A limited intermediate node buffering based RWA scheme in OBS backbone networks

Abstract

An all optical backbone Optical Burst Switched (OBS) network comprises of a multitude of optical transport sub-systems erected in commercial, residential as well as industrial ar-eas. The heterogeneous nature of the large volumes of traffic gen-erated by various applications and services ideally requires an op-tical backbone network infrastructure to accommodate it. Such a network must be continuously adaptable to the changing nature of the traffic as well as its spontaneous growth with time. In so doing, it has to ensure high end-to-end quality of service (QoS), availabil-ity as well as provision adaptable controllability in cooperation with peripheral (service) layer networks. To successfully design and deploy a cost-effective backbone network, consideration must be taken with regards to system configuration, as well as in applied devices manufacturing. This is to ensure that any component failure does not add any noticeable performance degradation as the network will quickly reconfigure itself accordingly. At operational level, ef-fective routing approaches are necessary to ensure minimized con-gestion as well as contention occurrences. The aggregation of both transit and local traffic at a node influences each other such as to aggravate congestion and to a certain extent reduce contention oc-currences (due to the streamline effect). In this paper, we propose a priority based intermediate Node Buffering based PIB-RWA scheme to combat the problem of contention occurrences and to prevent bursts discarding. It basically selects primary as well as deflection paths/links based on past contention frequency occurrences as well as current resources states in the candidate paths. Furthermore, the scheme also augments intermediate buffering provisioning for con-tending data bursts that are almost reaching the destination. Simu-lation results show that the scheme performs well in terms of key QoS metrics such as network throughput, data burst loss probabil-ities as well as load balancing.