Business design database with fast and large data updates

Tens of millions to hundreds of millions of data records
Device ID table sharding, by segment or modulo
Primary Key with long time
Finally, remove the auto-incrementing column.

If the concurrency exceeds a certain level (for example, 100 times per second)
Read/write splitting starts.
Primary database with no index write only
Create an index from the database and adjust the index size

If the concurrency is large (for example, writing-times per second or more), this is the case:
Raw data is stored using aof log
And then asynchronously stored in the database
Database re-read/write splitting

After writing logs, multiple workers write data to the database concurrently, almost instantly.

To put it bluntly, the key to a system of this scale is indexing.
Using relational databases is a fast and cost-effective index system.

 

Data archiving should also be done. cold/hot data is separated, and old data is merged according to different dimensions.