Bridging Network Monitoring and the Grid

Several network monitoring frameworks have been successfully deployed in various network domains, but they typically expose proprietary interfaces and feature poor interoperability. In addition, frameworks either focus on end-to-end or on domain edge-to-edge measurements. Consequently, they can only address a partial spectrum of Grid performance characteristics. As Grids are typically large-scale infrastructures spanning multiple network domains, the seamless view that is paramount to Grid middleware and operators, can be severely hindered. We analyse the requirements of important Grid middleware components: job scheduling and data transport. We propose a set of services for the publishing, discovery and gathering of information about network performance and framework capabilities. These expose interfaces that follow the Global Grid Forum recommendations. Finally, we review the implementation details of a prototype of the proposed architecture.