Design and Implementation of the Modified R-Tree Structure with Non-blocking Querying

In highly concurrent field such as location based services, massive objects are moving concurrently. Due to continuously changing nature of their location, traditional indexes cannot provide the real-time response since query processing is frequently blocked by node-split or region propagation as the locations of objects change. In this paper, the modified R-tree structure with lock-free querying for multi-dimension data, Rver-tree, is proposed. Basically, Rver-tree uses the new versioning technique. When updating data such as key update(region propagation) and index restructure(node-split), it never physically modify original data, rather creates new version for compensating data intactness. Due to data intactness, search operation can access data without any locking or latching by reading old version. In the performance evaluation, it is proven that search operation of Rver-tree is at least two times faster than a previous work.