Stress Testing Guide for Nginx On Ubuntu VPS

✅ Objective

To evaluate how our API /api/testendpoint handles different levels of concurrent user loads and request volumes on our VPS environment.

🛠️ Tools Used

• Apache Benchmark (ab) — Simple HTTP load testing tool.
• Netdata — Real-time system monitoring during tests.

🔥 Step 1: Install Apache BenchmarkOn Ubuntu VPS:

sudo apt install apache2-utils -y

🚦 Step 2: Running the Stress Tests🔸 Test 1 — Moderate Load:

• Command:

ab -n 500 -c 50 http://<server-ip>/api/testendpoint

• Purpose: Simulate 50 concurrent users making a total of 500 requests.
• Observation:
• Avg response ~1.3 seconds.
• No failed requests.
• CPU spiked to 100% during the test.

🔸 Test 2 — Lighter Concurrency:

• Command:

ab -n 200 -c 25 http://<server-ip>/api/testendpoint

• Purpose: Simulate 25 concurrent users with 200 total requests.
• Observation:
• Avg response ~594ms.
• No failed requests.
• CPU usage stayed moderate.

🔸 Test 3 — High Volume, Low Concurrency:

• Command:

ab -n 5000 -c 50 http://<server-ip>/api/testendpoint

• Purpose: Sustain 50 concurrent users over a long session (5000 requests total).
• Observation:
• Avg response ~1.3 seconds.
• Throughput ~38 requests/sec.
• Stable performance without failures.

🔍 Step 3: Monitoring with Netdata

• During each test, Netdata was used to monitor:
• CPU Utilization → Observed spikes to 100% during high concurrency.
• RAM Usage → Stayed stable; no leaks.
• Disk I/O → No significant bottlenecks.
• Network I/O → Spikes matched request throughput.
• PostgreSQL Metrics → Connections and query throughput were stable.

🎯 Observations Summary Test Concurrency Avg Response RPS Failures CPU Impact 1 50 1.3s ~38 0 100% spike 2 25 594ms ~42 0 Moderate 3 50 1.3s ~38 0 100% spike 🚀 Recommendations

• ✅ Current setup comfortably handles up to 25 concurrent users with response time under 600ms.
• ⚠️ 50 concurrent users causes CPU saturation, leading to ~1.3s response time consistently.
• 🔥 Immediate optimizations:
• Cache API responses for repeated queries.
• Scale VPS CPU from 1 core to 2-4 cores.
• Optimize PostgreSQL queries (check for N+1, indexing).

🔥 Request Flow During ab Test:

ab tool → NGINX → PHP-FPM (Laravel) → PostgreSQL (if DB involved)

✅ Conclusion

• System is stable under heavy load.
• No memory leaks or server crashes observed.
• Performance bottleneck is primarily CPU-bound, especially with Laravel app logic and database interactions during concurrent load.
• Netdata proved highly effective in real-time monitoring and bottleneck diagnosis.

This document serves as the record for how we conducted our stress testing, what was observed, and the actionable outcomes.