System design: Tushar roy | Design messaging/chat service like Facebook Messenger or Whatsapp

April 15, 2021

Here is the link. 

System design : Design messaging or chat service like facebook messenger. https://redis.io/topics/cluster-tutorial​ https://docs.datastax.com/en/cassandr...​ https://en.wikipedia.org/wiki/WebSocket

High quality content. I like to play a few more times and then write down some notes. After the first play, I decided to look into Redis and learn architecture of Redis by myself quickly. 

Facebook messages vs whatsapp / signal

Features/ topics

• one to one chating 
• online / sent/ read
• Sending pictures or big file
• database
• Security

Deep dive on the above topics:

Load balancer 

User A 

Cassandra

Redis - distributed cache 

HTTP - Server cannot initiate request to client - not working 

Alternative: websocket, BOSH, long polling over http 

websocket works best for chat application - sticky 

Heartbeat information 

Redis - distributed cache

Add heartbeat information to Redis - master/ slave, back and forth 

Keep memory - advantages - hold all the information in the memory 

User     |      | Date time

User A | N2 | 

userB   | N3 | ... time info

If N2 dies, User A will figure out, User A will get hooked to N3 instead N2 since N2 dies. 

If user A dies, then the information can be retrieved from Redis, last heartbeat happens. 

User A likes to send message to user B. 

N2 send hi to N3, N2 will not send message directly to userB; the flow is different from N3. 

Database - Cassandra DB

Picture - how two users chats 

Minimum three tables: 

User Table     Encryption 

-------------

User1  User2

Conversation table

Unread table

toUser   fromUser  time

userC     userB        t5

read table

--------------

fromUser   toUser    time

userA          userC     t5

That is my architecture, race condition. 

Next userA to send a picture to userB using messager. 

Picture - encrypted using key, 

Blob storage - every company has a blob storage

Send thumbnail, and if user has interest, then it will be downloaded. 

User case: Send message to a user offline 

----------------------

conversation id     time     text    

  s                            t1       {..}

  5                            t2       {hello} 

N1, N2, N3 manages thousands of sessions. 

Let us talk about optimization, basic search about this ...

Conversation

Optimize the storage - last few days conversation to blog storage - combined, saved as url - go back to history, it is slow - pull information from Blob storage, download - optimization for storing data

Search - by text - reversed index, search on chat history 

group chat - use it in group chat, group table, and then ...