Talking heads: What to consider when matching industrial IoT use cases to edge-cloud setups
Jean-Philippe Provencher, vice president, manufacturing strategy and solutions, PTC
“The governing factors are the facilities, the cost of downtime, and speed of product development.
“Cloud computing could hit a snag with these factors as facilities in lower cost areas may struggle with intermittent or costly network connectivity; unforeseen downtime could impede production and financials if the speed of production is drastically reduced.
“In short, a company that could fall back on paper would survive a network outage better than a company whose processes are real time and streamlined for development.”
Andreas Vogel, adviser, SAP
“There are three drivers for edge computing: data volume and bandwidth capacity, intermittent connectivity, and fast decision making.
“There is so much data generated by the devices that it is not economically viable to transmit it all into the cloud. This is mostly caused by the high sampling rate – 1,000 data points per second, for example – by modern sensors. The solution is to pre-process or filter the data at the edge to reduce the data volume. Alternatively the entire data processing could be done at the edge. In this case, the edge becomes an on-site micro data centre.
“Industrial operations need to continue whether internet connectivity is available or not. Operators use edge computing to ensure continuous operation of their equipment and production lines in case of Internet outages. Global manufacturers often operated in remote areas of the globe where the Internet is somewhat unstable and without redundancy.
“Modern production lines often operate at enormous speed and the time for a decision must match this speed. A round-trip to the cloud which can be hundreds of milliseconds and more might be too slow for certain operations. Edge computing can deliver responses in tens of milliseconds.”
Jesse Clayton, senior product manager, autonomous machines, Nvidia
“Different use cases require different system topologies. In some cases it is better to run AI workloads in the cloud or data centre where there are more compute resources available and it’s possible to aggregate data across multiple sources.
“For other use cases the AI must happen at the edge. In practice we see many organisations adopting a hybrid approach, where some analytics are processed at the edge, and then metadata is sent to the cloud for higher-level analysis.”
Bernd Gross, chief technology officer, Software AG, and chief executive, Cumulocity
“Important factors are quick deployment, scalability and an open platform architecture that facilitates the integration of existing systems. The decisive factor here is maximum efficiency through automated processes.
Some platforms offer integrated monitoring capabilities so devices and machines can be maintained or data analysed in a graphical interface without prior programming. These capabilities facilitate industrial pilot projects. Last but not least, it makes predictive maintenance easier, now a standard application in the industry 4.0.”
A new report and webinar on edge computing in industrial IoT setups, called AI and IoT at the cutting edge – when to move intelligence closer to the action, is available; go here for the report, go here for the webinar.