Load Testing

Test carried out on the theme "Car", in the context of a car event with 500 people interested in Authentify It technology over a short timeframe

Load testing is an essential practice to evaluate system resilience under high user demand. However, when hosted on DigitalOcean, our infrastructure benefits from automatic Cloudflare DDoS protection.

While this protection ensures strong security against malicious traffic, it also introduces limitations for synthetic load testing, as Cloudflare automatically filters or rate-limits what it interprets as potential DDoS activity.

As a result:

  • Load testing from external sources may not fully reflect real-world scalability.

  • To obtain accurate results, tests must be coordinated with Cloudflare’s security settings or executed through controlled environments and whitelisted IPs.

  • Despite these constraints, the system benefits from continuous monitoring and stress testing in staging environments, ensuring performance under realistic conditions.


Test objectives

  • Validate system performance under anticipated user loads

  • Identify potential bottlenecks and performance issues

  • Ensure system stability during traffic spikes

  • Verify response times remain within acceptable thresholds


Test scenarios

1. Soft

  • Users: 500 concurrent users

  • Timeframe: 6-minute window

  • User Flow:

  • All 500 users reach the Tesla Example claim URL within a 2-minute window

  • 400 users (80%) complete the claiming process and navigate to the Car about page

  • 200 users (50% of those who claimed) proceed to the passport and install the Apple Pass for a multi-owner car approximately 45 seconds after claiming

2. Medium

  • Users: 500 concurrent users

  • Timeframe: 4-minute window

  • User Flow:

  • All 500 users reach the Tesla Example claim URL within a 2-minute window

  • 400 users (80%) complete the claiming process and navigate to the Car about page

  • 200 users (50% of those who claimed) proceed to the passport and install the Apple Pass for a multi-owner car approximately 45 seconds after claiming

3. Hard

  • Users: 500 concurrent users

  • Timeframe: 2-minute window

  • User Flow:

  • All 500 users reach the Tesla Example claim URL within a 2-minute window

  • 400 users (80%) complete the claiming process and navigate to the Car about page

  • 200 users (50% of those who claimed) proceed to the passport and install the Apple Pass for a multi-owner car approximately 45 seconds after claiming


Test results

CPU Server for all 3 tests

Goal: CPU under 55%

Host: DigitalOcean AppPlatform PaaS

Ressources:

  • 8GB RAM

  • 4 Dedicated vCPUs

  • 700 GB Bandwidth

Scale: Auto

Containers: 2 to 4 with CPU Threshold at 50%

Results:


Soft test

Result:

  • Ended after: 5min56.3s

  • % checks: 100%

  • % fails: 0%

  • API Crash & restart: 0

  • Median request time: 338.69 ms

  • Max request time (p(95)): 1.99 s

  • Number of containers needed: 2

  • Container auto scale triggered: No

  • Result screenshot


Medium test

  • Ended after: 3min52.6s

  • % checks: 100%

  • % fails: 0%

  • API Crash & restart: 0

  • Median request time: 348.12 ms

  • Max request time (p(95)): 1.64 s

  • Number of containers needed: 2

  • Container auto scale triggered: No

  • Result screenshot


Hard test

  • Ended after: 2min05.2s

  • % checks: 100%

  • % fails: 0%

  • API Crash & restart: 0

  • Median request time: 1.01 s

  • Max request time (P(95)): 4.48 s

  • Number of containers needed: 2

  • Container auto scale triggered: No

  • Result screenshot

Last updated