二进制ABR流量控制算法的建模与分析

面向连接的ATM网络通过一系列的流量管理机制为各种应用提供纯粹的服务质量保证（QoS)，其中ABR业务流量控制的作用尤其重要，在二进制流量控制方案中，信元速率和队列长度的大幅振荡降低了链路利用率，严重地影响了算法的效率，但它的简洁性对设计高性能交换机又极具吸引力，该文基于流体流理论建立了二进制流量控制的分析模型，并用仿真试验验证了模型的正确性，通过小信号局部线性化方法得到源终端系统调节速率的线性模型，分别将标准EFCI算法和作者改进的基于概率标识的p-EFCI算法等效为带继电器特性的饱和特性的非线性环节，用控制理论中描述函数的分析方法研究了两种算法的稳定性，在理论上为p-EFCI算法在鲁棒性方面优于EFCI算法找到了根据。

Modeling and Analysis of Binary ABR Flow Control

ATM networks provide pure QoS(Quality of Service)for diversified services through a series of traffic management mechanism, and the ABR(Available Bit Rate) flow control is especially important. In the binary flow control scheme, cell rate and queue length oscillate with great magnitude to reduce link utilization, and make a negative impact on switch performance, so the standard EFCI algorithm is regard as ineffecient, however, its simplicity is attractive to high performance switch design. In this paper, the model of the binary ABR flow control is deduced based on fluid flow theory, and the correctness of this model is validated through comparing the analytical solution with the simulation results. Then, the second order system presenting the source end system how to regulate the cell rate, is obtained by local linearization about the equilibrium point, moreover, the standard EFCI algorithm and the improved p EFCI algorithm are approximated as the typical nonlinear components, and they are relay with hysteresis loop and saturated component respectively. Subsequently, we investigate the stability of the two algorithms using the describing function approach, which is very mature and effective and widely used in analysis and design of classical nonlinear control system. The conclusion is that the p EFCI scheme is more robust than the standard EFCI algorithm because the relay with hysteresis loop has negative imaginary part of the negative of the inverse of the relative describing function, which provides more chances to intersect with the Nyquist diagram. It is intersection that leads to the self oscillation occurred in queue and ACR. However, for the saturated component included in p EFCI algorithm, the imaginary part is equal to zero. Finally, we emphasize that theoretical analysis is rather important during algorithm design because the performance of heuristic algorithms depended on intuition, just like EFCI, is not always reliable, and point out the farther studying direction, i.e., optimizing the binary ABR flow control applying with the design approach in control system.