Moving from the bare metal to the cloud, the containerized HPCC Systems cloud native applications enable its users to improve their deployment onto the cloud. Kubernetes automation enabled even easier management of the system by treating different aspects within the application as pods and Helm eases the complication of version updates. However, configuring the infrastructure we generate as a result of deployment, debugging and reconfiguring the resources manually can be a daunting task and slows down the deployment, since we have to manually build resources, and virtual machines that are needed for the infrastructure in particular order from scratch. In order to resolve this issue, we employ Terraform and explore the deployment of the HPCC Systems containerized application onto AWS Elastic Kubernetes Service cluster (EKS). During the first part of this internship, we developed a hpcc-aws-terraform module. This consisted of building a necessary AWS infrastructure such as virtual private cloud (VPC), subnets, necessary security group, EKS cluster and node group. AWS VPC with a range of allocated IP addresses allows us to create resources needed to deploy our containerized application in an independent and isolated environment. After the initialization of AWS VPC, we can create an AWS EKS cluster to run node instances which allows us to deploy the HPCC Systems application. For the second part of the internship, we analyze the performance of the infrastructure by running terasort. We will pay attention to how different node instance sizes affect the cost of AWS service. This project not only enables the seamless deployment of HPCC Systems containerized application to AWS EKS, next to Microsoft Azure, The analysis provides insights on how different instance types affect the cost of running the application with AWS.

Presentation

In this Video Recording, Hiroki provides a tour and explanation of his poster content.

Automation of HPCC Systems Cloud Native Deployment to AWS with Terraform

Click on the poster for a larger image.