Slotted store-and-forward optical circuit-switched networks: A performance study

Optical switching techniques have benefited greatly from slotted network/node operations in terms of network performance, resource utilization, and computational complexity, especially for networks that are offered short, interactive flows. When we consider the delivery of bulk data flows, with each flow consuming a significant portion of the link bandwidth for a considerably long duration (of hours or days), then such fine granular slotted operations may be inefficient and unnecessary. In this paper, we investigate the applicability of coarser granular slotted network operations in networks where the majority of traffic is contributed by bulk data flows, and we consider how to find appropriate slot sizes, given the traffic demand distributions. Our study reveals that the choice of slot size is in fact a trade-off between bandwidth utilization efficiency and the size of the reservation window. We also find that for a particular example network scenario, when the slot size is equal to half of the mean flow size, the probability of request blocking can be reduced from 0.076 to 8.5 ×10
⁻⁵
, and the number of network reconfigurations can be reduced by a factor of five, compared with the un-slotted case.