Hi Felix, nice and very useful article that summarizes a lot of work on your part!
A few things that might be worth correcting in your table:
- Most if not all of the products support ACID properties for single record transactions, and what they lack is multi-record ACID. In most of these products the data for a given record in many cases “fits” within a single document/row, and so the need for multi-record ACID is greatly diminished. I’m not sure ACID is really binary the way your table suggests, especially in light of the distributed nature of these products.
- MongoDB does not support your definition of “update anywhere” — at any given time, there is a single primary copy of the data for writes.
- MongoDB provides an in-memory storage engine, but this is not open source (in case this matters for your analysis).
- MongoDB seems to meet your definition of elastic — you can add and remove shards at any time, and the data is automatically distributed across nodes (see this link and this one).
- Cassandra does not support full text search. Datastax packages Cassandra with Solr in their commercial product, but I don’t think that’s quite the same thing as a native database feature.
- Cassandra does not support secondary indexes in the way other products in your analysis do — there are many caveats to their use, see this link.