Percolator is a feature in Elasticsearch 0.15 that allows users to register queries against an index and later send documents to identify which queries match them, effectively reversing the traditional search process. Instead of indexing documents and running queries against them, users can register queries and then percolate documents to find matching queries. This feature can be particularly useful for scenarios like monitoring real-time data streams, such as tweets containing specific keywords. Queries are stored as documents in a special _percolator index, which provides persistence and the ability to filter queries using additional criteria. The percolation process leverages all the shards of an index for efficient processing, and the local execution preference enhances performance. Users can also include filtering criteria to narrow down query matches, and increasing the number of replicas can enhance percolation capabilities.