[Task]: Identify LOE for Supporting AWS QuickSight vs Hosted BI Tool

[acouch]

Summary

We want to identify the level of effort for supporting AWS QuickSight vs a self-hosted tool like Metabase or Apache Superset. Both of those tools use Postgres for user / data storage and have supported docker containers. We would use the same ECR/S strategy we use for the frontend and backend. We will not need dev/stage versions.

For QuickSight, LOE (in points) should include:

• User permissions that will need to be added and configured
  • Some will be needed to share dashboards publicly, some for adding non-dev users
• Terraform implementation
• Logging / monitoring if needed
• Other security issues

For self-hosted tool:

• Setup of terraform to support ECR and ECS hosting
• Setting up Postgres database in AWS
• Logging / monitoring

Acceptance criteria

• Steps for supporting each of the two options are detailed enough to be actionable and comprehensive
• Level of effort in estimated points is applied to each step and overall
• Major questions are surfaced

Notes

• What is the LOE and other trade-offs to consider to set a self-hosted tool vs. something out of the box from AWS?

[coilysiren]

LEO Quicksight: 2 ~ 3 points. It's already setup with user management, you essentially just "turn it on" in the UI and then everything else is simple configuration.

LEO Self Hosted: 16 ~ 24 points, eg. 2 ~ 3 sprints. This includes the time to get the compute running (ECR / ECS / EC2 / whatever it is), the data store that it will need (RDS or otherwise), and figuring out the access control (ie. user accounts, and more importantly the VPN)

[widal001]

Closing because issue was marked as 'Done' in https://github.com/orgs/HHS/projects/13

[coilysiren]

Here is the result of my scoping,

QuickSight: 10 points
Metabase (or anything self hosted): 24 points