But the cost of a server big enough to handle a million simultaneous users can be cost prohibitive if you can get away with a few smaller nodes and something that scales horizontally... not to mention failover/redundancy which can be downright expensive and/or painful in many SQL variants.
Re:The Marketing Behind MongoDB
by manigandham on Wednesday August 30, 2017 @12:36AM
(#15129757)
That's the whole point - worry about that when you actually have a million simultaneous users, which is a huge number. There are plenty of options like reading from replicas, using a caching layer, or moving specific data to another datastore - all work that's better done when you actually have success and the resources to understand your data access patterns and implement a smart solution when it's needed.
Basic replication has been part of every major relational database for decades. Where is it expensive or painful? Do you really need 100% uptime? There's a vast ecosystem for relational databases that can help with all of this and now you have CitusDB, MemSQL and CockroachDB if you need built-in next-gen scaling and availability.
ORMs can do 95% of most project data access needs and a little bit of SQL knowledge allows for easily creating and editing tables - or use the numerous GUI tools out there.
If anything, SQL actually lets you be more agile by allowing for quick queries to fix data or do a complicated 1-time report without having to write up and deploy code. This probably has more to do with the SQL language itself which is why it's surprising that MongoDB never tried to implement any of it.