Most people label Riak a key/value store, as does the author, but I get a more useful mental picture thinking of it as a distributed file system. The same kind

Most people label Riak a key/value store, as does the author, but I get a more useful mental picture thinking of it as a distributed file system. The same kinds of things that are fast on a file system are fast with Riak (direct access), and the same kinds of things are relatively slow on both (find(1)ing a file based on a pattern match compared to M/R for data). They both also excel at storing binary data assuming you understand the limits. This model has helped me make better decisions about when & where to use it and the cost involved. For example, if you're on AWS, just use S3 and keep metadata in an RDBMS or search engine.