Matthew Dillon’s written up a description of how indexing (via B+trees) will be handled in his as-of-yet-unnamed clustering file system.