Verifying the performance benefits of caching probabilities in probabilistic databases

Probability calculation has been observed to dominate performance for some queries in probabilistic databases. Current probability storage structure is suspected to be sub-optimal. This paper focuses on researching possible performance benefits of caching probabilities in probabilistic databases, focused on the probabilistic database management system DuBio. Controlled experiments have been run to measure performance benefits while accounting for different queries, database structures and server delays. This paper presents results showing under which circumstances caching is advisable.