Techniques are described herein for efficient and scalable processing of complex sets of XML schemas. The techniques described herein provide for reducing duplication of schema elements in volatile memory by building an XML schema in-memory model that stores repeating schema elements in in-memory data structures that are separate from in-memory data structures that store the parent schema elements which logically include or otherwise refer to the repeating schema elements. The techniques described herein also provide for faster generation of an in-memory model of an XML schema by pre-loading, in data structures on persistent storage, of schema elements from dependent XML schemas that are referenced and/or incorporated by the XML schema. The techniques described herein also provide for efficient processing of inter-dependent XML schemas by tracking all unresolved schema elements from dependent XML schemas and freeing the portions of volatile memory, which are used to process schema elements from the dependent XML schemas, as soon as the dependent schema elements being processed are stored in data structures on persistent storage.