I originally designed this with the understanding that the focus was on the API and code design so I chose python and flask because of it's friendly structure, however with the limited number of required endpoints this design may be slightly overkill. The system consists of three docker containers: nginx, python, and postgres. Nginx fields and buffers the HTTP request/reponses, postgres is the storage medium, and the python container runs the application. There are two docker networks to isolate the database from the gateway server, with the application in the middle.

The application itself consists of two main files: the app.py file and the models.py. The models script. The models script contains the definition for the SensorSample object which represents a sensor bundle. The model contains much of the necessary validation and handles most of the communication with the database via SQLAlchemy. The app.py script contains all the endpoints. There is third file which just contains a simple object defintion used to represent a standard response to a malformed PUT request payload.

I worked out the design and built the application before I realized there was a limitation on using an application server so mine does include gunicorn, because the focus of our in-person meeting seemed to be on API design rather than performance optimization with lower-level concurrent, and didn't have time to rewrite it. Had I realized the focus was on concurrency and peformance I would have rather built this application in Golang without a big HTTP framework like flask and rather just use the built-in HTTP server, but for API design python+flask was faster.

This project includes a docker-compose.yml file which spins up several docker containers hosting the application and it's required resources. To run the application, all you need is docker and docker-compose installed on your machine. Once you have those, you can just run:

$ docker-compose up

The API is a simple HTTP API.

• Gian Biondi [email protected]