On the impact of relocation on network dimensions in resilient optical Grids.

Optical networks, with their high bandwidths and low latencies, are excellent to support Grid applications, esp. those arising in so-called eScience applications. In this paper, we consider the resulting so-called optical grids.We address the issue of resiliency against network failures and show how the Grid-specific anycast principle can be exploited in providing shared protection. Since in Grid scenarios users generally allow the Grid system to decide upon the location where jobs are executed, we exploit relocation to alternate backup sites in case of failures. We provide integer linear programming (ILP) solutions to the resulting routing and wavelength assignment problems, as well as a scalable heuristic. A case study on a European network topology shows that this relocation allows savings in the order of 16% of total wavelength capacity compared to traditional shared path protection.