ManyMetrics is an open-source event tracking infrastructure based on AWS. Think Heap or Mixpanel but data warehouse first and it is in your AWS infra, you own the data and it is crazy cheap.
The infrastructure is provisioned using terraform, which sets up the following resources:
- API powered by API Gateway and Lambda
- S3 storage in the Apache Iceberg format
- AWS Athena for querying data with SQL
- A JavaScript client library that can be used to send common product analytics (feel free to use the API instead!)
The code is licensed under the MIT license.
ManyMetrics uses terraform for setting up resources in an AWS account.
-
Confgiure AWS access in your terminal
-
Run the following command (replace the name)
cd sample terraform apply -var="name=sample"This command should
- Create ECR repositories
- Build and upload docker images
- Create other AWS resources like API Gateway and Lambdas
-
Run
terraform output- you will need the values from the output in a few steps below -
Update
browser_sdk_url/sample.htmlwith the values from theapi_instance_urlandbrowser_sdk_urloutputs -
Render a sample page with enabled tracking
cd browser-sdk
python3 -m http.server 8000
open http://0.0.0.0:8000/
- Open development tools in your browser and watch events being sent to the API when clicking buttons
- Open Athena in AWS Console (use the AWS region where you deployed ManyMetrics to) and select the database from the
glue_database_nameoutput - Run the query to see the events
select * from events;. You should see the events
The calculations below do not include S3 storage and a few others costs like network costs but they should be relatively small.
| Service | Cost | 1kk events | 10kk events | 100kk events |
|---|---|---|---|---|
| API Gateway | $1 per 1kk req | $1 | $10 | $100 |
| Lambda | $0.0000000167 per ms [1] | $0.10 | $1 | $10 |
| Kinesis | $0.015 per hour [2] | $10.8 | $10.8 | $10.8 |
| Total | $11.9 | $21.8 | $120.8 |
- Assumes that 10,000 events are processed in roughly 1 min in Lambda.
- One shard of provisioned Kinesis, more shards might be required if the volume is higher than 10k events/min and data should be avialable immediately
Use pre-commit.
pip install pre-commit
pre-commit