iot-ml-deployment

Deploy machine learning model on IOX-Gateway

Sample Code Description:

This sample usecase demonstrates how a pre trained machine learning model can be deployed on the IOX gateway

Modules Description:

1. IOTMLmodel: Linear Regression ML model building from sample IOT dataset - 'Household power Consumption' dataset downloaded from https://data.world/databeats/household-power-consumption . Two different sample linear regression models are built, trained on the sample data and stored as pickle models for deployment on Edge device i.e, IOX Gateway

2. IOXMLApp: Demonstrtaes pre-trained ML models ML can be deployed on a webserver with an REST API to predict the 'Global Active Power' of a household based on 'hour'. Two models 'iotmode.pkl' or iotmodel2.pkl' can be used for deployment by changing the file name

3. sampleclient: Sample rest client application that sends the data in a json format , invokes the API get predicted output and prints the json response on screen (see screenshot below) the ip address of edge ML server and payload is selected in 'sample.data.py'

Technologies:

1. Python3
2. Numpy
3. Pandas
4. Scikit-learn
5. Pickle
6. Flask
7. Requests

Installation and Configuration:

Clone this repo first and follow the steps to run each Component

git clone https://github.com/wiprodevnet/iot-ml-deployment.git

Create a ML model (Optional)

Run the IOTLMLmodle python program to generate ML models. This will create 2 models that needs to be copied to IOXMLApp folder to create the package. Alternatively you could use the prebuild models 'iotmodel.pkl or iotmodel2.pkl'

cd IOTMLmodel

python3 iot_ml_model.py

Build , Package IOXApp for Docker

Build a docker image ,create a package and deploy the package as follows

Build docker image

sudo docker build -t iox-mldeploy-test .

Check built images

sudo docker images

Create a package for deployment

sudo ioxclient docker package iox-mldeploy-test:latest .

Further reference https://developer.cisco.com/learning/modules/iox-basic/iot-iox-app-docker/step/10

Deploy IOXApp for Docker
With the IOT ML docker image for IOX is built and connection to sandbox is established, next step is to deploy to IOXGateway.
More details on deployment refer to
https://developer.cisco.com/learning/modules/iox-basic/iot-iox-app-docker/step/11
https://developer.cisco.com/learning/modules/iox-basic/iot-iox-app-docker/step/12

Check Deployed IOXApp

Sample rest client application sends the data selected in 'sample_data.py'

python3 iotclientapp.py

iot-ml-deployment sample Output:

1.Sample output for single variable ML Model i.e, iotmodel.pkl

2.Sample output for multiple categorical variable ML Model i.e, iotmodel2.pkl

Prerequisites

Previous IOx labs needed to complete this lab
You will need to have completed the learning the following labs:

1. Introduction to Cisco IOx - https://developer.cisco.com/learning/tracks/iot/IoT-IOx-Intro/intro-2-iox/step/1
2. Intro to Containers - https://developer.cisco.com/learning/tracks/iot/IoT-IOx-Apps/Containers-101/step/1
3. Intro to Docker - https://developer.cisco.com/learning/tracks/iot/IoT-IOx-Apps/docker-101/step/1

Docker technologies

You should have a basic understanding on how to build and run a docker application.

Docker Platform

To create a Docker image and push it to the Docker Hub, you will need the correct Docker tools for your platform which are available at https://www.docker.com/products/overview.
Download and install the correct version for your operating system.

IOx Dev Environment As A Service

This lab has instructions for using a docker container that sets up a web service you can access from your browser. The web service is a browser based version of Visual Studio Code, developed by Coder Com (https://coder.com) and utilizes docker to run the service. Since it is Visual Studio Code, it it provides a text/code editor and comes pre-built with ioxclient and everything you need to build IOx applications. You will need the Docker Platform on your desktop or developer machine to run this environment.

Cisco AnyConnect Client

To access the DevNet Sandbox, you will need to use the Cisco AnyConnect Client for Virtual Private Network (VPN) access to the IOx instance. If you need to install it, you can find it at this link.

Note: If you are working on a DevNet Lab workstation, this software is already installed.

ioxclient

You will need the ioxclient to package and deploy the IOx Application. If you need to download it, you can find it at this link.
Note: If you are working on a DevNet Lab workstation or you are using the "IOx Dev Environment As A Service" container, this software is already installed.

git

One way to get the application template code is to use git. We have the option to download the code, so this step is optional. If you are working on a DevNet Lab workstation, the git software is already installed. If you need to install it, you can find it at this link.

Note: You can verify the installation of git by opening a command prompt and running:

git --version

Additional Note: If you are working on a DevNet Lab workstation or you are using the "IOx Dev Environment As A Service" container, this software is already installed.

Once you have verified the prerequisites to complete this lab you will:

0. Reserve an IOx Sandbox instance
1. Setup a Local Container for the IOx Developer Environment
2. Create a small IOx Application to run on the Sandbox IOx device
3. Package the IOx Application with the ioxclient Docker tools
4. Connect to the DevNet Sandbox
5. Deploy the IOx Application
6. Monitor and review the IOx Application
7. Test the output of the IOx Application using the sampleclient

Refer below link for more details

• https://developer.cisco.com/learning/modules/iox-basic/iot-iox-app-docker/step/1