Edge Computing 101: Requirements for IIoT Edge Computing Solutions

Edge Computing 101: Requirements for IIoT Edge Computing Solutions

There are many companies out there that promise to deliver “edge computing” for IIoT operations. But what does “edge computing” really mean and what is its underlying value?

Why applying edge computing solutions

By applying edge computing to streaming data, manufacturers can meet the real-time demands of high-stakes industries and utilities without straining their bottom line with increased storage or bandwidth costs, while improving the data sent to the cloud. We identify the top two “must haves” for organizations looking to implement edge computing to improve operational efficiency, reduce waste, optimize resources and more.

The Challenge for Industrial Cloud Computing

For most industrial settings, relying exclusively on the cloud to process data is insufficient. Cloud technologies are extremely powerful, yet inherently have heavy footprints and are not designed to process data in real-time. While the cloud is valuable in creating fleet-wide models against many locations’ worth of data, they are costly and ineffective for real-time applications. As an example, by FogHorn’s estimates, the 180,000 oil wells in Texas would spend over $1.49 billion every year to send one terabyte of data to the cloud every day, assuming the storage cost on the cloud was $0.023.

In addition, sending data to the cloud often causes slower processing times and requires more bandwidth to transfer and store it, meaning industrial locations will have to spend more to process the data they need – even if the bandwidth is available, which is often not the case.

Neither of these approaches are viable options for industries that do not have “all-you-can-eat” bandwidth and storage or have access to an uninterrupted connection.

Process data at the edge to minimize data transfer and save money

For many industrial companies, the economics and availability of bandwidth to stream a large (think terabytes, if not petabytes) of individual data points off-premise for analysis is not practical, sustainable or, in some cases, even possible. The easiest example is a very remote location in energy, agriculture or mining that does not have access to affordable and consistent connectivity. For these industrial locations, edge intelligence, meaning analytics, machine learning and larger data processing, must take place on the edge right next to the machine to reduce the amount of raw data processed in the cloud.

True edge intelligence reduces bandwidth demands by an order of magnitude, yet is able to accomplish this with minimal compute footprint. Many so-called “edge computing” systems are still entirely dependent on moving data to the cloud for analysis. This is not true edge computing. Cleansing, processing and analyzing the data locally allows real-time computations to be made and then sends a cleaned-up set of summarized metadata to the cloud. This reduces bandwidth and storage cost, while ensuring better, cleaner data for further cloud analysis.

Edge computing requirements

Real-time decisions with a minimal footprint

True edge intelligence must be able to perform real-time computations on vast, dirty sets of raw OT data. The results must be made available to operators immediately to reduce unplanned downtime and increase production yield. The advantage of placing this intelligence at the edge is computing with minimal latency and the ability to act natively on streaming, time-series data. In other words, these real-time decisions are best made and applied as close to the data source as possible.

This functionality, represented by streaming edge analytics and edge machine learning, must be small enough to fit and run on an IoT gateway or an even smaller device, next to or even on the actual machine. Edge devices are typically resource-constrained, and should be closer in size to a box that fits in your pocket than a server that fits in a closet.

Why value edge computing solutions?

Why are these key tenets to the value of IIoT edge computing? Without them, the edge is simply a dumb layer that gives cloud companies another line item to sell. For example, based on our research, real-time decisions on yield optimization can save a factory $14 million annually, and a wind farm can save over $33 million in hosting and cellular costs by not sending unnecessary data to the cloud. Intelligent data processing and real-time decisions located at the edge represents trillions of dollars in potential industry value.

Edge intelligence approaches that include edge analytics and machine learning are exceptionally impactful for IIoT. They provide faster insights, smaller bandwidth and storage bills, and are designed specifically to handle complex industrial settings. Gracefully turning raw OT data into actionable results and clean summary metadata, true edge computing solutions improve industrial operations – creating staggering amounts of value immediately.

 

This article was written by Matthew C. King. He develops new innovative solutions in emerging technology spaces. Matt is responsible for evangelizing, designing and assuring success in new technologies, working closely with partners and flagship customers to define these categories. Matt brings over five years of technology experience to FogHorn Systems.