A Study of Data Migration Efforts in NoSQL Databases

Despite the undeniable benefits of NoSQL databases in modern database management, workload on these systems constantly changes, so the schema used to design a NoSQL system becomes inefficient. It requires schema refactoring and data migration. Empirical evidences prove that these types of migrations operations are complicated and lack a research. The current paper contributes to the extensive research of factors that influence on migration of data between NoSQL schemas. After thorough literature review, the methodology of graph-based modelling was chosen to evaluate the corelation between structural complexity and query performance time. The case study started with an analysis of a SkyServer database and mapping of the queries into evolution graphs. This allowed to measure several metrics, as graph-based and entropy, that defines structural complexity, and make a juxtaposition between them and query performance metrics. The unexpected knowledge and insight gained by the analysis was that in NoSQL migration situation entropy and graph-based metrics do not provide a consistent direct statistical correlation between average time of execution of queries. A minor extension of a research caused an emergence of another insight: a positive corelation between two group of metrics, representing structural complexity from different perspectives. This insight proposed to use those features that could be easily extracted and linked more directly to the effort estimation.