"Before Clustrix an entirely different kind of datastore, such as MongoDB, would be used. This requires developers to split up their data across multiple server

"Before Clustrix an entirely different kind of datastore, such as MongoDB, would be used. This requires developers to split up their data across multiple servers and use a map reduce algorithm in order to query across all the data." Uhm, no. Sharding is orthogonal to map-reduce, and you wouldn't use map-reduce on MongoDB or otherwise for anything OLTP. (MR on Mongo is SLOW) Something that makes sharding easier/more automagic is nice though. I hate sharding anything that isn't Redis or MongoDB.