基于测量的接纳控制研究

与传统的接纳控制算法相比，基于测量的纳控制有诸多优点，首先它无需知识应用的流量模型，其次它能动态适应网络的负载变化，提高网络资源的利用率。文中分析了基于测量的接纳控制的基本思想，并在此基础上提出和实现了一种自适应的接纳控制算法（Adaptive Measurement-Based Admission Control,AMBAC).作者通过实验对该算法进行了验证，发现在系统资源利用率（或接纳能力）接近的情况下，与传统的（固定时间窗口的），MBAC相比，AMBAC能达到更低的平均分组丢失率。

An Adaptive Measurement-Based Admission Control Algorithm

In contrast to traditional admission control mechanisms, the mostattractive feature of Measurement-Based Admission Control (MBAC) is that it does not require an apriori traffic model, because it is very difficult or even impossible for the user or application to come up with a tight traffic model before establishing a flow. Other advantages of MBAC include that an overly conservative specification does not result in an over-allocation of resources for the entire duration of the session, and it can adapt to the changing traffic load dynamically, so it improves the network utilization while offering quality of service to users.This paper first studies how MBAC works. Existing MBACs adopt the fixed-length Time-window Measurement Mechanism to estimate the network traffic load LCL. There are two important parameters (measurement window T and sampling window S) that impact the estimation of LCL. Because S has smaller impact than T, in this paper, we only consider T. In MBAC, small T means more adaptability and higher resource utilization, but larger T results in greater stability and lower resource utilization. Hence, to select an appropriate T is very important for MBAC. To solve this problem, we propose an Adaptive Measurement-Based Admission Control (AMBAC) algorithm. In AMBAC, we set two thresholds: LTmax and LTmin. When the measured traffic load LCL is larger than LTmax, our algorithm enlarges T automatically, which makes AMBAC more conservative and hence decreases the network's admission ability. When LCL is smaller than LTmin, our algorithm shrinks T, which improves the network's admission ability. When LCL is between LTmax and LTmin, our algorithm does not alter T. By altering T, AMBAC makes the network adapt to the changing traffic load dynamically, so the network utilization is improved. To evaluate AMBAC we implemented our algorithm on FreeBSD. We test it under different traffic scenarios and compare it with the traditional MBAC. Our simulation results show AMBAC can get lower packet-loss while achieving a high level of utilization.