Hash tables are a ubiquitous component of computer science. They offer a wide range of functionality such as query(key), insert(key), etc. How can such a data structure be stored in a particularly space-efficient way? The question becomes particularly interesting when only a subset of the functionalities are supported, which can result in astonishing space savings. For this reason, my research deals with so-called retrieval structures and perfect hash functions. Orthogonally, I work on both CPU and GPU parallelization.