Analyzing data quality trade-offs in data-redundant systems

Abstract

For technical and architectural reasons data in information systems are often redundant in various databases. Data changes are propagated between the various databases through a synchronization mechanism, which ensures a certain degree of consistency. Depending on the time delay of propagating data changes, synchronization is classified in real time synchronization and lazy synchronization in case of respectively high or low synchronization frequency. In practice, lazy synchronization is very commonly applied but, because of the delay in data synchronization, it causes misalignments among data values resulting in a negative impact on data quality. Indeed, the raise of the time interval between two realignments increases the probability that data result incorrect or out-of-date. The paper analyses the correlation between data quality criteria and the synchronization frequency and reveals the presence of trade-offs between different criteria such as availability and timeliness. The results illustrate the problem of balancing various data quality requirements within the design of information systems. The problem is examined in selected types of information systems that are in general characterized by high degree of data redundancy. © 2008 Physica-Verlag Heidelberg.