Requirements

Functional Requirements:

-Post a comment on post/video/image

-Reply to the comment and create a hierarchy.

-Display comment with pagination - sort by time

Non-Functional Requirements:

• High availability
• High read:write ratio
• Scalable
• Secure
• Low latency

Calculations:

100 DAU

20 % comments at least once

2 avg comments

40 mil comments per day

Peak is 10 times higher

Storage capacity req:

We can revisit this but right now considering 50-100 TB per year

Traffic :

We can revisit the numbers

~ 500 wps

r: w ratio 100 :1

Database design:

Strong consistency within single post

Eventual consistency across broader system

• No SQL DB seems like a good choice - Scalable , can store nested comments easily and can scale
• We can use Redis and write through the cache to achieve strong consistency and low latency

Choose Nosql over postgres here to complex avoid joins and complicated fetches over multiple tables - based on time, based on which user comment on what would need joins between user table comments table etc and we would need good sharding to scale.

Entities:

comments: id/uuid, post_id, parent_id, depth as int, author_id, text, user_id

post_comment_aggregates: post_id, total_counts, last_comment_at, votes_count

API Design

POST /comments

{post_id, parent_id, body}

response : {id, depth:}

GET /comments/{id}

response { body}

DELETE /comments/{id}

We can get votes and ratings

High-Level Design

Main components:

UI client to post comments

CDN - to serve, image, video you are commenting on , caches to HTML, css

API gateway: Auth, rate limiting , routing

Load Balancer: route to concurrently running commenting services ( scaled horizontally )

Services:

Auth service

Comment Service

Aggregation service

Redis between Nosql and services, read replicas to scale reads and fault tolerance.

Kafka for real time comment aggregation feed aggregation services

Detailed Component Design