Derive GitLab Team member badge from gitlab-com group
What does this MR do?
The current way of determining whether a user is a GitLab employee is done by checking the domain of the user's primary email.
This MR determines the GitLab Team member by checking if the user belongs to gitlab-com group. Internally, it uses the similar cache mechanism defined for Performance bar that checks if a user belongs to a certain group.
The performance analysis for this approach is listed below:
Related to !28837 (comment 328910484)
Mentions #216802 (closed)
Analysis tool
The tool used for the analysis is GitLab Performance tool (GPT).
Since the GitLab employee badge will be referenced extensively in issues and MR, the following web tests were performed as part of this exercise:
Test result data
Analysis
- Project Issue:
| Metric | Comments | Data |
|---|---|---|
| TTFB Average | Almost identical for both approaches. On an average, revised approach performed better. |  |
| Breaking point (BP) | For both approaches, the breaking point was 60s - 100rps |  |
| RPS result | RPS for requests were under threshold upto BP. Post BP, the revised approach performed negligibly better (0.34s). |  |
| Overall Result Score | Upto BP, the ORS was above the desired value of 90%, after which it decreased drastically (below 45%). Revised approach superseded by a low margin (3.4%). |  |
- Project MR:
| Metric | Comments | Data |
|---|---|---|
| TTFB Average | Almost identical. Existing approach performed better on an average. |  |
| Breaking point (BP) | For both approaches, the breaking point was 20s - 20rps |  |
| RPS result | Below min. threshold for all tests. Difference got significantly higher with increasing requests. Low Standard deviation (0.2605). |  |
| Overall Result Score | All tests scored below the desired value of 90%. On an average, revised approach faired better. |  |
Inference
- In comparison, revised approach performance was slightly better, not significantly
- Both approaches perform poorly with increased requests
- Project issue api performed significantly better than Project MR api in all metrics
Edited by 🤖 GitLab Bot 🤖