The success of NoSQL DBMSs has pushed the adoption of polyglot storage systems that take advantage of the best characteristics of different technologies and data models. While operational applications take great benefit from this choice, analytical applications suffer the absence of schema consistency, not only between different DBMSs but within a single NoSQL system as well. In this context, the discipline of data science is steering analysts away from traditional data warehousing and toward a more flexible and lightweight approach to data analysis. The idea is to perform OLAP analyses in a pay-as-you-go manner across heterogeneous schemas and data models, where the integration is progressively carried out by the user as the available data is explored. In this paper, we propose an approach to support data analysis within a high-variety multistore, with heterogeneous schemas and overlapping records. Our approach supports relational, document, wide-column, and key-value data models by automatically handling both data model and schema heterogeneity through a dataspace layer on top of the underlying DBMSs. The expressiveness we enable corresponds to GPSJ queries, which are the most common class of queries in OLAP applications. We rely on nested relational algebra to define a cross-database execution plan. The system has been prototyped on Apache Spark.

No abstract

Multistores are data management systems that enable query processing across different and heterogeneous databases; besides the distribution of data, complexity factors like schema heterogeneity and data replication must be resolved through integration and data fusion activities. Our multistore solution relies on a dataspace to provide the user with an integrated view of the available data and enables the formulation and execution of GPSJ queries. In this paper, we propose a technique to optimize the execution of GPSJ queries by formulating and evaluating different execution plans on the multistore. In particular, we outline different strategies to carry out joins and data fusion by relying on different schema representations; then, a self-learning black-box cost model is used to estimate execution times and select the most efficient plan. The experiments assess the effectiveness of the cost model in choosing the best execution plan for the given queries and exploit multiple multistore benchmarks to investigate the factors that influence the performance of different plans.