Grid Service Discovery Based on Cross-VO Service Domain Model

Abstract

The diversity of grid service originates from heterogeneous and dynamic nature of grid, and it poses a great challenge to grid service discovery. How to discover services satisfying users’ multiple requests meanwhile avoiding negative effect derived from requests and updates becomes increasingly important in grid environment. This paper proposes a Cross-VO (Visual Organization) service domain model for compensating deficiencies that traditional approaches exhibit in flexibility of discovery. Service domain is developed to make advantage of similarity among services. In this model each service domain is constructed through all services that have similar function in VOs and nodes in service domain connect according to unstructured P2P system. It breaks through resource restriction in a VO and satisfies users’ requests in great extent while achieves favorable scalability and flexibility. Both theoretical analysis and experimental results indicate that this model performs efficiently in high discovery success ratio, low average hops and messages even with low density and small TTL. Compared with non-domain grid system via the same discovery success ratio, our model outperforms it in both average hops and messages.