Simple system for profiling performance differences between message broadcast APIs. Accompanying code to the blog post which can be found here
Deployment uses Github actions and is very easy to set up. All you need is a deployment user in your AWS account:
- Fork this repo in your Github account.
- Add a new IAM policy in you AWS account. The
policy should use the permissions provided in
resources/deployment_role.json. - Create a new user for deployment purposes, add the policy you created in the previous step. Note: the user only requires programmatic access
- Copy out the Access Key and Secret key.
- Go to your forked repository in Github, choose Settings > Secrets and add the following:
- AWS_ACCESS_KEY_ID
- AWS_SECRET_ACCESS_KEY
- Create a new commit on the master branch (make a dummy change to the README for example) - this will trigger the Github actions pipeline and deploy the stack.
Other than the resources required for deployment, the following AWS Resources will be provisioned:
- Lambda:
- BatchProfilerIndividual
- BatchProfilerBatch
- BatchProfilerBoto3BatchUtils
- SQS Queue:
- message-batch-profiling-TargetQueue
To run the tests you can use AWS Lambda's in-built "Test Events" functionality. Setup tests against each of the deployed Lambda Functions:
{
"messages_to_send": 10,
"sqs": "True",
"kinesis": "False"
}Running this test will cause the Lambda to send the requested number of messages to the SQS queue. Note: A maximum of 10 should be used.