"First, there are many scenarios in which that server loses all its data no matter what. If there is water damage, fire, some hardware problems, etc..." Err!

"First, there are many scenarios in which that server loses all its data no matter what. If there is water damage, fire, some hardware problems, etc..." Err! So there are burglers and other evildoers sneaking through the city. Does that mean I should stop locking my door, since burglaries happen anyway? "In the real world, traditional durability often isn’t even done correctly. If you are using a DBMS that uses a transaction log for durability, you either have to turn off hardware buffering or have a battery backed RAID controller." Or you write to raw devices, which perform pretty neatly. The fact that a _lot_ of databases are set up and managed in a crappy way justifies that this is considered the norm? I don't think so. Here's for the disclaimer. I don't know anything about MongoDb and not a helluvalot about NoSQL databases. But I think the reasoning behind this post is not very sound. No matter which database.