System requirements
Functional:
Post Post
Share Tweets
Follow Users
View news feed
Like and react to posts
Follow and unfollow users
Account Authentication
Account Profiles
For posts the user should be able to edit the post
Upload photos and upload videos
Non-Functional:
Eventual Consistency on News Feed
Low Latency
Highly availability
Highly reliable
Fault tolerance
Capacity estimation
Let's assume 300,000,000 MAU
Unique DAU 10,000,000
Peak concurrent users: 1,000,000
QPS: 1,500,000
This would be a Read heavy system. Let's assume 20% of QPS are Writes: posts, likes, comments and the other 80% are Reads
For the actual storage:
300,000 QPS at something like 0.5KB
150,000 KB = 150MB * 60 = 9000 MB per minute = 9GB per minute 540 GB per hour
500GB *24 = 12,000 = 12 TB per day
3.6 PB per year
API design
Define what APIs are expected from the system...
post Posts (authentication_tokern, message) and returnes a 200 successs message
get Posts (authentication_token, user_id, hashtag_id, search_params_) and returnes a 200 successs message. this would also have pagination
patch Posts (authentication_tokern, message)
post Like
post comments
get Like
get Comments
Database design
Defining the system data model early on will clarify how data will flow among different components of the system. Also you could draw an ER diagram using the diagramming tool to enhance your design...
user table with user_id and other data like location
user settings table
all these tables should be sharded on the user_id field
the database should be multi-regional.
could use something like Spanner or MySQL. Spanner is likely better since it needs to be global
All other database is noSQL/ unstructured
Comments would be in a document store and in a nested format for quick reads
High-level design
You should identify enough components that are needed to solve the actual problem from end to end. Also remember to draw a block diagram using the diagramming tool to augment your design. If you are unfamiliar with the tool, you can simply describe your design to the chat bot and ask it to generate a starter diagram for you to modify...
Mobile Front End
Load Balancer/Api Gateway
Micro Services architecture
CDN (for popular posts)
Sharded database
Web socket
Request flows
Explain how the request flows from end to end in your high level design. Also you could draw a sequence diagram using the diagramming tool to enhance your explanation...
Requests flows through everything
Detailed component design
Dig deeper into 2-3 components and explain in detail how they work. For example, how well does each component scale? Any relevant algorithm or data structure you like to use for a component? Also you could draw a diagram using the diagramming tool to enhance your design...
The CDN would have both push and pull mechanisms. The CDN could use something like Redis and the video files could use something like Cloud CDN or Cloud flair
The load balancer would also have a rate limiter.
THe API gateway would have to have sophisticated routing services to make sure any pod traffic goes to pods in a healthy state if usiging something like Kubernetes
Trade offs/Tech choices
Explain any trade offs you have made and why you made certain tech choices...
I would use primarily NoSQL. the actual tweets, posts, comments, and replies could be stored in a NoSQL document type storage. I was additionally use Blob storage for the photos and videos that users upload.
-Focus on high availability
-quick write times
-highly reliable, can have a lot of replication
SQL is not as good a choice since the write time may be significant, but SQL can be used for overall user meta data that doesn't change as much and is more relational
Failure scenarios/bottlenecks
Highly popular posts with videos could be a failure scenario. Reading from Blob storage could flood the server with too many requests and lead
to unhealthy state -> smart pushing of content to the CDN
External API getting flooded
Too many simultaneous comments
Rate limiting more generally -> user tries to upload too much content
Future improvements
What are some future improvements you would make? How would you mitigate the failure scenario(s) you described above?
message queue for comments. really rely on eventual consistency
Message queue more generally, rate limits, "try again later messages". This could use kafka or rabbitMQ
Smart pushing to the CDN. Mix of push and pull. TTL the videos or keep an algorithm with a popularity counter that smart helps drop things from cash
External api -> use a rate limiter and quotas