Winning with the IoT: the vitality of edge computing to the enterprise

The Internet of Things (IoT) represents the most transformative technological application of the immediate future in terms of adoption rates and societal impact. Gartner projects 20 billion connected devices by the start of the next decade.

Moreover, the IoT will encompass three fundamental realms of society. They include aspects of personal life (via smart homes and wearables), public life (with smart cities and transportation infrastructure) and professional life (with sensor data applications). In most cases, each device will continuously transmit or receive data via direct internet connectivity.

Given the sheer number of connected gadgets and their significant expansion of enterprise networks, it’s impractical to constantly transmit all that data to centralized locations and then send results back for informed action—especially given the low latency of most IoT use cases.

“For all of that data to go to some centralized location and then return so that instead of being kind of dumb transmitters they actually benefited from this information, that requires a trip to the edge,” says Jack Norris, MapR senior VP of data and operations.

Edge computing is touted as the means of realizing the IoT’s projected impact. It’s a decentralized paradigm in which computational resources and connectivity are extended to the cloud’s edge for real-time applications. The most accessible edge benefits include:

♦ Latency elimination—By processing data at the edge of the cloud where devices are located, organizations bypass the time required to send the entirety of their data to a centralized location (and back). Edge processing occurs where the devices are for real-time results and action.

♦ Reduction of bandwidth—Constantly streaming data back and forth to centralized datacenters presents intense network demands. Organizations must continually pay to increase bandwidth and monitor traffic, yet still risk downtime for overloads. Edge computing removes the need for such bandwidth concerns in part via analytics filtering, in which only analytics results are transmitted to centralized locations. Randy Guard, CMO of SAS, says, “You’re not going to do all the analysis unless you’re catching some at the gateway at the edge, then do smart filtering for what you store in a central environment.”

♦ Use case optimization—By processing data at the cloud’s edge, organizations broaden enterprise intelligence in a timely manner for individual use case optimization. Thus, consumer applications yield higher levels of customer satisfaction; enterprise applications result in greater ROI.

Despite those gains, edge computing’s distributed architecture won’t completely supplant centralized clouds. The latter paradigm will likely persist in the IoT for:

♦ Security—The issue of security is central to IoT adoption rates and the edge in particular since it involves expanding protection beyond comfortable enterprise boundaries. Chet Wisniewski, principal research scientist at Sophos, says, “We’re really reliant on a centralized architecture that most of these devices are using to provide that security element.”

♦ Scale—Depending on how edge deployments are implemented, they may create inherent costs to scale for which organizations are not prepared.

♦ Aggregation—Many data deployments are enhanced by aggregating data for a holistic picture of how they apply to use case, which will almost always require synthesizing data between locations in centralized centers.

Consequently, a merging of the two architectures will almost certainly continue, with edge computing proving essential to the IoT’s adoption.

“Will the central store ever go away?” asks Nima Negahban, CTO of Kinetica. “No. But will a lot of the aggregates that are being conducted happen at the edge, and then just the aggregates be shipped up, so the central store just has a collection of aggregates? Yes.”