<< Previous Blog Post Next Blog Post >>
The Edge and the Cloud
In today’s world of complex Industrial IoT applications the cloud and the edge complement each other with functionalities distributed across the edge and the cloud. By cloud, we mean IoT platforms in the cloud.
Many IoT platforms with rich functionalities are continuously moving to the cloud, while at the same time more and more powerful and intelligent edge devices are introduced into the market. Machine learning and Computer vision, considered as a core cloud functionalities are getting deployed at the edge. The lines between the edge and the cloud are blurring.
This brings us to certain key questions.
How to distribute functionalities across the edge and the cloud?
What are the functionalities that should be at the edge?
What are the functionalities that should be part of the IoT platform in the Cloud?
Distributing functionalities across edge and the cloud
Any Industrial IoT application requires a range of functionalities to be implemented. Given below are some points to consider while distributing functionalities across the edge and the IoT platform in the cloud.
Acquiring raw data – Connecting with sensors and field data sources happens in real time at industrial environments like a manufacturing shop floor. Edge devices located in the industrial environment are best suited to do this activity.
Basic processing of Raw data – It is better to convert the raw data into useful information in near real time, so in principle this is also an edge functionality. Dong this at the edge gives the flexibility of transferring only the relevant information to the cloud platform. However, in certain cases where the edge device does not have the data processing capability, the basic processing can be performed in the IoT platform in the cloud. This would mean transmitting all the raw data to the cloud platform.
Aggregating and creating meaningful information packets – This is an edge functionality if the IoT gateway has data processing capabilities, else can be performed by the platform in the cloud. It is recommended to have this functionality at the edge.
Data ingestion from multiple assets – This is an IoT platform functionality in the cloud as it requires connecting to multiple IoT edge devices in the field.
Device management – Managing multiple devices (Provisioning, Authentication, OTA) is done by the IoT platform in the cloud.
Raw data storage – Edge devices can accommodate limited amount of raw data storage (Few days), but raw data storage for longer periods (years) is a cloud functionality. It is recommended to distribute this functionality across the edge and the cloud.
Large scale aggregation of data – This is typically done by the IoT platform in the cloud.
Running basic descriptive analytics – At an Industrial asset level, basic descriptive analytics can be run in an Intelligent edge device. If the descriptive analytics needs to be run on large data sets, it is done in the IoT platform in the cloud. It is recommended to distribute this functionality across the edge and the cloud.
Developing and training complex AI models – This is a cloud functionality as it requires powerful compute, large data sets, and model libraries.
Running AI models – AI models can be run both at the edge with powerful edge devices and in the IoT platform in the cloud. It is recommended to run these models at the edge if real time insights and actions are required.
Analytics cutting across Industrial assets, plants – Since this requires data from multiple entities (Assets, plants), this is a functionality for the IoT platform in the cloud that has access to all these data sets.
Real time analytics and actions – This is an edge functionality considering the extreme low latency requirements.
Visualization – Real time visualization at an asset level can be done at the edge. Complex visualizations across longer periods that require larger data sets needs to be done in the IoT platform in the cloud.
Distributing functionalities across the edge and the cloud depends on the functional and performance requirements of the IIoT applications. The points mentioned above act as a barometer to help you decide.
In the previous blog posts we explored various facets of Edge devices starting with an overview, examined the building blocks, delved into edge computing. In this post we looked at pointers for distributing functionalities across the Edge and the Cloud.
The next blog post is all about AI at the edge ...
Comments