Mongo Sharding

• Approach to horizontal scaling - break up a collection onto multiple logical hosts
• Want to spread data around to different databases
• Split data up from a particular collection
• Each shard is a replica set (at least 3 each)
• Queries are distributed
• Router - mongos - used to distribute the queries to the proper host.  You can have multiple mongos instances.  The mongos instances are similar to a replica set.  They are stateless, and if one goes down, another takes over.
• range-based approach
• Shard Key
• Queries are sent to the mongos router from the application
• If query does not include the shard key, mongos will send to all shards to figure out location
• Inserts require the shard key
• data without a shard key will be stored in S0 (left-most shard database)

Read More..