Collecting data on the cost of spot instances across the major Cloud Providers (AWS, Azure)

Objectives

• Create a DWH for staring pricing data from various cloud providers.
• Create orchestration that will pull data from the sources periodicaly.
• Create a dashboard that will show relevant metrics.
• Create a workflow to automaticaly deploy all of this.

Technology used

• Cloud: AWS
• Containerization: Docker with Docker-Compose
• Infrastructure: Terraform
• DWH: Redshift with dbt core.
• Orchestration: Prefect
• Data Transformation: Pandas
• Data Visualization: Metabase

Setup instructions

If you want to run it locally. Local setup

If you want to run it with GitHub Actions GitHub Setup

DWH Database Schema

Data Sources:

• AWS historical pricing.

• Azure price list API.

Tech Diagram

Dashboard

Link to dashboard

Insights

• Lower powered AWS spot instances are often as costly as on-demand. Even when available the savings are much less than typically advertized.
• Bigger instances come with bigger discounts. For example: m5a.large spot instances are 44% cheapter than on-demand. a1.medium spot instances are the same price as on-demand.

To-Do

• Add GCP Data.
• Add persistance for metabase.
• Add data quality tests to dbt flow.

Acknowledgements

Thanks to the instructors:

• Ankush Khanna
• Sejal Vaidya
• Victoria Perez Mola
• Kalise Richmond
• Jeff Hale
• Alexey Grigorev

Thanks to collegues:

• Anna Geller, her articles on Prefect DataOps have been a huge infuence.
• Andy Nelson, for telling me about ZoomCamp and generaly being a great mentor.
• Matt Little, his articles on Terraform and AWS were what made this entire project possible.

Contact information

• LinkedIn