Realtime Search at Twitter

At Twitter we're indexing hundreds of millions of tweets per day, and are serving more than 2 billion queries per day. The engine behind this immense load uses a modified Lucene version, which is optimized for real-time search. This talk will explain the modifications we made to Lucene, such as highly concurrent data structures, lock-free algorithms, improved garbage collection performance, etc.

Watch the video of Michael Busch's talk here.

Slides: 

Realtime Search at Twitter-mbusch-bbuzz12.pdf