Fullstack Home Assignment
This is a web interface that allows the user to create a playlist of YouTube videos, and plays them in order and a server backend, which stores, and serves the playlist.
Two projects are included: client and server:
- server: Socket.io server that serves the React application
- client: React application
Quick Start
Install both the client and server apps: a README.md file is available for each of the projects with detailed instructions.
And then:
- Start the server
- Start the client app
Scale (server)
Assuming there are very few playlist editors, but many viewers:
Implications
- Server will get hit more often with the following requests:
playlist:get- user just connected and gets the playlistplaylist:remove- user finished to watch a video and requests its removal from the list
Suggested optimizations
Background: Unlike this project's implementation, in which the playlist is stored in-memory, in production it will be persisted in a DB (restore-able in case of server restart or crash).
In-memory data store
To optimize playlist:get requests, the playlist can be loaded into memory (from DB) and serve clients from memory
- This solution will reduce DB load (fewer queries will hit the DB).
- Each change to the playlist (i.e.
playlist:remove,playlist:add) should then be applied both to DB and memory store, while updating the DB can occur async (eventual consistency).
Multiple Socket.IO servers (nodes)
In case the server is the bottleneck, it is possible to deploy multiple Socket.IO servers (nodes), while enabling cookie-based sticky session.
Clustered in-memory data store
In case memory is the bottleneck, there are in-memory data stores which support clusters, like Redis.