Jeff Kao

So if you need to rollback, you just switch everything over and all the data pointers also switch back.

And so that's a way to rollback.

And you don't have to reason about these many states.

It's just one self-contained thing.

And for that to work, you really do need something that's very efficient, essentially, both in

Because they're shipping so many data assets.

Right, because it's almost like, hey, you don't even need an external database for something like you would typically grab for in web apps like a Postgres or MySQL.

It's, hey, I'm going to use an embedded database or have a large in-memory index.

And that's your sort of state.

And so all that ships together in one whole unit or package.

So we use a couple of things.

And so we talk about this in, you know, the blog post that describes Verizon DB.

Yeah.

I would say our primary storage mechanism is this thing called RocksDB, which is an embedded storage or embedded database.

And the way people think about it is it's essentially the database that you use to build databases.

So, for example...

Facebook or Meta, they originally use this to power the storage layer of MySQL for them.

And I think there's a project called MyRocksMongoDB, for example.

They actually use RocksDB for some time.

If you look at CockroachDB, their storage layer for a long time was also RocksDB.