SPPS: A scalable P2P-based proximity-aware multi-resource discovery scheme for grids

Abstract

Grids are emerging as a novel approach of employing distributed computational and storage resources to solve large-scale problems in science, engineering, and commerce. Distributed hash table (DHT) middleware overlay has been applied to grids as a middleware for providing scalable multi-resource discovery. However, direct DHT overlay adoption breaks the physical locality relationship between nodes, making it difficult to discover physically close resources to requesters. Moreover, to achieve multi-resource discovery, some approaches relying on multiple DHTs need high DHT maintenance overhead and other approaches lead to imbalanced load distribution, resulting in low scalability. This paper presents a scalable P2P-based proximity-aware multi-resource discovery scheme (SPPS). It collects the resource information of physically close nodes together, and maps resource requests from requesters to the resource information pool of its physically close nodes. In addition, it relies on a single DHT and achieves balanced resource discovery load distribution, enhancing the system scalability. Simulation results demonstrate the effectiveness of SPPS in proximity-awareness, overhead reduction, and balanced load distribution in comparison with other approaches.