Designers of relational databases often face anomalies i.e. problems preventing such databases from normal performance. Studying and eliminating the anomalies constitutes the main aspect of the logical stage of databases designing as a whole because it allows for significant improving the earlier designed relational model. Unfortunately, this stage is the least automated one, and existing anomalies display their negative effect when the database is filled up and functioning already. This paper proposes a databases restructuring technique based on graph theory, n-tuple algebra and parallel computations.