An adaptive distributed index for similarity queries in metric spaces

Abstract

As the amount of data is growing rapidly, efficient and scalable index structures for managing large-scale data are attracting more and more attention. To efficiently query and manage the data in metric spaces, an adaptive distributed index, MT-Chord, is proposed. MT-Chord integrates Chord based routing protocol and M-tree based index structure to support efficient similarity query processing in metric spaces. Each index node has multiple replicas for load-balance and a cost model is presented to dynamically tune the number of replicas based on the query and update pattern at the granularity of each index node. MT-Chord is a truly scalable, efficient and adaptive distributed index structure for query processing in metric spaces, which is verified by our extensive experimental studies on three real-life datasets extracted from different data sources. © 2012 Springer-Verlag.