On the use of query-driven XML auto-indexing

Autonomous index management in native XML DBMSs has to address XML's flexibility and storage mapping features, which provide a rich set of indexing options. Change of workload characteristics, indexes selected by the query optimizer's "magic", subtle differences in the expressiveness of indexes, and tailor-made index properties ask-in addition to (long-range) manual index selection-for rapid autonomic reactions and self-tuning options by the DBMS. Hence, when managing an existing set of indexes (i.e., a configuration), its cost trade-off has to be steadily controlled by observing query runtimes, index creation and maintenance, and space constraints.