I just rewrote the server-side part of the site, this time using mongoDB instead of mysql.
While I can't speak for the performance of mongoDB, I did find the workflow to be very nice. There are three key things that make that the case: It's not a relational database, it's schema-less, and it uses BSON(Binary json).
Because it's not a relational database, I don't have to have to make as many seperate tables(or 'collections' in mongodb). For example, I put the comments in an array right inside each post. The data is still accessible out of context in the same basic way it is with a relational database: find the post based on id, then the comment based on id. You can also still search for a comment id directly without worrying about what post it's in.
MongoDB is schema-less, which means you don't have to spend the time to create a database structure beforehand (tables, columns and their type).
The coolest part to me is that it uses BSON for queries. This means the mongoDB driver can take data structures directly from my programming language, in this case perl, and use those as the query.
The three things combined means I'm working with a database more integrated with my programming language: The mindset is that I create a data structure in my programming language that represents what I want to store (such as a comment), and then I just store it. No need to create the structure beforehand, no need to stringify the data into a query, and no need to seperate it into chunks in a relational database style.