DotnetGoRustBenchmark

I have wanted to step into the shoes of a QA engineer so I decided to implement the same contract in three different technologies and run a simple benchmark

Table of contents

• General info
• Benchmark
• Technologies
• Setup
• Status
• Inspiration

General info

I have built three simple application which all implement the same contract which is:

curl -X GET -H "Content-Type: application/json" http://localhost:8080/helloWorld

responds with

{
    "message": "Hello world!"
}

The same response is for ports 8081 and 8082. And also second endpoint which retrieves simple user from MongoDb database:

curl -X GET -H "Content-Type: application/json" http://localhost:8080/users/1

and the response is

{
    "Id": 1,
    "Name": "Arthur",
    "Surname": "Morgan",
    "Email": "ArthurMorgan@mail.com"
}

For ports 8081 and 8082 response slightly differs, but its only a difference in casing. Overall the project looks like this:

Pic.1 Visualization of docker compose containers

Benchmark

Small disclaimer: I really recommend to take those benchmarks with more than a grain of salt and not to jump to any conclusions - I was curious what would be the result but at the end of the day those benchmarks are really closer to toy examples rather than to production grade stress test. I intentionally don't provide any comment about the results

Once the project was finished I wrote simple tests with k6. All three containers got limit of 1 CPU and 512MB of RAM limit in docker compose. MongoDB was not limited. Specification of the physical machine:

• CPU: Intel(R) Core(TM) i7-9750H CPU @ 2.60GHz
• RAM: 24GB
• OS: Windows 10 Home

After the docker compose was up and running tests were run one by one in order:

1. Go API - endpoint GET /helloWorld
2. .NET API - endpoint GET /helloWorld
3. Rust API - endpoint GET /helloWorld
4. Go API - endpoint GET /users/1
5. .NET API - endpoint GET /users/1
6. Rust API - endpoint GET /users/1

Below I am presenting the results of http_req_duration metric. Benchmark was started with command:

k6 run --summary-trend-stats "min,avg,med,max,p(50),p(95),p(99)" .\script.js

Endpoint GET /helloWorld has been tested with 100 preallocated virtual users, at rate 100 with duration of 10s:

	min [ms]	avg [ms]	med [ms]	max [ms]	p(50) [ms]	p(95) [ms]	p(99) [ms]
Go	0,511	1,78	1,46	46,23	1,46	2,92	5,41
.NET	0,511	3,13	1,85	110,38	1,85	4,72	48,26
Rust	1,05	2,64	2,05	84,04	2,05	4,42	9,33

Pic.2 Visualization of the data on a chart

Endpoint GET /users/1 has been tested with 100 preallocated virtual users, at rate 10 with duration of 10s:

	min [ms]	avg [ms]	med [ms]	max [ms]	p(50) [ms]	p(95) [ms]	p(99) [ms]
Go	25,55	32,55	30,39	46,71	30,39	42,4	45,55
.NET	7,71	46,42	11,46	724,33	11,46	310,26	680,92
Rust	19,16	33,78	26,82	267,57	26,82	48	187,92

Pic.3 Visualization of the data on a chart

Technologies

• .NET 8
• Go 1.22
• Rust 1.74
• MongoDb
• Docker
• k6 0.50.0

Setup

1. Run docker compose in src folder: docker-compose up
2. APIs should be accessible from http://localhost:8080, http://localhost:8081 and http://localhost:8082

Status

Project is: finished

Inspiration

Cause coding is fun