Business process management repositories manage large collections of process models ranging in the thousands. Additionally, they provide management functions like e.g. mining, querying, merging and variants management for process models. However, most current business process management repositories are built on top of relation database management systems (RDBMS) although this leads to performance issues. These issues result from the relational algebra, the mismatch between relational tables and object oriented programming (impedance mismatch) as well as new technological developments in the last 30 years as e.g. more and cheap disk and memory space, clusters and clouds. The goal of this paper is to present current paradigms to overcome the performance problems inherent in RDBMS. Therefore, we have to fuse research about data modeling along database technologies as well as algorithm design and parallelization for the technology paradigms occurring nowadays. Based on these research streams we have shown how the performance of business process management repositories could be improved in terms of loading performance of processes (from e.g. a disk) and the computation of management techniques resulting in even faster application of such a technique. Exemplarily, applications of the compiled paradigms are presented to show their applicability.