Cassandra vs. MongoDB vs. CouchDB vs. Redis vs. Riak Comparison

Mongo DB got a lot of traction very early on, despite some less than optimal defaults. As time went on, adoption grew, their marketing teams and support contracts also took them to the next level. I liked them early on, because it was such a natural pairing with Node on the server. The disconnect that you get from ORMs was gone in my mind, and despite the likes of Mongoose, I often don't see the point of the mappers, beyond basic data validation.

Cassandra is at the tops because it was a very early BigTable-like solution that came out of Facebook. It has many options to tweak, and scales near linearly. But at the cost of having to do for yourself what you might take for granted in other databases.

Couchbase is interesting, I also think that Redis has become interesting as well, but would suggest separate instances depending on the use case.

The one that I think doesn't get enough consideration is probably RethinkDB... from a dev-ops perspective it's hands down one of the better NoSQL databases to work against. From a dev perspective, the fluent API is sometimes better, other times harder than say MongoDB's, but it works. It's about as close as it gets to easy to deploy-scale in my mind, and would love to be able to use it more.

From my own experience, most of my data fits well in either traditional SQL, a bigtable-like solution where most queries are single-key (Cassandra) or an in-memory kv store such as with Redis. Mongo has traction, and RethinkDB is my preferred choice where I want tiered data in a single record (document store), which serves several other areas in between traditional SQL and the likes of Cassandra.

Another interesting one not covered is etcd...