Riak is an open-source, distributed key/value database for high availability, fault-tolerance, and near-linear scalability. In short, Riak has remarkably high uptime and grows with you.
As the modern world stitches itself together with increasingly intricate connections, major shifts are occurring in information management. The web and networked devices spur an explosion of data collection and access unseen in the history of the world. The magnitude of values stored and managed continues to grow at a staggering rate, and in parallel, more people than ever require fast and reliable access to this data. This trend is known as Big Data.
Aside: So What is Big Data?
There's a lot of discussion around what constitutes Big Data. I have a 6 Terabyte RAID in my house to store videos and other backups. Does that count? On the other hand, CERN grabbed about 200 Petabytes looking for the Higgs boson. It's a hard number to pin down, because Big Data is a personal figure. What's big to one might be small to another. This is why many definitions don't refer to byte count at all, but instead about relative potentials. A reasonable, albeit wordy, definition of Big Data is provided by Gartner:Big Data are high-volume, high-velocity, and/or high-variety information figures that require new forms of processing to enable enhanced decision making, insight discovery and process optimization.
The sweet spot of Riak is high-volume (data that's available to read and write when you need it), high-velocity (easily responds to growth), and high-variety information figures (you can store any type of data as a value).
Riak was built as a solution to real Big Data problems, based on the Amazon Dynamo design. Dynamo is a highly available design---meaning that it responds to requests quickly at very large scales, even if your application is storing and serving terabytes of data a day. Riak had been used in production prior to being open-sourced in 2009. It's currently used by Github, Comcast, Voxer, Disqus and others, with the larger systems storing hundreds of TBs of data, and handling several GBs per node daily.
Riak was written in the Erlang programming language. Erlang was chosen due to its strong support for concurrency, solid distributed communication, hot code loading, and fault-tolerance. It runs on a virtual machine, so running Riak requires an Erlang installation.
So should you use Riak? A good rule of thumb for potential users is to ask yourself if every moment of downtime will cost you in some way (money, users, etc). Not all systems require such extreme amounts of uptime, and if you don't, Riak may not be for you.