Empirical evaluation of guarded structural indexing

Traditional indices in relational databases are designed for queries that are selective by value. However, queries can also retrieve records on their relational structure. In our research, we found that traditional indices are ineffective for structurally selective queries. To accelerate such queries, socalled 'structural indices' have been applied in graph databases. These indices group together structurally similar nodes to obtain a compact representation of the graph structure. We studied how structural indices can be applied in relational databases and evaluated their performance. Guarded bisimulation groups together relational tuples with similar structure, which we use to obtain a guarded structural index. Our solution requires significantly less space than traditional indices. At the same time, it can offer several orders of magnitude faster query evaluation performance.SCOPUS: cp.p