Bringing Intelligence to the Edge with IoT Gateway

by | Feb 7, 2018 | Edge Computing

In the previous blog, my colleague David Bericat discussed why Internet of Things (IoT) architecture should be built with open source. One of the core components of end-to-end IoT architecture listed in that article was an intelligent IoT gateway that can process data near its source in near real time and filter/prioritize the actionable data. In this article, we’ll explore the reasons behind the need for an intelligent IoT gateway.

Enterprise-Grade IoT

It’s important to realize the differences between consumer IoT (used for personal or home monitoring/control) and enterprise or industrial IoT (used for managing critical infrastructure, improving productivity/safety or enabling a new product or service). Enterprise IoT requires a fundamentally different approach than the fully integrated, black-box approach used for consumer IoT solutions.

Deploying industrial or enterprise-grade IoT at scale within a company presents several unique challenges:

Amount of device data: Instrumenting a factory or smart city will generate a huge amount of data. In this 2014 article, Sham Chotai from GE talks about a single blade in a gas turbine being able to generate 500GB data/day.
The device-to-cloud model, where all sensor data flows to the cloud for processing, does not scale effectively as the costs can rise pretty dramatically. This doesn’t even include the transmission cost over pricey cellular networks.

Fast vs. Slow data: All data from devices can’t be treated the same way. Non-critical data (slow data) about environmental conditions doesn’t need to be treated with the same urgency as critical data (fast data) that needs to be acted upon urgently. Slow data may need to be processed (via aggregation, transformation, or summarization) further at the edge before being routed to various applications or databases.

Business continuity: If core business functions are dependent on the cloud, then losing cloud connectivity will severely affect business functions. Consider the case of smart parking, where cloud connectivity failure or degradation could cause parking mayhem in a city. This requires basic parking services to continue functioning through local computing capabilities independent of access to the cloud.

Time sensitive: For IoT use cases related to critical infrastructure, the latencies associated with transmitting data to the cloud are not acceptable because decisions need to be made in near-real time. Even more serious are situations where critical decisions need to be made within a very small timeframe, e.g., inclement weather causing internet reliability issues while there is a transformer failure that could take the whole electrical grid down.

Securing devices: The sensors/actuators are constrained devices that generally lack the security mechanisms of more robust systems. Bringing them online without adequate protection is a recipe for disaster. However, it’s not easy to harden these fixed-function devices without significantly changing their footprint (cost, power envelope, real estate).

Legacy devices: Some of the industries’ infrastructure (e.g., oil rigs, factories, and office buildings) have been getting instrumented over the last several decades. Any IoT solution will need to connect with these devices. The existing sensors, actuators, and other devices often use fieldbus protocols (e.g., Modbus, Profibus, BACnet) that may not easily connect with modern back-end systems. It may be economically infeasible to replace or upgrade these legacy devices without causing business disruption.

Integration with existing infrastructure: The IoT data will need to integrate with existing systems, databases, applications, and services across a heterogeneous IT infrastructure. This requires capabilities to transform and route the IoT data as and when needed by various systems.

Intelligence at the Edge

One of the ways to solve the challenges mentioned above is by adding another layer to the architecture between the end-devices (sensors, actuators, and other embedded systems) and the back-end. This approach in IoT architecture moves data processing and decision making closer to the data source. This middle tier helps brings intelligence towards the edge and acts as a gateway between the other two tiers. The implementation of this middle tier can be provided by a system known as an intelligent IoT gateway.

The intelligent IoT gateway is an important component of industrial IoT, as it can process data near its source in near real time and filter/prioritize the actionable data. Business-critical functions can also continue even with the loss of internet/cloud connectivity. The device traffic can be separated for each segment, lowering the network bandwidth requirements and costs. The intelligent IoT gateway can act as a bridge to the legacy devices through fieldbus protocol adapters, thus bridging the operations technology (OT) and information technology (IT) worlds. This allows IoT data to be integrated with existing IT systems and enable companies to better leverage their existing IT infrastructure.

The intelligent IoT gateway also serves to aggregate the data from end devices–this includes collecting, transforming, and summarizing the data stream into useful chunks. Instead of all the sensor data, a small subset of data and, more importantly, the actionable data is sent to the cloud. The data can also be prioritized and routed to when and where it’s needed, such as critical alerts that must be prioritized and routed to a different endpoint than the rest of the data. The additional value-added services, like the aforementioned smart parking use case, can include demand-based pricing, video analytics, and guided parking, all provided through cloud infrastructure. Being close to the data source, the intelligent IoT gateway can process information in near real time and do away with the latencies associated with the cloud–a key requirement for critical use cases.

Through the use of open source solution like Eclipse Kura on an intelligent IoT gateway, developers access IoT device data without worrying about the underlying hardware interfaces or device drivers. Eclipse Kura also provides connectivity to an IoT integration hub, which manages devices and their data using open source solutions like Eclipse Kapua.

The middle tier also serves to help secure the downstream devices (sensors, actuators, and other embedded systems) from the security risks posed by the public internet. This middle-tier firewall protects the downstream devices by leveraging security best practices like SELinux and container policies. The software stack on these intelligent IoT gateways can be maintained, patched, and upgraded years after the initial deployment. Best of all, these systems can be provisioned at scale using existing IT infrastructure.

Conclusion

The intelligent IoT gateway is a key requirement for industrial IoT, as it can bridge the end devices and back-end systems (whether on-premises or cloud). By bringing intelligence closer to the data source, the actionable data can be acted on in near real time (a must for life-preserving safety systems). The business-critical needs can continue to function during the loss of internet/cloud connectivity. Network bandwidth is conserved as the data is summarized/aggregated before sending it to the back-end.

The middle tier also provides a firewall to protect the end devices from the threats associated with being online. Finally, businesses can better leverage their existing OT and IT infrastructure through the intelligent IoT gateway. This provides businesses more control over their IoT deployments.