Deploying Analytics on an IoT Platform

In the previous chapters, we looked at the differences between different types of analytics. We implemented some examples of prognostic and diagnostic analytics in the world of the I-IoT. We also studied how to deploy our analytics on the most common platforms and how to use open source technologies.

In this chapter, we will finalize our exercise by delivering the algorithms developed in Chapter 14, Implementing a Digital Twin – Advanced Analytics. In particular, we want to highlight the major differences between three platforms: AWS, Azure, and GCP. We will discover that all three platforms adopt the same principles of providing a computational infrastructure for training and a service-oriented platform for using the analytical model.

In this chapter, we will explore the following topics:

• Deploying diagnostic analytics using the...

Technical requirements

In this chapter, we need the following prerequisites:

• Git: https://git-scm.com/downloads
• Python 3.7: https://www.python.org/downloads/
• Anaconda 5.2: https://www.anaconda.com/download/
• Docker: https://www.docker.com

For the complete solution to the exercises, please go to the repository at https://github.com/PacktPublishing/Hands-On-Industrial-Internet-of-Things.

Working with the Azure ML service

The Azure ML service is a service to train and deliver a model as a containerized application. When we have built the model, we can easily deploy it in a container such as Docker, so it is very simple to deploy to the Azure Cloud. The Azure ML service can work in collaboration with Azure Batch AI, advanced hyperparameter tuning services, and Azure Container Instances.

The Azure ML service is different from the Azure ML Studio. The Azure ML Studio is a collaborative visual workspace where we can build, test, and deploy analytics without needing to write code. Models created in the Azure ML Studio cannot be deployed or managed by the Azure ML service.

The basic steps to develop our analytical model with the Azure ML service are as follows:

1. Preparing the data
2. Developing the model with a rich tool, such as Jupyter Notebook, Visual Studio Code,...

Implementing analytics on AWS SageMaker

AWS SageMaker is a fully-managed service that enables data scientists to build, train, and deploy ML models at any scale. AWS SageMaker is based on Jupyter Notebook, so that developers can use a familiar user interface to build their own analytics. The basic concepts of SageMaker are the same as Azure ML. We can build our analytics on Jupyter and our training cluster through a Python API, and then deploy our model as a web app that can be consumed through a REST API. SageMaker also supports built-in algorithms to train our model. These include K-Means, K-Nearest Neighbors, Linear Learner, Neural Topic Model (NTM), Principal Component Analysis (PCA), and Random Cut Forest.

Evaluating the remaining useful life (RUL) of an engine with SageMaker

...

Understanding the advanced analytics capabilities of GCP

GCP is very focused on the machine learning algorithm and provides a huge variety of technologies to build an analytical model. These include BigQuery, Dataflow, Data Studio, Prebuilt-Model, and Engine ML. GCP has also developed its own Tensor Processing Unit (TPU) processor to speed up the adoption of ML both on the cloud and on the edge (https://cloud.google.com/edge-tpu/). GCP supports ML through the Google Cloud ML service (https://cloud.google.com/ml-engine/).

ML Engine

The process to develop a model using the ML Engine is quite similar to that we have seen in the Working with the Azure ML service and Implementing analytics on AWS SageMaker sections:

1. Develop the...

Discovering multi-cloud solutions

Choosing the right cloud for our platform is a complex and risky job that is likely to have a significant impact on our business. One option is to use inter-cloud communication, or Docker, Jupyter, and Python, to be as agnostic as possible with regard to the cloud. There are, however, other interesting multi-cloud solutions. We will explore these in the following sections.

PyTorch

If we are working with Python, deep learning, and GPU, we can get interesting benefits from PyTorch (https://pytorch.org). PyTorch is supported by Azure, AWS, and GCP. It can also be installed on-premises.

Chainer

...

Summary

In this chapter, we discovered how to deploy analytics on Azure ML and AWS SageMaker, using the algorithms developed in Chapters 13, Understanding Diagnostics, Maintenance, and Predictive Analytics, and Chapter 14, Implementing a Digital Twin - Advanced Analytics. We looked at the differences between the deployment methodologies of Azure, AWS, and GCP, and we also explored the new trend of IoT Analytics.

This chapter is the last chapter of our journey into Industrial IoT, which we started in Chapter 1, Introduction to Industrial IoT. We began by looking at the differences between the IoT and the I-IoT. The first four chapters talked about the most important data sources in the industrial sector and the differences between them. We looked at the OPC UA and learned how it is the new upcoming standard in the industrial sector. We also understood the differences between data...

Questions

1. What is a Jupyter Notebook?
   1. An IDE in which to develop Azure ML or SageMaker analytics
   2. A general-purpose interactive IDE for Python and other languages
   3. A notebook used by NASA
2. Which one of the following steps is more appropriate to deploy analytics on Azure ML, SageMaker, or GCP Analytics?
   1. Prepare the data, train the model, test the model, deploy the model
   2. Prepare the data, build the model, deploy the model, monitor the model
   3. Prepare the data, build the model, (train the model, test the model), deploy the model, monitor the model
3. What is the basic idea of SageMaker and Azure ML?
   1. To build the model as a web application on a containerized application
   2. To learn the model using FPGA and GPU
   3. To allocate a computational cluster in which to test the analytics

Further reading

Additional resources can be found at the following links:

• AWS SageMaker: https://aws.amazon.com/sagemaker/
• Predictive Maintenance with Greengrass and AWS IoT: https://aws.amazon.com/blogs/iot/using-aws-iot-for-predictive-maintenance/?nc1=b_rp
• RUL with Azure ML: https://docs.microsoft.com/en-us/azure/iot-accelerators/iot-accelerators-predictive-walkthrough
• Azure ML with GPU: https://azure.microsoft.com/it-it/blog/what-s-new-in-azure-machine-learning-service/