DPLA uses the Linked Data Platform (LDP) standard to manage harvested content and metadata. Our LDP server and RDF repository runs on Apache Marmotta, backed by a PostgreSQL database. Broadly, we handle two kinds of resources through LDP: DPLA MAP records (as LDP RDF Sources with type ORE::Aggregation), and binary "Original Records" (as LDP non-RDF Sources).
See also KriKri RDF Database Review