ReX: Representative extrapolating relational databases

Generating synthetic data is useful in multiple application areas (e.g., database testing, software testing). Nevertheless, existing synthetic data generators are either limited to generating data that only respect the database schema constraints, or they are not accurate in terms of representativeness, unless a complex set of inputs are given from the user (such as the data characteristics of the desired generated data). In this paper, we present an extension of a prior representative extrapolation technique, namely ReX [20], limited to natural scaling rates. The objective is to produce in an automated and efficient way a representative extrapolated database, given an original database O and a rational scaling rate, s∈Q. In the extended version, the ReX system can handle rational scaling rates by combining existing efficient sampling and extrapolation techniques. Furthermore, we propose a novel sampling technique, RVFDS for handling positive rational values for the desired size of the generated database. We evaluate ReX in comparison with a realistic scaling method, namely UpSizeR [43], on both real and synthetic databases. We show that our solution statistically and significantly outperforms the compared method for rational scaling rates in terms of representativeness.