Ah, yes... and 640kB of RAM was supposed to be enough, too. Or 4B IP addresses.
If anything, my experience is whenever somebody sets close hard limits like that somebody else is going to pay for it.
The service I work on creates up to couple millions of unique objects per second, on a single node in a batch process. It tries to do a large batch of work as fast as is possible. It fetches I think something like 2GB of data per second from MongoDB node, processes it and creates up to 5 million objects in span of 2-3 seconds. We also have realtime services that are about 20-30k TPS per node.
We are currently using MongoDB and it uses plenty bits for that:
If anything, my experience is whenever somebody sets close hard limits like that somebody else is going to pay for it.
The service I work on creates up to couple millions of unique objects per second, on a single node in a batch process. It tries to do a large batch of work as fast as is possible. It fetches I think something like 2GB of data per second from MongoDB node, processes it and creates up to 5 million objects in span of 2-3 seconds. We also have realtime services that are about 20-30k TPS per node.
We are currently using MongoDB and it uses plenty bits for that:
https://docs.mongodb.com/manual/reference/method/ObjectId/