weather service

This is the test for backend developers.

Please do not push your solution to a publicly available repo.

Introduction

This is a Python/Flask API to get a forecast for London during Oct 2017.

Environment set up

Checkout the code and set up the service ready for development as follows:

# Set up the package to run in development mode:
python setup.py develop

# Running the webapp locally reloading when changes are made:
FLASK_APP=app.py FLASK_ENV=development flask run

# Optional: Running any test you might choose to write:
pytest -sv

Check the service is running after set up

When the app is running you can do the dollowing curl request:

curl http://localhost:5000/ping
{
  "name": "weatherservice",
  "status": "ok",
  "version": "1.0.0"
}

Pair-programming tasks

• Implement the get forecast endpoint.
• Build and run the API inside a docker container.
• Discuss interview questions.

Get forecast endpoint behaviour

The task is to implement the get_forecast() inside backend.py, returing data to implement the following GET requests

A GET request for a known date and time

curl http://<host:ip>/<city>/<date>/<hour minute>/

curl http://localhost:5000/london/20171018/1500/
{
    "description": "light rain",
    "humidity": "100%",
    "pressure": 1015.55,
    "temperature": 290.44
}

A GET request for an unknown date and time

curl http://localhost:5000/london/21171005/2200/
{
  "message": "No data for 2117-10-05 22:00",
  "status": "error"
}

Questions for interview

Please prepare answers for these questions in advance of the interview.

• How would you return data for a request between two data points in the dummy data.
• How will you scale the weather service to handle increasing load.
  • For example, scaling the systems to 1k Requests Per Minute (RPM) and then 5k RPM.
• Where do you think the bottlenecks lie in scaling such a system?
• How do you ensure that the API response time is always within tolerable range?
• What if any data storage would you use as the backend for the weather service?
  • If you had a storage backend, how would it scale?

Optional question:

• How might we minimise the use of the system resources (cloud hosting costs)?
  • The system should still be able to scale as needed.