However, this ‘cloud’ only model have been found to be sub-optimal for some of the Industrial IOT (IIOT) use cases having these requirements:
- Adhering to latency budget for critical industrial applications
- Avoiding unnecessary exposure and migration of data to remote datacenters
- Adhering to regulatory compliance
- Bandwidth constraints for sending entire data to cloud.
These IIOT requirements paved the foundation for a ‘hybrid’ model, where computing is shared between edge devices (which are closer to the source of data) and cloud resources. The ‘hybrid’ model has been called with various names – ‘edge-to-cloud’ by GE for their Predix platform or ‘fog computing’ by Cisco. Note that the public cloud vendors are also adding new capabilities to their IOT offerings to support this model. One example is AWS Greegrass offering.
Typically, in the ‘hybrid’ model, the most latency sensitive data are analysed on the nodes closest to the devices generating the data. Less latency sensitive data can be either processed at aggregator nodes present on the network ‘edge’ or sent to cloud for analysis and storage based on defined criteria.
While ‘hybrid’ model does meet some of the IIOT requirements, there are still practical challenges today for doing advanced analytics leveraging machine learning at ‘edge’ devices. This is primarily due to the complexity and resource requirements for the machine learning models. Deep learning on resource-constrained devices is seeing significant research and industry focus in recent times driven primarily by these use cases.
A practical and cost-effective solution to leverage advanced analytics with machine learning on the ‘edge’, for IIOT, is the next frontier in the evolution of IOT space. Who knows we might see a Predix like platform leveraging machine learning right at the ‘edge’ via a combination of hardware and software innovation!!
Please do share your thoughts or questions in the comments section. Would be very interesting to hear and learn on this exciting space from experts in the field.