Performance Testing for APIs: Basic setup with K6

Performance testing is a crucial part of modern software development, ensuring that your APIs can handle real-world traffic and deliver a smooth user experience. It is the process of evaluating how a system behaves under various levels of load. For APIs, this means simulating multiple users making requests to ensure the API remains responsive and stable.

There are several major kinds of performance testing, each with a specific purpose:

1. Load Testing: Measures how the system performs under expected user loads. The goal is to identify performance bottlenecks before the software application goes live.
2. Stress Testing: Determines the system’s robustness by testing it under extreme conditions, often beyond its expected capacity, to see how it handles high traffic or data processing.
3. Soak (Endurance) Testing: Checks the system’s stability and performance over an extended period, ensuring it can handle sustained use without degradation.
4. Spike Testing: Evaluates the system’s reaction to sudden large spikes in the load generated by users.
5. Scalability Testing: Assesses the system’s ability to scale up or down in response to increased or decreased load.

K6 is an open-source, developer-centric tool for load testing APIs and web applications. It’s popular because:

• It uses JavaScript for scripting, making it accessible to most developers.
• It’s easy to automate and integrate into CI/CD pipelines.
• It provides rich metrics and reporting.
• It’s lightweight and fast.

While K6 is a great choice, there are several other widely used tools for API performance testing:

Each tool has its strengths—choose based on your team’s language preferences, required features, and integration needs.

Let’s walk through a real-world example using this repository.

// Controller action
[HttpGet]
public async Task<IActionResult> Get() {
  var result = await _context.Books.Select(b =>new {
    b.Id,
    b.Title,
    b.Author,
    BorrowCount = b.Borrows.Count
  }).OrderByDescending(b =>b.BorrowCount).FirstOrDefaultAsync();

  if (result == null || result.BorrowCount == 0) return NotFound("No borrow records found.");

  return Ok(result);
}
                

import http from 'k6/http';
import {
    check,
    sleep
} from 'k6';

export const options = {
    vus: 10, // number of virtual users
    duration: '10s', // test duration
};

export default function() {
    const res = http.get('http://localhost:5127/mostborrowedbook');
    check(res, {
        'status is 200': (r) => r.status === 200,
    });
    sleep(1);
}  
                

cd MostBorrowedBookApi
dotnet run
                

k6 run k6-mostborrowedbook.js
                

k6 run k6-mostborrowedbook.js --out json=results.json
npx k6-reporter results.json
                

The full demo code for this article is available at: ajaysskumar/API-Performance-Testing

This repository contains:

• MostBorrowedBookApi: ASP.NET Core Web API with an in-memory database, seeded with thousands of books and borrows at startup. The endpoint /mostborrowedbook returns the book with the highest borrow count.
• k6-mostborrowedbook.js: K6 test script for load testing the API.
• run-perf-test.sh: Bash script to run the API and K6 test in one shot.

After running the test, K6 outputs a summary like this:

Running K6 test...

          /\      |‾‾| /‾‾/   /‾‾/   
     /\  /  \     |  |/  /   /  /    
    /  \/    \    |     (   /   ‾‾\  
   /          \   |  |\  \ |  (‾)  | 
  / __________ \  |__| \__\ \_____/ .io

     execution: local
        script: k6-mostborrowedbook.js
        output: -

     scenarios: (100.00%) 1 scenario, 10 max VUs, 40s max duration (incl. graceful stop):
              * default: 10 looping VUs for 10s (gracefulStop: 30s)


     ✓ status is 200

     checks.........................: 100.00% ✓ 60       ✗ 0   
     data_received..................: 14 kB   1.2 kB/s
     data_sent......................: 5.8 kB  509 B/s
     http_req_blocked...............: avg=1.63ms   min=1µs      med=6µs      max=11.72ms p(90)=11.09ms p(95)=11.33ms
     http_req_connecting............: avg=589.03µs min=0s       med=0s       max=7.52ms  p(90)=2.06ms  p(95)=4.84ms 
     http_req_duration..............: avg=874.47ms min=758.66ms med=823.23ms max=1.18s   p(90)=1.15s   p(95)=1.17s  
       { expected_response:true }...: avg=874.47ms min=758.66ms med=823.23ms max=1.18s   p(90)=1.15s   p(95)=1.17s  
     http_req_failed................: 0.00%   ✓ 0        ✗ 60  
     http_req_receiving.............: avg=1.09ms   min=13µs     med=48µs     max=11.35ms p(90)=1.23ms  p(95)=10.06ms
     http_req_sending...............: avg=1.43ms   min=5µs      med=27.5µs   max=9.13ms  p(90)=8.22ms  p(95)=8.46ms 
     http_req_tls_handshaking.......: avg=0s       min=0s       med=0s       max=0s      p(90)=0s      p(95)=0s     
     http_req_waiting...............: avg=871.94ms min=758.58ms med=823.17ms max=1.17s   p(90)=1.13s   p(95)=1.16s  
     http_reqs......................: 60      5.302087/s
     iteration_duration.............: avg=1.87s    min=1.75s    med=1.82s    max=2.2s    p(90)=2.18s   p(95)=2.19s  
     iterations.....................: 60      5.302087/s
     vus............................: 10      min=10     max=10
     vus_max........................: 10      min=10     max=10


running (11.3s), 00/10 VUs, 60 complete and 0 interrupted iterations
default ✓ [======================================] 10 VUs  10s
        

Here’s a breakdown of the key metrics:

If you use k6-reporter, you’ll get an HTML file with charts and detailed breakdowns.

Think of your API like a busy coffee shop. Performance testing is like simulating hundreds of customers ordering coffee at once. You want to ensure the baristas (your API) can handle the rush without slowing down or making mistakes. K6 helps you create that simulated rush and measure how well your API copes.

Performance testing is essential for delivering reliable APIs. K6 makes it easy to simulate real-world load, catch bottlenecks, and ensure your API is production-ready. With a simple setup and powerful reporting, it’s a great tool for developers and DevOps alike.

Try it out with this repository and see how your API performs!