Among the biggest challenges of the Industrial Internet of Things (IIoT) is the installation, management, and integration of an unprecedented array of data-producing connected devices. The increased loads of data created by the IIoT’s proliferation of connected devices places a great strain on industrial networks.
To date, several technologies are being used to ease this burden. For example, MQTT (message queuing telemetry transport), an industrial communication protocol which uses a publish/subscribe transmission paradigm to limit the volume of data exchange on a network, has significantly reduced bandwidth usage. In addition, the rise of more sophisticated simulation technology—often driven by artificial intelligence (AI) and machine learning—has allowed for explosive advances in the prediction and forecasting of issues that may arise within networked systems as they become more complex. This applies in many areas, including: Supply chain management, in which parallel planning for various possible scenarios has been made possible by virtual optimization models; predictive maintenance, where machine learning algorithms trained on large quantities of data can help operators anticipate equipment failures before they occur; and in the use of interactive digital twins that can help plant managers spatially optimize their facilities.
Now, the combined capabilities of simulation and MQTT are being extended to the testing of large-scale IoT networks via the use of Swarm, a new offering from HiveMQ, a provider MQTT broker software. HiveMQ’s Swarm is a distributed platform that can create hundreds of millions of virtual network connections that simulate devices, messages, and MQTT topics. From there, the Swarm software is able to develop scalable and reusable scenarios that mirror real-world device behaviors.
According to Mary Brickenstein Ofschen, technical writer at HiveMQ, the custom scenarios created by Swarm allow end-users to scale the number of devices involved in a simulation by changing simple parameters and observing how individual assets will behave within the overall system as a result. This can help determine how to distribute a load intelligently across a network for the purpose of stress testing.
“For many companies, it is very important to understand exactly how their current development plans impact the entire system. For example, what will the backend infrastructure look like if the number of products sold each year doubles. The same is true for the industrial IoT space. Let’s say you want to expand your gateway infrastructure, add a new SCADA system, or increase the primary and secondary applications in your Sparkplug space,” Ofschen said. “You need to discover the potential bottlenecks in your system, but it’s hard to do. The ability to build reusable, declarative scenarios that HiveMQ Swarm provides [means that you] can build scenarios that truly match your planned use case, put the scenarios on version control, and copy and modify your testing parameters as needed. You can even build your own library of different test scenarios and use them to automate the testing of your entire solution.”
Swarm features multi-cloud functionality and can also be deployed on a local machine with no additional requirements. Currently, the software supports up to 10 million MQTT connections. In addition to its simulation capabilities, Swarm offers built-in monitoring, logging, and reporting, as well as REST API compatibility, making it possible to integrate with a centralized infrastructure.