Storing rare numeric data (e.g. inverted index) - any conventions?

Is there an acceptable way of storing and accessing sparse numerical data (e.g. inverted index / search engine term on a matrix of documents)? RDBMS seems unacceptable for this type of data, but it would be nice if it was stored in some kind of database (saved to disk, worked as a server, etc.). Is there an acceptable solution for this kind of problem (for example an existing database capable of supporting such a model)? Does anyone know how Google stores and accesses its indexes so quickly?