I was recently going through some system design videos and stumbled upon youtube system design. And some of the specs mentioned in the video made me wonder how youtube handles all the comments on its platform. There must be billions of entries daily in the "comments" table.
Is all that stored in a single table? Is a table with trillions of entries maintainable?
How does fetching comments for a particular video works? Shouldn't the latency be super high considering the data to look through?
If there is some indexing based on, say video_id, would the index be maintainable?
10 year old videos can have 10 year old comments as well as 1 day old comments. So, does partitioning play any kind of role here?
Another important thing would be comment threads. How are those maintained?
I tried googling the stuff but could not find anything relevant. Any help would be appreciated.