On Fully Decentralized Resource Discovery in Grid Environments

摘要: Computational grids provide mechanisms for sharing and accessing large heterogeneous collections of remote resources such as computers, online instruments, storage space, data, applications. Resources are identified based on a set desired attributes. Resource attributes have various degrees dynamism, from mostly static attributes, like operating system version, to highly dynamic ones, network bandwidth or CPU load. In this paper we propose peer-to-peer architecture resource discovery in collection resources. We evaluate request-forwarding algorithms fully decentralized architecture, designed accommodate heterogeneity (in both policies types) dynamism. For this, build testbed that models two usage characteristics: (1) distribution peers, varies the number frequency shared resources; (2) requests patterns analyzed our up 5000 where each peer provides information about at least one resource. learned approach is not only desirable administrative reasons, but it also supported by promising performance results. Our results allow us characterize correlation between characteristics.