Study On Several Key Technologies For Burst Loss Control In Optical Burst Switching Networks

There are three typical schemes in optical switching field: OCS(Optical Circuit Switching), OPS(Optical Packet Switching) and OBS(Optical Burst Switching). Compared with OCS and OPS, OBS performs better due to the following two main characteristics. The first is that the OBS adopts the one-way reservation scheme and uses smaller switching granularity, so the transmission delay and bandwidth utilization of OBS are better than those of OCS. The second is that the burst control packet and burst data packet in OBS are separated in time and space domain(there is an offset time between them, and they are transmitted on different wavelengths), and hence the burst control packet can be processed via mature electronic technologies at every forwarding node after O/E conversion. OBS can make the burst data realize end-to-end transparent transmission in an all-optical manner without the need of optical memory and optical logical processor, which overcomes the disadvantages of OPS. Therefore, OBS is becoming a promising technique for future optical switching networks. Some large companies, such as Huawei Technologies Co., Ltd., keep investing a lot of manpower and material resources for the applied research on corresponding key technologies.However, due to lacking of commercial optical memory and the separation between burst control packet and burst data packet, the burst loss ratio of OBS is not ideal. In order to decrease the burst loss ratio of OBS networks to a practical level, a number of researchers have devoted themselves into the studies on decreasing burst loss. With the same purpose, this dissertation focuses on several key technologies for decreasing burst loss in OBS networks, which mainly include following three aspects: 1. A high-speed channel scheduler for OBS networks based on MaximumChannel Utilization with Void Filling(Max-CU-VF) scheduling algorithmThe channel scheduling is the process that each forwarding node in the OBS networks schedules the burst data packet on a data channel in advance according to the information carried by the corresponding burst control packet. Not only does an ideal scheduler need to assign a proper channel for a burst data packet rationally, but also its speed should be fast enough. Otherwise, many bursts have to be discarded due to the early exhaustion of the offset time(the burst data packet catches up the burst control packet) or the overflow of waiting queue. Recently with the explosive growth of cloud computing, massive video, online game data and so on, the bursts created in a unit time are increasing sharply. Therefore, the burst loss problem caused by slow scheduler is becoming a bottleneck that limits the development of OBS networks.This dissertation proposes a high-speed scheduler for OBS networks based on the Maximum Channel Utilization with Void Filling(Max-CU-VF) scheduling algorithm. Compared with existing scheduling algorithms, such as traditional LAUC-VF, Max-CU-VF changes the criterion of choosing an optimal data channel from the voids between burst data packets(fine granularity) to the data channel utilization(coarse granularity), which can promote the processing speed by removing the requirement on the detailed information between burst data packets(e.g. the start and end times of voids). By this way, Max-CU-VF can reach a faster speed and higher channel utilization at the same time, which results in the improvement of the burst loss ratio in OBS networks. 2. A streamline effect based routing strategy for multi-flow in OBS networksA good routing strategy is able to realize load balance and avoid congestion bottleneck, so it can reduce the burst loss ratio and enhance the throughput. However, the existing OBS routing strategies, Deflection Routing(DR), Shortest Path First(SPF) and so on, still have some shortages, such as lacking of flexibility or never considering the streamline effect. The streamline effect is the contention effect when many flows from different inputs contend for the same output in the case of being short of commercial optical storage equipments. A significant characteristic of streamline effect is that the bigger flow always has smaller burst loss ratio. Hence, for a flow, it is not proper to choose the path with the minimum end-to-end burst loss ratio. Instead, the flow should consider what influence its path selection will bring to other flows. Therefore, the whole flows in the OBS networks need to care about each other as they are routing, which will lead to the rational traffic routing for the whole burst loss of OBS networks.Considering that point, this dissertation designs a streamline effect based routing strategy with high speed and good burst loss for multi-flow. When there are multiple flows needing to be routed in the OBS networks, they could rapidly choose their optimal paths from the alternative paths in parallel according to the streamline effect in two times, which can avoid the traffic contention and load unbalance existing in DR and SPF as well as effectively reduce the burst loss ratio in OBS networks. 3. A Dual Price-based Congestion Control(DPCC) mechanismThe network congestion always leads to serious packets loss. Till now, many OBS network congestion control schemes, the congestion control schemes based on dynamic adjustment of the burst data sending rate or the feedback information that reflects the network traffic situation, usually provide several local services for current node or take reactive operation after the congestion happening. If those dynamic congestion problems can be solved in a proactive manner previously with a global perspective, the result will be much better.Network Utility Maximization(NUM) has found many applications in Internet congestion control protocols, network efficiency fairness, wireless network rate allocation, etc. Inspired by above researches, this dissertation designs a Dual Price-based Congestion Control(DPCC) scheme from the perspective of NUM. Since the quality of data transmission(e.g. end-to-end reliability, data sending rate, network delay, etc.) in OBS networks always influence the users’ satisfaction(i.e. network utility), the quality of data transmission can be promoted by the way of maximizing network utility, which results in the good control of the traffic conflict and congestion. DPCC can adjust each user’s sending rate and end-to-end reliability(the probability of data successfully arriving at the destination) according to the feedback information that reflects the network congestion situation or burst loss ratio(e.g. congestion price and reliability price) so as to maximize the network utilization and realize effective burst loss control.