Book Image

IoT Edge Computing with MicroK8s

By : Karthikeyan Shanmugam
Book Image

IoT Edge Computing with MicroK8s

By: Karthikeyan Shanmugam

Overview of this book

Are you facing challenges with developing, deploying, monitoring, clustering, storing, securing, and managing Kubernetes in production environments as you're not familiar with infrastructure technologies? MicroK8s - a zero-ops, lightweight, and CNCF-compliant Kubernetes with a small footprint is the apt solution for you. This book gets you up and running with production-grade, highly available (HA) Kubernetes clusters on MicroK8s using best practices and examples based on IoT and edge computing. Beginning with an introduction to Kubernetes, MicroK8s, and IoT and edge computing architectures, this book shows you how to install, deploy sample apps, and enable add-ons (like DNS and dashboard) on the MicroK8s platform. You’ll work with multi-node Kubernetes clusters on Raspberry Pi and networking plugins (such as Calico and Cilium) and implement service mesh, load balancing with MetalLB and Ingress, and AI/ML workloads on MicroK8s. You’ll also understand how to secure containers, monitor infrastructure and apps with Prometheus, Grafana, and the ELK stack, manage storage replication with OpenEBS, resist component failure using a HA cluster, and more, as well as take a sneak peek into future trends. By the end of this book, you’ll be able to use MicroK8 to build and implement scenarios for IoT and edge computing workloads in a production environment.
Table of Contents (24 chapters)
1
Part 1: Foundations of Kubernetes and MicroK8s
4
Part 2: Kubernetes as the Preferred Platform for IOT and Edge Computing
7
Part 3: Running Applications on MicroK8s
14
Part 4: Deploying and Managing Applications on MicroK8s
21
Frequently Asked Questions About MicroK8s

Creating a Kubeflow pipeline to build, train, and deploy a sample ML model

In this section, we will be using the Fashion MNIST dataset and TensorFlow’s Basic classification to build the pipeline step by step and turn the example ML model into a Kubeflow pipeline.

Before deploying Kubeflow, we will look at the dataset that we are going to use. Fashion-MNIST (https://github.com/zalandoresearch/fashion-mnist) is a Zalando article image dataset that includes a training set of 60,000 samples and a test set of 10,000 examples. Each sample is a 28 x 28 grayscale image with a label from one of 10 categories.

Each training or test item in the dataset is assigned to one of the following labels:

Table 9.1 – Categories in the Fashion MNIST dataset

Now that our dataset is ready, we can launch a new notebook server via the Kubeflow dashboard.

Step 1 – launching a new notebook server from the Kubeflow dashboard

You can start...