Query-oriented Relaxation for Cardinality Assurance

Although a large number of queries used in applications ranging from web search to business intelligence have associated cardinality constraints, current database engines have minimal support for ensuring query cardinality. This leads to two main types of problems: the empty result-set problems and the too few/too many problems. Inability to meet query cardinality constraints requires the user to undertake a frustrating trial-and-error process that can be extremely cumbersome and time-consuming. In fact this process wastes system resources without any guarantee of success. A possible strategy for solving the cardinality assurance problem is query relaxation. However, obtaining exact cardinality assurance with query relaxation has been proven to be NP-Hard. In this work we present QRelX, a novel algorithm for cardinality assurance using query relaxation. QRelX not only relaxes queries efficiently but also ensures that result queries have minimal relaxation. Our algorithm uses QXForm, a query space transformation framework, to enable the relaxation of join and select queries. Following this transformation, QRelX uses layer-based navigation and incremental cardinality estimation to find result queries that meet the expect cardinality but are also close to the original query. Our preliminary experimental results indicate that QRelX is computationally more efficient than traditional cardinality assurance algorithms and it indeed successfully finds minimal relaxation queries.