Somshubhro Pal Choudhury, MD, Analog Devices India, recently told the Financial Express that directly transporting sensor data to the Cloud via a gateway is little more than a recipe for disaster.
“This leads to a surge in OPEX. The CAPEX investments will also have to be made at the same scale. The radio that enables wireless data transmission has to be always on, which will abuse the battery capacity,” he explained.
“IoT is all about creating a ‘sensor to the cloud’ system, which is solving a business problem at a good ROI. On the contrary, we are spending more energy transmitting wirelessly compared to doing computing at a given node. It requires a fat pipe.”
The right way, says Choudhury, is to execute the initial analysis on premise and send only the relevant data to the Cloud. It should be noted that TechTarget’s Alan R. Earls expressed similar sentiments in a recent article titled “Edge computing becomes ‘Wild West’ of IT world.”
“End users don’t necessarily care where and how data is stored and processed. They just want to know the job will get done quickly, securely and with as few difficulties as possible. Oh, and inexpensive, too,” he wrote. “With so much data coming from or going to the edge, schemes are afoot to do more processing there, to extract value from the data sooner and to reduce the volume traveling over scarce network bandwidth.”
As Sylvain Fabre, research director at Gartner, points out, there can be significant advantages to edge computing from an industrial or corporate perspective. More specifically, says Fabre, processing data locally translates into faster results, while reducing data transport, with only “results,” rather than raw data, likely to be sent to a central location.
Another important angle to consider is security, says Earls, as processing data locally on an edge node means it is kept in an internal environment. Patrick Gill, a Principal Research Scientist at Rambus, concurred with Earls’ assessment.
“OEMs can [now] build simpler sensing nodes that compute sensing results in place, avoiding the security issues of transmitting the user’s potentially sensitive data to a remote server,” he explained.
One example of a simple-function, sensor-laden endpoint where initial processing occurs on the edge is Rambus’ lensless smart sensors (LSS).
“We realized that if LSS is going to function in IoT nodes independent of an Internet connection [during inevitable downtime], we needed to reduce its computational requirements dramatically,” Gill explained. “So we optimized our optics for computationally light functions, and in doing so we’ve derived a system so resource-light that one of our team members has ported image reconstruction to an Arduino Due.”
With a minimal class of computational requirements, says Gill, OEMs can build simpler sensing nodes that compute sensing results in place – avoiding the security issues of transmitting the user’s potentially sensitive data to a remote server.
“Put simply, the less IO in IoT, the more robust the T will be,” he concluded.