Index Compression algorithms used by Google

PForDelta is another interesting approach that isn't mentioned here. It claims slightly better compression than VByte, and significantly faster (~2x) decompression on modern superscalar processors:

http://www2008.org/papers/pdf/p387-zhangA.pdf

I've yet to try it in practice.

Obligatory: http://nlp.stanford.edu/IR-book/information-retrieval-book.h...

Edit: link to compression intro page (very readable) http://nlp.stanford.edu/IR-book/html/htmledition/index-compr...