Research On The Congestion Control Technologies In Intelligent Optical Networks

The essential problem of optical network congestion control is to minimize network congestion under certain network resources. The emergence of dynamic and intelligent optical networks brings new problems to network congestion control technology, however, provides a vaster space for their development. In this dissertation, congestion control technology is systematically studied in both single-layer wavelength-routed networks (WRN) and multi-layer optoelectronic hybrid networks, where theoretical analysis, software simulation, and hardware experiments are utilized synthetically. A soft preemptive scheme, a wavelength buffering scheme, a postponed short lightpath teardown strategy and a wavelength-alternation-based concurrent collision avoidance scheme are proposed in the dissertation, all of which can enhance overall performance of the optical networks.The uneven space distribution characteristic of optical network is one important reason that leads to low network utilization. In order to overcome and utilize this network characteristic, a soft preemptive scheme (SP) is proposed in the dissertation. This scheme allows new connection requests "soft preempt" existing connections when they fail to find a route. Before the preemption, the system will calculate new routes for the connections to be preempted. Only when these connections are switched to the new routes can the preemption be operated. In this way, the quality of service (QoS) of existing connections can be guaranteed. Simulation results show that the soft preemptive scheme can reduce overall network blocking probability by more than one order of magnitude under a light traffic load. The dissertation further studies on the combination of soft preemption and wavelength conversion. The new scheme helps the network obtain a better blocking performance.The dissertation analyzes the uneven time distribution characteristic of optical network and finds that the instantaneous network utilization is very bursty under light loading. In order to utilize this characteristic, a wavelength buffering scheme (WB) is proposed in the dissertation. To make better utilization of network resources, this scheme reserves a portion of wavelength resources under light loading and releases them when the load goes up. The dissertation further proposes a resource buffering scheme (RB) which works in multi-layer networks at sub-wavelength granularity level. Simulation results show that both buffering schemes achieve better blocking performance and the resource buffering scheme performs comparatively better.The "local feature" of the dynamic routing optimizing algorithms restricts them to achieve a better possible performance. To solve this problem, a postponed short lightpath teardown (PLT) strategy is proposed to take global optimizing factors into dynamic routing and wavelength assignment (RWA). This strategy postpones the teardown process of the lightpaths with small physical hop number in a multi-layer network when there is no traffic on them. In this way, short lightpaths are reserved and a comparatively stable and global optimal virtual topology (VT) is constructed to reduce the overall blocking probability. Meanwhile, this strategy can also reduce the signaling flow of the optical control plane significantly.For distributed optical networks (DON), resource collisions brought by concurrent connection requests may lead to a sharp increase of blocking or even a network collapse. To solve this problem, the dissertation proposes a wavelength-alternation-based (WA) concurrent collision avoidance scheme, which tries to calculate routes for concurrent connection requests in different wavelength channels. This scheme can notably avoid link resource collisions during the establishment process of concurrent connections without changing current network's routing protocol, and further reduce network blocking probability.