Adapting the content native space for load balanced indexing

Abstract

Today, there is an increasing demand to share data with complex data types (e.g., multi-dimensional) over large numbers of data sources. One of the key challenges is sharing these data in a scalable and efficient way. This paper presents the design of ZNet, a P2P network for supporting multi-dimensional data. ZNet directly operates on the native data space, which is partitioned dynamically and mapped to nodes within the network. Particular attention is given to reduce load imbalance among nodes due to skewed data distribution. Results from an extensive simulation study show that ZNet is efficient in routing and processing range queries, and provides good load balancing. © Springer-Verlag Berlin Heidelberg 2005.