Instagram feed- WIP

🔃all HLD design articles

· Subtopic
· Requirements
· Data access pattern
· Data schema
∘ How to get number of likes for comment/post?
∘ Following/Follower feature
· High Level Design
· TODO
· Resources

Subtopic

• generate feed
• celebrity problem
• image upload and processing flow + CDN integration

Instagram Image upload + CDN — [Notes]

Alex xu — design template

Requirements

• Store/Get Images
• Like + Comment + Like a comment on Image (Can we comment on a comment? i.e. recursive comments?)
• Follow someone
• Publish a news feed

Data access pattern

Data access pattern --

* Get all the posts made by a user + sort by timestamp
* Get all the comments made on a post + sort by timestamp
* Get the number of likes on a post/comment
* Get list of users who have liked a post
* Get all posts made by a user's followers and sort by timestamp

Data schema

TODO — what database to use? and did you decide on the same?

• Likes Table: ID, Type, Active, ParentID, UserID, timestamp
  [ParentID — can be a comment or a post,
  Active — means is active or deleted
  Type — comment or a post]
• Post Table: UserID, PostID, text, Image URL, timestamp
• Comment Table: ID, text, timestamp, PostID

How to get number of likes for comment/post?

Approach # 1

• select (*) from like where PostID = ‘abc’;
• very slow

Approach # 2

• hold “likes” in a column in likes/posts table

Post Table: UserID, PostID, text, Image URL, timestamp, likes
Comment Table: ID, text, timestamp, PostID, likes

• whenever there is a like on the post insert in like table and do +1 in likes column
• this column is an aggregated value and doesn’t belong with posts/comments table data.

Approach #3

• Create an Activity table [ParentID, Likes]

Following/Follower feature

Data access pattern --

* Who follows user Hawking?

    Select * from following where followeeId = 'Hawking';

* Which users does user Hawking follow?

    Select * from following where followerId = 'Hawking';

Following Table

High Level Design

Design 1 — Not scalable at posts service level

Design 2 — Precompute

• Problem here is that when a celerity creates a post there will be lot of cache update + lot of notifications

Design 3 — Precompute with celebrity problem

TODO

• More detailed read to resolve celebrity problem — push vs pull vs hybrid? — algorists
• Image upload and processing flow
• Image render flow + CDN integeration

Resources

• Arpit bhiyani
• Interview ready
• Alex xu template
• scaler videos