Designing an Architecture for Edge Success

infrastructure bridge
Dell Technologies

As the volumes of data from connected devices or the Internet of Things continue to grow, IT leaders are finding the need to rethink their strategies and architectures for data processing and analytics. In particular, they must consider the relative advantages of processing and analyzing data at the Edge, as opposed to doing the work in the cloud.

According to a new report based on an extensive survey conducted by S&P Global, 451 Research titled “Designing for Edge and IoT Success,” choosing where to run the workloads is a complex decision for most enterprises. Additionally, as per the report, “Edge and IoT data should not be framed as an ‘edge vs. cloud’ topology; rather, the two should work in tandem where it makes the most sense for the purposes of performance, security, reliability and cost.”

So how do enterprises make these determinations on workload placements effectively? As some of the key levers impacting workload location, the report highlights the following:

  • Application requirements — The requirements of real-time applications favor Edge computing over cloud computing, given the latency inherent in wide area networking (WAN) connections, if available, to the cloud, the firm found. In the manufacturing sector, for example, Edge and near-Edge use cases might include production monitoring and condition-based maintenance due to time sensitivity.
  • Cost — A key cost driver for cloud compute IoT deployments is the frequency of messages between the cloud and the end asset. If these are simple keep-alive messages, triggers or rules, they add up to increased traffic, compute and cloud storage, which drives up the cost of data transport. In its report, 451 Research cites the example of a manufacturing use case where
    the wide-area circuit costs for sending production-monitoring data to the cloud are a recurring fee that exceeds the cost of the edge hardware, support and electricity.
  • Network connections — The availability and speed of network connections for many IoT applications is inconsistent. These network costs were a key lever in many of the use cases detailed in the firm’s report.
  • Data volumes — The volume of data generated, especially from video and radar sensors, can range from tens of terabytes to petabytes. The cost of WAN links, cloud compute and long-term storage of this large datasets heavily impact the five-year costs involved in cloud-centric scenarios, with costs of 7x to 61x over comparable Edge computing scenarios.

For its report, S&P Global, 451 Research examined six use cases in three industries — manufacturing, oil and gas, and smart cities. Edge computing was found to be less expensive over a five-year period than comparable cloud-based implementations in these use cases. The report notes that these industries and applications were heavily skewed toward real-time and other production applications, which demand more than casual, low-frequency applications in less real-time industries.

“Two key drivers to success in IoT projects are a clear understanding of the intended outcome(s), which requires understanding of the applications and workloads themselves, and a firm grasp of the costs involved in delivering against these outcomes,” the report notes. “This serves not only IT teams that want to grow internal skills but gives business units and operational leaders the data they need to use in refining and modeling business operations.”

The bottom line

S&P Global, 451 Research concludes that while there is no “one size fits all” solution or design for an Edge and IoT deployment, the use cases presented in its paper heavily favor Edge computing over a five-year period from a cumulative cost basis. Other use cases that are less data-intensive or cost-prohibitive for storage or bandwidth could also favor deployment of edge computing.

“Ultimately, network designs will have edge compute and storage paired with cloud-based AI models and integration with cloud-hosted applications such as manufacturing historians or CRM,” the firm notes. “It will be the combination of these two, the edge and the cloud, leveraging each when best suited to a particular use case or environment, that will prove the IoT topology of the future.”

To learn more

For the full story, see the S&P Global, 451 Research paper “Designing for Edge and IoT Success.” To explore 5G-enabled Edge solutions, and learn how different industries are putting them to work to drive measurable outcomes, visit DellTechnologies.com/Edge and Intel.com/Edge.

Copyright © 2020 IDG Communications, Inc.