This repository contains a simple Next.js application and a k6 script for performance benchmarking. The Next.js application includes an API that sleeps for 50ms and then generates a random number. The k6 script measures the application's performance on a DigitalOcean App Platform server.
Benchmark results are at the bottom of this README.
You need to have Node.js and npm installed on your machine to create the Next.js application. Install k6 for the performance testing.
The Next.js application contains a single API at pages/api.js. This API waits
for 50ms, generates a random number, and returns it in a JSON response.
Follow these steps to deploy the Next.js application to the DigitalOcean App Platform:
- Sign up for DigitalOcean App Platform and click 'Create App'
- Select your repo and branch, then hit 'Deploy'
Your Next.js application should now be live on DigitalOcean.
We use k6 for performance testing due to its flexibility in defining performance
test scenarios. The k6 script (bench.js) included in this repository ramps up
the load to 50 virtual users (VUs), holds it for 3 minutes, and then ramps down.
To run the k6 script, use the following command:
k6 run bench.jsThe script will stop if more than 1% of the requests result in an HTTP error or if the 95th percentile response time exceeds 1 second.
The k6 output includes several useful metrics, such as the number of total requests, the number of failed requests, and the average response time. By reviewing these metrics, you can understand how many requests per second your Next.js app can handle on a single DigitalOcean server before it starts to slow down.
- The k6 output shows that our server handled 78.9 requests per second (RPS) on average, with a maximum of 50 concurrent users.
- CPU usage hovered around 63% during the test.
- Memory usage was constantly at 100%. This indicates that our application is memory-bound.
- The average response time was 507.82ms.
- The 95th percentile response time was 793.04ms. This means that 95% of requests were processed within this time, but 5% of requests took longer.
- No HTTP requests failed, indicating that the server was able to handle the load without dropping requests.