Multi-Dimensional Range Querying Using a Modification of the Skip Graph

摘要

Skip graphs are an application layer-based distributed routing data structure that can be used in a sensor network to facilitate user queries of data collected by the sensor nodes. This research investigates the impact of a proposed modification to the skip graph proposed by Aspnes and Shah. Nodes contained in a standard skip graph are sorted by their key value into successively smaller groups based on random membership vectors computed locally at each node. The proposed modification inverts the node key and membership vector roles, where group membership is computed deterministically and node keys are computed randomly. Both skip graph types are modeled in Java. Range query and node mobility simulations are performed. The number of skip graph levels, total node count, and query precision are varied for query simulations; number of levels and total node count is varied for the mobility simulation. Query performance is measured by the number of skip graph messages used to execute the query while mobility performance is measured by the number of messages transmitted to maintain skip graph coherence. When the number of levels is limited and query precision is low, or when query precision is matched by the number of levels in the skip graph and total network node counts are increased, the modified skip graph transmits fewer messages to execute the query. Furthermore, fewer update messages are needed to fix lost node references due to mobile nodes.