Search

xisting peer-to-peer approaches for resource location based on distributed hash tables focus mainly on optimizing lookup query resolution. The underlying assumption is that the arrival ratio of lookup queries is higher than the ratio of resource publication operations. We propose a set of optimization strategies to reduce the network traffic generated by the data publication and update process when resources have dynamic-valued attributes. We aim at reducing the publication overhead of supporting multi-attribute range queries. We develop a model predicting the bandwidth reduction, and we assign proper values to the model variables on the basis of real data measurements. We further validate these results by a set of simulations. Our experiments are designed to reproduce the typical behaviour of the resulting scheme within large distributed resource location system, like the resource location service of the XtreemOS Grid-enabled Operating System.