WHERE IT’S ALL COMING FROM
Where is all the edge and IoT coming from these days? “Every process can be sensored, monitored, analyzed, and improved,” said Wright. “All physical processes can also be digital twinned to store and repeat best performance runs, model certain use cases, and allow greater autonomous or remote control.”
Among the many sources are “appliances, meters, home tech, wearable technology, fleet vehicles, rideshare vehicles, including bikes and scooters,” said Kanwar. “There are also many kinds of industrial processes and sensors—control systems— producing data. We’re also seeing the expansion of IoT data from smart city infrastructure, including energy-consuming venues and buildings, and energy generation infrastructure such as windmills and solar farms.”
Sensors alone “record empirical data about the environment, manufacturing process, logistics and movement, and many other measurable, quantitative qualities,” said Currie. “Imaging and other nonstructured data are also gathered. Unstructured data often requires correlation with the structured sensor data. For example, images of objects during a manufacturing process are analyzed for defects. Simultaneously, data is collected from manufacturing control systems on the rate of movement, so a potential defective object’s location is known so it is automatically removed from a fast-moving manufacturing process.”
Above all, “surveillance cameras are by far the number-one source,” said Nilsson. “There are almost 100 million cameras in use in the U.S. alone, and roughly 1 billion globally. Each camera can generate roughly the same amount of data as the average Netflix movie every day. That’s a lot of video to process—vastly more than anyone could watch in a lifetime, produced every single day.”
DATA, UNTOUCHED
The challenge is that many companies don’t necessarily have a grasp on all the data flowing in from the edge and IoT. While it holds great potential, “often this data may be brought into a central hub but just sits and accumulates, not used to build insights or take action,” said Jain.
Call it data deluge—which “continues to be a problem, especially when raw data from IoT devices is flowing to data centers in the cloud for processing,” Lewis cautioned. “Many organizations collect IoT data and then have no idea what to do with it. Organizations need to spend time designing data collection, filtering summarization and storage strategies so that the data that is sent to the cloud is only what is needed, and that data is not stored beyond its utility, especially if stored in raw form.”
It’s not just a technical issue—the importance of edge and IoT data needs to be recognized at the business level. “This needs to be reimagined as a data problem,” said Currie. “Once the challenge is understood as a data problem, then the available IoT data will have a purpose. Data management is very important, as actionable data is used at the edge with aggregated or meta sent on for centralized analysis.”
In the case of video surveillance, for example, “I would estimate that 99% or more of all data is transferred over a network, stored for somewhere between 30 and 90 days, and then deleted without ever being watched or analyzed,” said Nilsson. “Some of that data will be fed into various analytics, but most companies probably do not have a full grasp on all of the specific data being gathered.”
Making data actionable to the business also requires investments, as the delivery of value through edge and IoT data requires “efficient processing and storing various data flows, while protecting the sources and data,” Currie continued. “Security is a particularly interesting challenge, as IoT devices that are affecting operations have never been exposed to corporate networks or the internet now have a new area of exposure to mitigate.”
Along with data infrastructure challenges, edge and IoT computing introduce more unstructured data, as well as data quality and integration challenges. This makes “successfully implementing the transformational initiatives at the edge an uphill task,” said Kanwar. “Autonomous cars, for example, rely on data to differentiate between vehicles and pedestrians when operating. Situations like these depend heavily on core data from a central source, such as location maps or a driver’s portable profile. The precision of the outcomes at the edge depends heavily on the quality of the core information that is universally shared across the entire service. Putting data at the edge doesn’t necessarily improve outcomes—data still needs to be high quality and accurate.”
While internet, cloud, and storage fees may give companies a sense of how much data is actually flowing into their systems from the edge and IoT, analysis needs to be stepped up to better determine its business value, said Wright. “What some companies do not grasp from IoT is the ability to fully use the data to gain information and intelligence. This requires a purpose-driven analysis of the data and possible modification of collection as insights are gained. Having more data collected allows future machine learning insights to be gleaned from anomalies of failures. Conversely, generating purpose-driven intelligence may allow a reduction of different sensor data that ultimately points to the same information. At the beginning of IoT, most of the intelligence obtainable was abandoned for simple remote control. In some cases, this has not changed much.”