Ant colony optimization for the routing of jobs in optical grid networks

Grid networks provide users with a transparent way to access computational and storage resources. The introduction of (dense) wavelength division multiplexing techniques have made optical networks the technology of choice for data-intensive grid traffic. In a grid network scenario, users are generally more interested in the successful completion of their jobs than in the location where the actual processing occurs. Job routing and scheduling in current generation grid networks are managed by resource brokers, which assign each job to a resource and route the job in a unicast way. An anycast approach using grid-aware network algorithms would bypass the need for a resource broker and increase scalability. We propose several anycast algorithms for job routing in optical grid networks, based on the concept of ant colony optimization, which draws parallels between the behavior of ants gathering food and the routing of packets inside a network. Simulation results show an increased performance of our algorithms over more classical unicast-based protocols, even though this is accompanied by a slight increase in complexity.