Jessie Mao is a junior at Lambert High in Suwanee, Georgia.

Poster Abstract

HPCC Systems Platform Cloud Solution is a High-Performance Cluster Computing platform built for high-speed data engineering which supports cloud-based deployments (large or small) such as Azure, AWS, etc or local testing/development such as Docker Desktop. Users normally modify values.yaml or helm install command-line options to configure the cluster before deploying it. Currently, the default values.yaml file in HPCC-Platform provides lots of settings as examples. A drawback to this, however, is that it has too much information, and is not easy to configure, track, and maintain. It can also be inconvenient in some situations. For example, if someone has limited resources, they may not be able to start a default cluster due to its size. Being able to deploy smaller configurations and a combination would help production, training, debugging, and testing. As of now, we provide two solutions: overrides and HPCC-lite with a script that can be used to deploy both. The overrides solution utilizes the default values.yaml file while using other files to modify it. Overrides can be used to make small changes to the values.yaml, and mainly concentrates on Roxie and Thor. An example is nothor-tworoxies.yaml which modifies the default file to deploy only two Roxies without thor or hthor. The HPCC-lite, on the other hand, does not require a custom values.yaml file,so can be used with other files to create more scenarios. This means it doesn’t have the same problem with overrides when it comes to the amount of changes because HPCC-lite is much more customizable. This flexibility is because people are making all the files from scratch. A downside of HPCC-lite, however, is that it requires more work than overrides because all the files are made by the user.

Presentation

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

HPCC Systems Deployment with Various Helm Chart Configurations

Click on the poster for a larger image.