Performance Test Suite

Overview

Execute load testing, stress testing, and performance benchmarking to identify bottlenecks, establish baseline metrics, and verify SLA compliance. Supports k6 (recommended), Artillery, Apache JMeter, Locust (Python), and autocannon (Node.js).

Prerequisites

• Performance testing tool installed (k6, artillery, locust, jmeter, or autocannon)
• Target application deployed in a production-like environment (not local dev)
• Baseline performance metrics or SLA targets (e.g., p95 < 200ms, 99.9% availability)
• Monitoring stack accessible (Grafana, CloudWatch, Datadog) for resource metrics during tests
• Test data sufficient to avoid cache-only responses

Instructions

1. Define performance test scenarios based on production traffic patterns:

• Load test: Simulate expected peak traffic (e.g., 500 concurrent users for 10 minutes).
• Stress test: Ramp beyond expected capacity to find the breaking point.
• Spike test: Sudden burst of traffic (0 to 1000 users in 10 seconds).
• Soak test: Sustained moderate load for extended duration (1-4 hours) to detect memory leaks.

1. Create test scripts targeting critical endpoints:

• Identify the top 5-10 most-hit API endpoints from production access logs.
• Include both read (GET) and write (POST/PUT/DELETE) operations.
• Simulate realistic user behavior with think time between requests.
• Use parameterized data to avoid cache-only hits (randomize query parameters, user IDs).

1. Configure load profiles:

• Define virtual user (VU) ramp-up stages (e.g., 10 VUs for 1 minute, then 50 VUs for 5 minutes).
• Set test duration appropriate to the scenario (load: 10-15 min, soak: 1-4 hours).
• Configure request timeouts matching production settings.

1. Execute the performance test:

• Run from a machine with sufficient network bandwidth and CPU.
• Avoid running from the same host as the application under test.
• Monitor application metrics (CPU, memory, DB connections) during execution.

1. Analyze results against SLA thresholds:

• p50, p90, p95, p99 response times.
• Requests per second (throughput).
• Error rate (target: < 0.1% for load test, higher tolerance for stress test).
• Resource utilization (CPU < 80%, memory < 85% at peak load).

1. Identify and document bottlenecks:

• Slow database queries (check slow query logs).
• CPU-bound operations (profiling data).
• Memory leaks (growing RSS over soak test).
• Connection pool exhaustion (database or HTTP client).