Research On The Key Techniques Of Congestion Control For DTN

For the realization of interconnect and interoperate among heterogeneity Opportunistic Networks, as well as the reliable transmit, delay and disruption tolerant network(DTN) emerges as a new network architecture. Contrast with the packet, end-to-end communication, address addressing, full connection, DTNs make use of the bundle, hop-by-hop communication, name addressing, local connection instead. Along with the breakthrough in best-effort service, as a overlay network, DTNs can also be easy to compatible with other networks , such as TCP/IP network.DTN is being the highlight in academic and industrial circles day by day.Congestion control is a classical network question, plays a vital role in enhancing the availability and reliability of networks, as well as in increasing the through-out. Meanwhile, DTNs utilize the custody transfer, which requests the reliability of hop-by-hop transmission. Considering that, the traditional congestion control's constraint conditions are changed, makes the congestion control to be a new research question. In this article, we make a depth investigation on the key technology about DTN networks congestion control, based on the ideas of hop-by-hop congestion control. In specific, we research the questions about DTNs nodes'hierarchical clustering, the way of multi-attribute criterion in choosing the next hop, the local congestion control mechanism and the optimal management in nodes'buffer, the coordination between multi-path routing and fair-share scheduling. The chief contributions and innovation points of this article are as follows.(1) Comprehensive using of DTNs local shortest path routing algorithm and limited flooding algorithm, it establishes a clusters-based hierarchical cluster protocol ACHR, basing on distributed 2-hop clustering method and articulation point.The basic idea of ACHR is it adopts local visibility routing strategy, using the 2-hop neighborhood method to construct DTNs local domains whose visibility is 3. In these domains, it realizes management and scheduling in a certain extent, which reduces the amount of data inter-communication among domains. First of all, the identification of the articulation points within the domain is the premise of eliminating imbalance. According to the relationships between nodes in networks, distribution characteristics of the neighbors , the node's location and other factors, the article show us a 2-hop articulation points analysis model. Secondly, ACHR splits the nodes'routing into two parts: inner-domain and inter-domain. If the destination is in the domain, inner routing strategy is used, otherwise, it uses the inter ones. Tow nodes in one domain must be connected by direct or indirect path. Inner routing strategy uses the synchronous optimize-shortest-path method, the non-articulation points are prior to be chosen as the translation points, to achieve the best efficiency of all nodes, at last enhance the system's load balancing. In comparison, the exchange of massages between nodes among domains can be done depend on nothing but the mobile nodes'massage-carrying. Routing strategy uses the asynchronous restricted flooding routing, it considers the articulation points as translation points firstly. It passes the massages to node who has the maximum probability to encounter with destination, in order to enhance deliver probability. ACHR has the advantages in simplicity, availability and well- expansibility. The result shows us that, in the case of strong community, by transferring the articulation points'data traffic, it can achieve the aim that proportionate resource consuming in difference domains and keep steady delivery rate and delay time.(2)Combining with layer-crossing optimization theory, it presents a comprehensive criteria for the next hop, which is suitable for DTNs environment. Base on it, a Multi-Attribute Congestion Aware Routing(MACAR) has been designed.On the choice of next hop, DTN-related routing and congestion mechanism have provided us lots of incommensurable and even conflictive evaluation criteria. But due to the failures of algorithms in dealing with some evaluating questions such as the conflicts in multiple criteria, its performance could be further optimized. MACAR considers the various attributes'influences on networks performance through cross-layer mechanism. The predefined criteria have been summed up as coming under three top attributions: node itself static attribute, receive intention attribute and receive-and-forward attribute. It build the IDC AHM, to make the assessment results be comprehensive, objective and effective. On the basis of that, an uncertainty multi-attribute and congestion-aware autonomous decision-making algorithm MACA has been proposed. Using of decision theory, MACA firstly blends the top attributes'priority preference relations which have confidence level, and then determines the next hop's superiority (inferiority) of potential partial order by partial order distance formula, at last may be get preferential decision on choice of next hop. The algorithm is expressing simply and has low computation complexity. The results shows that, because of the using of multi-attribute analysis method, the rationality of choosing next hop has been enhanced. MACAR may access a higher data deliver probability with lower time delay, and may achieve the load-balancing routing performance to some extent, effectively alleviate the congestion phenomenon.(3)Starting with the hop-by-hop congestion control options, it proposes the node-priority-oriented and autonomous-relief congestion algorithm ARBR(Adaptive Role-Based Routing). It achieves the load balancing of nodes by dynamic node priority allocation. Back-pressure is a representation of hop-by-hop congestion feedback mechanism. ARBR has defined the measurement of nodes'intensity of distributed load by buffer and link factors, using the income-venture analysis price mechanism to adapt to the dynamic variation of available resource. Node can adjust the priority parameters adaptively. In the case of heavy load, it only agrees to accept the data relay request from specific nodes, which makes the congestion information be fed back as soon as possible, and the node can select the spare next-hop adaptively in time, so that it can make full use of the random communication opportunities, enhance the network's through-out rate. The algorithm can accurately quantify the congestion, simulation results show that, the algorithm can bring us with shortest time of all data delivered and maximum number of data delivered in any time.(4)Propose a load-balancing multi-path flow transport mechanism ALAP-LB ( Adaptive Load-Aware Routing-Load Balance).The basic idea of ALAP-LB is to use the local connectivity of networks, according to the nodes'resource occupancy condition, it conducts multi-path scheduling in network traffic. The key points of agreement includes two parts: load sensor and load balancing. The implementation could be divided several sub-operations, those are message selection, routing requests, multi-path generation, path selection, routing adjustment and so on. According to the self-adaptive path adjustment, the agreement can effectively change the data packets from heavy-load-path to light-load-path in the case that additional cost is slightly, which ensuring the a balanced distribution of network traffic flow and balance transfers. Simulation results shows that, in some aspects such as packet loss rate and end-to-end delay, ALAP-LB has a good performance, it can adapt to the dynamic environment of Ad-Hoc Networks.