After running make start docker-compose will build and start 3 containers:

• test - it will start end-to-end tests. It will write test data directly to backing redis and get data from proxy
• proxy - proxy with LRU cache. Data can be fetched via HTTP or RESP
• redis - backing redis container

The most common implementation of LRU cache is to use doubly linked list and hash map.

However, JavaScript has a specific implementation of Map. According to specification:

The Map object holds key-value pairs and remembers the original insertion order of the keys.

This allows to use just a standard Map and make the code of ./proxy/src/cache.js very simple and clean

O(1)

Run tests:

make test

Run only proxy:

make start

Configuration can be changed via environmental variables:

• REDIS_URL - Address of the backing Redis
• CACHE_TTL - Cache expiry time
• CAPACITY - Capacity (number of keys)
• PORT - TCP/IP port number the proxy listens on

When you run make start default values are taken from .env file and are passed as environmental variables to a container.

By default tests will be run for HTTP server. In order to run them for RESP, you need to set PROXY_PROTOCOL=RESP in .env

• devops - 2h
• development - 4h
• resp protocol - 3h
• documentation - 1h

To be fair, the concurrent processing is not really parallel, it's asynchronous. However, the requirement stated:

When multiple clients make concurrent requests to the proxy, it would execute a number of these requests (up to some configurable limit) in parallel (i.e. in a way so that one request does not have to wait for another one to complete before it starts processing)

The app meets this requirement. To be trully parallel, we need to use cluster module to fork processes.