Towards consumer-oriented photonic Grids

Grid technology is currently reaching a level of maturity allowing deployment of large test infrastructures world wide, leveraging huge computing facilities to the academic and scientific community (important examples of such infrastructures include the EGEE [1] and TeraGrid [2] deployments in Europe and the US, respectively). Two important trends can however be discerned in current grid developments: the tendency to deploy dedicated optical network facilities [3] (instead of sharing network connectivity with other applications) and the adoption of grid technology beyond the e-science community. Especially the envisaged usage of grid-aware applications by home-users entails developments towards more scalable and hence distributed resource management schemes. Current grid systems are indeed specialized in handling (extremely) large jobs submitted by a relative small number of users. Future applications including multimedia editing [4], online gaming and virtual reality immersion [5] have the potential for an enormous user base. The number of jobs launched will not only increase because of the increase of users but also due to higher degrees of parallelism (resulting in a larger number of smaller jobs) typically encountered in multimedia applications. These new application areas for grids result in novel requirements: scalable job management (centralized and hierarchical approaches currently adopted do scale to the envisaged number of service requests), high connectivity for end users, provisioning of computing facilities for residential users, offering guarantees on service levels (service availability and guarantees on job completion time) and flexible resource management (limited or ideally no resource configuration to limit the overhead relative to the job execution time).