Admin dashboard for benchmark visualization in the APEx Algorithm Services Catalogue

[JanssenBrm]

To enhance visibility into the current status of benchmarks, an admin dashboard should be developed and integrated into the APEx Algorithm Services Catalogue. A Proof of Concept (POC) for this solution already exists parquet-dashboard, currently deployed at apex-algo-dashboard.netlify.app.

Proposed Features

1. Integration with APEx Algorithm Services Catalogue
   Embed the benchmark dashboard directly into the APEx Algorithm Services Catalogue for streamlined access and improved usability.

2. Authentication and Role-Based Access
   Initially, only users belonging to the administrator group (configurable) in Keycloak will have access to the dashboard. They will be able to view benchmarks for all services.

3. Future Enhancements
   In a follow-up iteration, extend access to all APEx users, allowing them to log in and view benchmarks relevant only to their projects or services.

[JanssenBrm]

Steps:

• Create a mockup

Considerations:

• Show more information on the failures
  • Error
  • Date of execution

[rahmandawibowo-vito]

I created an initial mockup design for the admin dashboard here https://drive.google.com/file/d/1-cxzLk9HfuIyv6JWjSPX4LK11ob-D5rN/view?usp=sharing @JanssenBrm

The important part is the user can visualize the data and the charts using a date range filter. Is there perhaps any better way to achieve this than querying from all the parquet files @jdries ?

[JanssenBrm]

@p3dr0 / @HerveCaumont - Feel free to add your feedback to the updated design.

[JanssenBrm]

Update 20/02

• Add to the guidelines (best practices) that a platform is able to provide metrics for the benchmarks.
• Further discussion needed to see how to formalize the structure of the metrics

[JanssenBrm]

• @JanssenBrm to schedule a meeting to further discuss this

[JanssenBrm]

@rahmandawibowo-vito - Let's create our first example of this dashboard based on the suggested design. I believe this is irrelevant of the numbers that are provided by the platforms.

[JanssenBrm]

Feedback 30/10

• @HerveCaumont suggested making the benchmark time window more explicit in the user interface to avoid confusion about the stability status. For example adding "Testing period" label around the date selection to make clear that the status is based on the selected period.
• @Patrick1G proposed adding a column for the last test date and, if possible, reasons for job failures (if possible). The latter one is also available in the GitHub tickets that are created when a benchmark fails.