer housing 12 IoT emissions sensors is also equipped with a Hewlett Packard Enterprise Edgeline server, where the data collected is analyzed and checked for quality before it's sent over the network to the school's data repository. Data collected by the Phenomobile is also streamed directly from the field to the central data center.
"The servers do error-checking to make sure we're sending good data back to the main data center," Smoker says. "For example, if the data doesn't meet a certain criteria, it could be because a sensor may have been hit by lightning."
As intriguing as the technology might be, Smoker said he doesn't want to lose sight of the team's ultimate goal.
"This is about feeding the world," he said. "We don't have much time. By 2050, it's predicted we have to double food production to meet the needs of the world's population."
4 questions about using edge computing in IoT
Before deciding whether edge computing is a good fit, enterprises need to consider four major questions, according to Michael Tennefoss, vice president for strategic partnerships at Aruba, a Hewlett Packard Enterprise company. While the use of edge computing, cloud and in-house data centers can be complementary, the use of any option should help meet a company's business goals.
Companies shouldn't feel rushed to move everything to the cloud or decide all processing must now take place on the edge, Tennefoss advised. Case in point: Manufacturers that run multiple plants around the world may want to incorporate edge computing on the factory floor to track production defects, but aggregate data in the cloud or in an in-house data center when they want to illuminate how their plants are performing compared to one another.
Among the key questions to consider:
How fast does the company need the data?
A production line processing 1,000 bottles per minute requires fast detection and response times to identify defects in real time. Compare this environment with municipal wastewater processing, where fluid levels change over a longer period of time. Enterprises should determine how fast a response time they need at the edge and whether they can accept the latency associated with processing that data at a cloud facility farther away. Sometimes, companies blend both edge and cloud compute. For example, if a company has multiple plants and wants aggregated performance data across those facilities, it could use edge computing to support high-speed decisions and process the aggregated data in the cloud.
To what extent can the company tolerate a WAN outage?
For companies where an outage would have severe effects, using edge computing to process data close to where it's generated makes sense. Other enterprises could tolerate outages without much impact. Enterprises that value uptime and want to use cloud computing will want to consider a network design that encompasses redundant WAN links, using a mix of cellular, internet or satellite connectivity.
What can the company afford?
Data transmission costs can rise quickly, especially if high-cost services -- such as satellite communications or cellular -- are in use. An alternative is to process data at the edge and send a compressed and aggregated data dump via an internet connection at the end of each day or at some other time that's more cost-effective.
Does company policy require certain data to stay local?
In many manufacturing and defense applications, or in situations where companies are dealing with sensitive intellectual property or trade secrets, the data may have to stay local. As a result, companies may want to keep all their data at the edge or only send relevant summary data sets to a central data center over a secure private cloud. For security and compliance reasons, an all-cloud-computing option may never make sense.
