用于Ad Hoc网络的自适应多速率多播拥塞控制策略

多播提高了链路的传输效率,但易于造成网络拥塞.因此,在网络中实施多播拥塞控制至关重要.然而,由于AdHoc网络的两个本质特点,为Internet设计的多播拥塞控制不适合AdHoc网络:(1)无线多跳连接引起了信息流之间在时间域和空间域的竞争;(2)节点频繁移动导致了网络状态不断变化.首先提出了链路干扰集的概念来描述信息流竞争的特点,将网络状态不变的小时间段内的多速率多播拥塞控制问题表达成一个非线性优化问题,联合运用罚函数法和次梯度法获得此问题的优化解,相应地提出了一种有效的分布式迭代算法.在此算法基础上,针对网络状态的时变性,设计了一种基于状态检测和滚动优化的自适应多速率多播拥塞控制策略——AC2M2.仿真结果表明,分布式算法能够快速收敛到最优解;AC2M2(adaptive congestion control strategy for multirate multicast sessions)策略对网络状态的变化具有较好的自适应能力,所获得的网络性能比TCP-Reno要优越得多.

Adaptive Congestion Control Strategy for Multirate Multicast Sessions in Ad Hoc Networks

It is vital to implement multicast congestion control in networks as multicast improves the link's transmission efficiency, but it is apt to cause network congestion. However, multicast congestion control designed for Internet isn't fit for Ad Hoc networks, due to the following features: (1) multi-hop wireless-based transmission resulting in the contention relationship between flows in both the time domain and the spatial domain; (2) and frequent node mobility leading to the time-varying network situations. In this paper, the notion of the link's interference set is introduced to describe the characteristics of the contention relationship between flows, and a nonlinear optimization problem is presented to formulate the multirate multicast congestion control problem for small time interval where network situations can be regarded as time-invarying. The penalty function method and the subgradient method are jointly applied to obtain the optimal solution, and thus the distributed iterative algorithm is proposed. On the basis of this distributed algorithm, an adaptive congestion control strategy for multirate multicast sessions (AC²M²) is proposed to cope with the time-varying network situations, by means of network status detection and receding optimization. Simulation results show that the proposed distributed algorithm can quickly converge to globally optimal solutions; and, compared to the AIMD algorithm in TCP-Reno, AC²M² is more adaptive to time-varying network situations, and achieves better network performance.