Beyond 100 million entities: large-scale blocking-based resolution for heterogeneous data