Optimized dimensioning of resilient optical grids with respect to grade of services

Abstract

An optical grid network geographically integrates distributed computing/information resources with high speed communications. Network dimensioning, maximization of services, and job scheduling are some of today key arising issues in optical grids. Since the last decade, many projects have been conducted in order to provide computational and information facilities in the academic as well as in the business communities. In this paper, we study the network dimensioning and the maximization of IT services in optical grids. We propose a scalable optimization model for maximizing IT services under link transport capacities. We assume the use of the anycast routing principle to identify the server nodes for executing the jobs, and a shared path protection mechanism in order to offer protection against single link/node failures. We also investigate different calculation methods of the link transport capacities in order to maximize the grade of services, while taking into account the bandwidth requirements. Computational results are presented on different traffic distributions. They show that the proposed link dimensioning can save more than 35 % bandwidth in optical grid networks, in comparison with the classical link dimensioning strategies. We also investigate the different protection schemes against single link failures, single node failures, single node and server node failures, and compare their bandwidth requirements, as well as their impact on the grade of services (GoS). Results show that there is no significant increase of the bandwidth requirements and no meaningful impact on the GoS when moving from a single link protection scheme to a single node (including server nodes) protection scheme.