Zookeeper
I recently needed to solve the following problem: Given a typical search application we would integrate multiple indexes with distinct content (e.g. news, images, web search results etc.). For performance reasons, each index would be available on a set of distinct machines. To make things even more interesting there may be serach servers sharded over many machines. During operation time there is some chance that one might wish to move, shut down or bring up additional indexes.
One component that helped a big deal solve this problem quickly was the Apache Hadoop sub project zookeeper. The API is really easy to understand: Getting to a prototype sort of proof of concept didn’t take more than an hour, maybe two. Polishing up the solution and integrating it with at least the most important existing systems made up for about one and a half of a week.
Thanks to all the zookeeper developers: For contributing the project to Apache, keeping up improving it and providing helpfull documentation (though there are still some “tbd” marks in some places).