About

Simple sample projects to showcase image size and container/k8s deployment memory footprint for the following languages:

Rust
Go
C#
Python
Java

`Hello, world!`

Simple Hello, world! examples only containing the necessary code and no additional frameworks/libaries. Only prints Hello, world! message and waits for key input to stop. The docker files are all multi-stage builds and use Alphine base images as these are small, secure and generally well maintained.

`Hello, world!` rest API

Simple Hello, world! REST Api examples using popular libraries for the language:

Rust -> warp
Go -> net/http
C# -> AspNetCore
Python -> flask
Java -> spring

All examples expose a simple API which can be reached here:

http://127.0.0.1:8080/hello

This will return the following JSON message:

{
  "messsage": "Hello, world!"
}

The docker files are all multi-stage builds and use Alphine base images as these are small, secure and generally well maintained.

Building and running the images/containers locally

To build and run the containers you can use the snippets below from the respective folders:

#`Hello, world!`
docker build -t container-name .
docker run -it  container-name:latest

#`Hello, world!` rest API
docker build -t container-name .
docker run -p 8080:8080 container-name:latest

Deploying on K8s

For the K8s deployment, only the Hello, world! rest API examples where used with preprepared images which can be found here:

To deploy them you can use the deploy.yml files below from the respective folders:

kubectl apply -f build -t deploy.yml

This will create a simple Deployment with 3 replicas and a nodeport to run the requests against.

How results where gathered

Locally

All local builds and runs where done with where done with docker-19.03-12-stable under Windows 10 Professional.

Image size was determinded with:

docker image ls

Container runtime memory usage was determinded with:

docker stats

K8s deployment

K8s deployment runtime memory usage was determinded by aggregating memory usage for each pod. The memory usage per pod was measured with the K8s metric server.

Request simulation

The requests against the Hello, world! rest API examples where executed with the HelloWorldApi/apitest.py python script.

Results

`Hello, world!` locally

Memory was measured after the Hello, world! message waiting for key input.

Image	Size	Memory
rust-hello-world	8.77MB	844KiB
go-hello-world	7.77MB	7.164MiB
csharp-hello-world	87.3MB	5.355MiB
python-hello-world	42.7MB	5.062MiB
java-hello-world	345MB	11.38MiB

`Hello, world!` rest API locally

Memory was measured after running a 1000 requests against the API`s.

Image	Size	Memory
rust-hello-world-api	10.9MB	904KiB
go-hello-world-api	13MB	7.195MiB
csharp-hello-world-api	105MB	28.86MiB
python-hello-world-api	46.3MB	34.7MiB
java-hello-world-api	361MB	116MiB

`Hello, world!` K8s deployment

Memory was measured after running a 10000 requests against the API`s through a nodeport

Image	Size	Memory usage for 3 pods
rust-hello-world-api	10.9MB	3MiB
go-hello-world-api	13MB	21MiB
csharp-hello-world-api	105MB	96MiB
python-hello-world-api	46.3MB	73MiB
java-hello-world-api	361MB	363MiB

Notes

The Alpine base images really help out C#, Java and Python as using the normal ones would increase the size of them 2 to 8 fold. The Rust and Go images on the other hand could even be made smaller if the scratch base image is used.

seriva/container-test