Although I am a heavy user/believer of MongoDB, one caveat that is overlooked in this article is the administration part of Mongo Clusters.
Even the simplest replication needs 3 servers, add sharding to the dance for extra performance and the server counter jumps up.
For startups this is a major decision to consider as a full-time ops guy isn't always affordable.
Luckily, PaaS services as MongoLabs and MongoHQ save the day.
You could use replication (master/slave) with 2 servers and you don't have to use sharding for anything. How is MongoDB not the same as every other database ?
If anything MongoDB is by far the easiest database I've ever used for setting up clustering/sharding.
Master / slave is deprecated. Replica sets can run on 2 nodes however, running on an even number of nodes in a set is a very bad idea as a primary node failure will cause the rest of your nodes to go read-only. Always run an odd number. Avoid master / slave.
Even the simplest replication needs 3 servers, add sharding to the dance for extra performance and the server counter jumps up. For startups this is a major decision to consider as a full-time ops guy isn't always affordable. Luckily, PaaS services as MongoLabs and MongoHQ save the day.