在以測量值為基準的允入控制法中所設計的目標使用量調整法

Abstract

以測量值為基準的允入控制法(MBAC)主要是要讓整合性服務網路能達到更高的使用量，能夠既滿足服務品質，又能夠使網路達到最高的使用量，正是MBAC的設計目標。過去有人研究得到了一個結論，就是到目前為止所發展的MBAC都具有相同的效能，只要我們能研究出調整MBAC參數的方法。因著這個結論，我們特別研究了兩個MBAC參數的性質，測量窗與目標使用量。在我們的研究中發現，如果我們想要針對封包的落失機率作限制，測量窗是很不適用的調整參數。因此我們進一步地發展了調整目標使用量來滿足封包落失機率的限制方法。這方法基本上要依靠線上測量網路流量的特性，但這個測量結果又通常不是準確的，為了解決這個問題，我們更進一步設計了一套輔助的門閂演算法。從模擬實驗中可以看出，我們的設計不但有效，而且在整體效能上也表現得不錯。

Measurement-Based Admission Control (MBAC) is provided to give the higher link utilization for Integrated Services Networks. Achieving both QoS constraint and optimizing the link utilization are the subjects to design the MBAC algorithm. It has been claimed that the MBAC algorithms developed so far give the same performance if we can properly tuning their parameters. From this issue, we discussed the properties of two parameters in the Measurement-Based Admission Control Algorithm, Measurement Window and Utilization Target. We found that the Measurement Window is not a good tuning knob for the loss probability constraint. Thus we further derived the loss probability constraint methodology by resizing the Utilization Target. The methodology needs the evaluations of traffic behaviors. But the on-line statistics from the measurements are usually inaccurate. To overcome this problem, we introduced the auxiliary policies called the Latch Algorithm. The simulation results have shown that the design of the admission control in our work is robust and performs well.