B-trie

The B-trie is a trie-based data structure that can store and retrieve variable-length strings efficiently on disk. [ Citation | title=B-tries for Disk-based String Management | first1=Nikolas | last1=Askitis | first2=Justin | last2=Zobel | year=2008 | isbn=1066-8888 (Print) 0949-877X (Online) | pages=pp. 1-26 | url=http://www.springerlink.com/content/x7545u2g85675u17/ | journal=VLDB Journal ]
The B-trie was compared against several high-performance variants of B-tree that weredesigned for string keys. It was shown to offer superior performance, particularly under skew access (i.e., many repeated searches). It is currently a leading choice for maintaining a string dictionary on disk, along with other disk-based tasks, such as maintaining an index to a string database or for accumulating the vocabulary of a large text collection.

References