CloudServiceBenchmarking

Rodamap

• Set-up simple client-server architecture (containerized)
• add variety of interceptors
• create load generator/benchmarking tool
• GCP setup
• run Benchmarking Experiments
• Analyze results
• Write report

Benchmark Configuration

📢 Important: The benchmarking experiment takes its configuration from this GitHub Repository. Thus, any changes configured via the config.yml file in the client & server directories need to be pushed in order to be applied in the next experiment run.

Server

Specify which interceptor to include on the server side by toggling the respective boolean in the server/config.yml file (for full list, check out the file).

Example for the Prometheus Metrics Interceptor:

# Prometheus Metrics Interceptor
metrics:
  enabled: true

Client

Define the load pattern to generate in the client/config.yml file. In the example below, the client will generate a load of 100.000 requests over 100 iterations each, resulting in 10.000.000 requests throughout the whole experiment. With each iteration the number of concurrent (parallel) requests is increased.

total_requests: 100000
max_concurrent_requests: 100
min_concurrent_requests: 1

Warmup

Specify the amount of requests sentto the server before the experiment is prperly started in the client/config.yml file. In the example below, the client will send 100.000 requests before the experiment is started.

warmup_requests: 1000000

GCP Setup

Prerequisites

• Terraform
• GCP Account
• gcloud CLI

Service Account

Create a service account in GCP and download the JSON key file. Rename it to credentials.json and place it in an /.env directory of this repo.

SSH Key

Create an SSH key pair and place them in the /.env directory of this repo.

Terraform

From the /tf directory of this repo, initialise terraform

terraform init

Then run the terraform script to create the GCP resources

terraform apply

⏳ optional: SSH into the (client) VM. You'll find the current result progress in the /home/ubuntu/CloudServiceBenchmarking directory. Check out the result.csv file for the current progress.

🏁 After the Experiment is done, the results.csv file will be copied to your local machine automatically.

To delete the resources, run

terraform destroy

Local Setup

Compile Locally

From the /server directory of this repo, compile the server binary.

go build

Build Docker Images

From the top of this repo, build the docker images for server and client.

docker-compose build

Run containers

Runs the server container first, and then the client container. The client container needs to be run with the --network="host" flag to allow it to connect to the server container.

docker-compose up

🏁 After the Experiment is done, the result.csv file will be copied to your local machine automatically.

Bring down containers

docker-compose down