Optical burst switching (OBS) has been proposed as the next generation optical network for grid computing. In this paper, we envision a heterogeneous Grid served by an Optical Burst Switching framework, where grid traffic co-exists with IP and/or a 10 GE based traffic to achieve economy of scale. This paper addresses the latency that Grid jobs experience in OBS networks. The injection of jumbo size grid jobs can potentially affect the latency experienced by IP/10GE traffic. Simulation results have shown that in Grids served by an optical burst switch, grid jobs consistently have lower latency than co-existing IP/10GE traffic, with a slightly elevated latency of IP/10GE traffic when the size of grid jobs increases. We conclude that given the fact that OBS can efficiently handle the enormous amount of bandwidth made available by DWDM technology, Grid over Optical Burst Switching is a cost effective way to provide grid services, even for latency sensitive grid computing applications.
Yuhua Chen, Wenjing Tang, Pramode K. Verma