Storage system for randomly named blocks of data

A method for constructing an index suitable for indexing a large set of records identified by long generally randomly distributed record names and for answering membership queries about the set, the method comprising assigning each different record a different record name, determining that a new record name is not already represented in the index by checking an entry in a first level index that is shorter than the complete new record name, and determining that a queried record name is already represented in the index by determining that the queried record name is represented in a second level index that contains enough information to reconstruct the complete queried record name.