Kafka is a distributed publish-subscribe system for processing large amounts of streaming data. A single Kafka broker can handle hundreds of megabytes of reads and writes per second from thousands of clients. It's scalable, durable, distributed by design and totaly awesome.
The script installs by default the latest kafka release, and running on Scala 2.9.2 (as opposed to the 2.8.0 by default). I haven't tested it with newer versions of Scala, so it might break. Feel free to report if you have experiences.
The default configuration installs it as a single broker setup running on localhost, under process supervision by supervisord.
To set it up on/as a cluster, you do need to adapt it, but it should be pretty easy starting from here.