ankit-22 / assignment

Its a great assignment that I came across recently.

Introduction

I came across this assignment while applying for a startup. The questions were of good quality, so decided to share here. It was a take home assignment but it was expected to be done in 2hrs. I completed it along with the documentation in around 2.5 hrs. You can read the question in assignment.txt file. I'll explain my solutions below.

Solution

Q1

The project is in folder q1q5/ I have created 4 Django Models: User, Post, Comment and Connection.

The User model will store user data (name, email, password, etc.). The Post model will store the posts made by users (title, content and author(Foreign key to User model)). The Comment model will store the comments made on post (content, date, post on which comment is present(Foreign key to Post model), author(Foreign Key to User model), parentComment(Self Reference Foriegn key to handle comment replies)). The Connection model will store two foreign keys on User model representing their friendship.

Scalability Comments

Lets break our scalability problem into 2:

• Quick Queries

  • Fetching data for a user will be simple as email can be indexed or primary key can be used to search
  • Fetching data for posts of a user is simple again as we have the foreign key.
  • Per post we also need to query comments, this should be a different async call to load paginated comments lazily. We can move the logic of structuring the heirarchy of replies using the parentComment foreign key to the client side.
  • We can fetch friends of a user by checking both foreign keys. We can move the logic of finding if 2 users are friends to the client side.

• Storing Data and Horizontal Scaling

  • For Users model, we can have millions of entries (A few GBs) without need for horizontal scaling, but we may use sharding to scale for future.
  • Assuming an average user will make 1000 posts, this can be in billions (A few TBs). Normally dbs can handle such loads but we can implement a caching mechanism at microservice level to reduce costly fetch operations on recent data. We may also need to use sharding mechanism.
  • Assuming an average post will have 1000 comments, this data will be of a few 100 TBs. We will need both a caching mechanism and sharding of data to store it on a distributed system. We may also need a master-slave design to ensure data availability and consistency.
  • Assuming an average user will have 1000 friends, this data will again be of a few TBs. Caching could help to save costly fetch operations on highly active users. Sharding is not necessarily required but we can have that to scale for future.

Q2

I have created an node script in q2q3 folder. Install the dependecies by running npm install. You can run the srcipt using npm run q2 <your_limit>. The default limit is 10, you can provide any number between 1 and 10 as there are only 10 entries in box office section.

I have also dockerised the script. You can simply build the docker image: sudo docker build . -t <your_tag> Then run the docker image using: sudo docker run <your_tag> <your_limit>

Q3

I have used Redis as a Messaging Service. I have a redis instance running in background. I have written the consumer in NodeJS q2q3/q3_consumer.js and publisher in Python q2q3/q3_publisher.py.

Start the redis instance on localhost:6379 using redis-server

To run the consumer:

npm install
npm start

To run the publisher: Install requirements: pip install -r requirements.txt and run python q3_publisher.py

Q4

I have created a simple python script to first squeeze all question blocks together, then sort them based on question number.

rm assignmentNew.txt
python q4.py
cat assignmentNew.txt

Assumptions:

• Each question is in new line and starts with Q has a number after it and then a )
• Bonus Question is a part of Q5
• Subquestions will be below the parent question

Q5

In the project same as Q1, I have added a decorater in facebook_lite.views named allow_post_only. This decorator checks the request method and only allows POST, it responds with 405 on any other method.

You can install the requirements using: pip install -r requirements.txt. You can run the django app using:

python manage.py makemigrations
python manage.py migrate
python manage.py runserver

You can test it by importing the postman collection file: Facebook_Lite.postman_collection.json.

assignment's Issues

konsi company?