

A debate has developed inside IT over whether Internet-of-Things (IoT) processing should take place mostly in central data centers, mostly in the cloud or at the edge where the data is generated. At the recent HPE Discover 2016 conference, Hewlett Packard Enterprise Co.’s Dr. Tom Bradicich and National Instruments Corp.’s Eric Starkloff appeared on theCUBE to present their argument that most processing will be at the edge for seven reasons (see video below). The two companies have an obvious agenda – they jointly announced a partnership with General Electric Co. to provide turnkey IoT edge processing systems based on HPE’s new family of Edgeline IoT converged systems.
That, however, does not invalidate their arguments, writes Wikibon Research Director Peter Burris. Wikibon agrees with their reasoning and comes down on the side of edge computing for most IoT processing.
IoT, Burris writes, bridges the information technology (IT) and operating technology worlds that until now have been kept separate for various reasons. This has several implications. Among them is that the analog operating world generates essentially infinite data, Starkloff said on theCUBE. How much data is digitized depends on how often the continuous sensor stream is sampled.
The seven arguments they presented for IoT at the edge were:
These are all practical issues that dictate that most processing, particularly by time-sensitive applications, is as close as possible to the sensors generating the data. This is particularly true when those locations are isolated (such as ocean drilling rigs), demand microsecond response (as in driverless vehicles) or are user-facing, such as sensors that monitor the health of patients with chronic conditions.
In the full Professional Alert, available to Wikibon Premium subscribers, Burris discusses and validates each of these arguments and looks at some of the larger implications of the combination of the IT and OT worlds.
THANK YOU