DistriPlan: an optimized join execution framework for geo-distributed scientific data

Scientific data is frequently stored across geographically distributed data repositories. Although there have been recent efforts to query scientific datasets using structured query operators, they have not yet supported joins across distributed data repositories. This paper describes a framework that supports join-like operations over multi-dimensional array datasets that are spread across multiple sites. More specifically, we first formally define join operations over array datasets and establish how they arise in the context of scientific data analysis. We then describe a methodology for optimizing such operations—components of our approach include enumeration algorithms for candidate plans, methods for pruning plans before they are enumerated, and a detailed cost model for selecting the best (cheapest) plan. We evaluate our approach using candidate queries, and show that the optimization effort is practical and profitable—query performance was improved significantly using our approach.