Storage scheme of video viewing records of users in station B

At noon B When I'm standing there watching the video , Suddenly curious B How to save the video viewing record of the station .
First , Because the video records viewed by multiple terminals are synchronized , That is, whether you watch it on your mobile phone , Or web viewing , Or a new device to watch , Videos can jump directly to the last viewing position , So this record should not be stored locally .
secondly ,B The users of the station are up to 2020 end of the year ,MAU（ Monthly living ） reach 2.02 Billion , Total number of videos 2020 According to statistics, it is about 7 Ten million , Suppose you press 1 100 million pieces , Even on a per capita daily basis 100 A video , that 10 Every user also has 36.5 Ten thousand records . How to store such a large sparse matrix ？ Traditional database is obviously not suitable .
My first thought should be HBase, Single instance supports one billion rows and one million columns . Because video content is growing infinitely , And users are limited , So you can take the user id As rowkey, Video content id As column To record the viewing position of each video of the user .
But the number of videos has now reached 100 million , And it will continue to grow , A million trains is obviously not enough . This may require consideration of multiple instances , That is, put the video content id According to a certain hash The rules （ Or directly in order , Easy to expand ） Stored in different instances , such as 1 One hundred million videos are saved for each instance 100 Ten thousand , branch 100 An example , When querying the user's video viewing location , Just according to the video id Go to the specified instance first , Then through the user id Primary key + Video content id Column query location record .
however , After thinking about it, I feel that this is a relatively clumsy scheme .
The above is just a question that pops up in my mind , Let's keep a record of , Specifically B Station or other video websites （ For example, Teng Youai ） What technical scheme is adopted , I will continue to search and think .