Here’s a nice introduction by Todd Huff to the topic of graph databases: what they are, and why they’re relevant. The author starts by trashing all of the candidates:

So relational database can’t handle complex relationships. Graph systems are opaque, unmaintainable, and inflexible. OO databases loose [sic] flexibility by combining logic and data. Key-value stores require the programmer to maintain all relationships. There, everybody sucks 🙂

And then Todd gets into a nice discussion of one graph database, Neo4j. He cites a piece comparing Neo4j with Hadoop. Hadoop’s great for shallow data reductions, like log processing, but really bad for deep relationships.
And don’t just read this piece; bookmark it! Because at the end, Todd includes an excellent bibliography of related articles.