在多速率及多網路基地台WiFi系統之網路選擇最佳化

Abstract

因為現在WiFi需求增加，對於使用無線網路的效能要求也提高，為了提供更好的QoS (Quality of Service)，我們需要在多速率且多AP的環境下提供最佳效能。此處討論及分析效能時以throughput跟delay為主，並參考目前的無線網路環境，使用IEEE 802.11標準。考慮在多AP的環境下，當一個node可能同時可連線到多個AP，這時選擇哪一個AP連線就要考慮對整體throughput跟delay的影響，就是網路選擇(network selection)問題。在這個問題下，我們想要決定一個架構使得系統整體效能最大，換句話說就是系統throughput最大及每個封包的delay最小。將此網路選擇問題寫成數學式，並用最佳化的方式解決。為了解最佳化問題，我們根據模擬觀察及分析，寫出可代表不同情況下系統效能的utility function，並使用utility代表實際的網路效能。我們提出了疊代法(iterative method)做為解多速率環境下網路選擇問題的方法。經由設計適合的初始狀態及改變現在連線狀態的政策，我們就可在有限時間內得到或逼近最佳的網路架構使得整體效能最大。 決定utility function時，首先考慮系統裡只有一個AP的情形，加上由於無線網路的環境，node與AP的距離不同會造成不同的傳輸速率。考慮在多速率的環境下，我們可以將不同傳輸速率的node分時傳輸，避免了高速率跟低速率node一起傳輸所下降的throughput，這個方法稱為分群 (grouping)。上述的在單一AP下會影響throughput的各種參數和多速率的問題下使用分群的情形一併考慮進來就可得出utility function。 決定使用疊代法解最佳化問題後，我們比較在不同網路架構下疊代法跟其他方法所得出的結果差異，無論是改變系統的使用者數目還是改變封包產生的速率，疊代法都是最接近最佳解的方法。並且在最好的情況下疊代法比其他方法高出約15%的效能。

Due to increasing requirements of wireless network, users’ requirements of performance in wireless network also increase. To provide better Qos (Quality of Service), we have to provide good performance in multi-rate and multiple APs scenario. Here we discuss throughput and delay for performance, and use IEEE 802.11 standard for wireless network in the following discussion and analysis. Consider that there are many APs in the system, each client can choose which station to connect now, and different connections affect throughput and delay of system. This is a network selection problem of multi-rate system. In this problem, we want to find a configuration such that the system performance is maximal, which means maximal throughput and minimal delay. We can solve the network selection problem in optimization, and use utility function to represent real performance in this problem. With utility function, we can find optimal configuration which causes maximal performance by using iterative method of utility function optimization in multi-rate network selection problem. After design of iterative algorithm, it is ensure the method can get or close to optimal configuration, and has the best performance than other methods. To decide utility function, consider there are one AP in the system, find the variables which affect performance. In addition, different distances from client to AP cause different transmission rate. In multi-rate environment, we can divide the stations by transmission rate into different groups, so it reduces the reduction of throughput if high rate stations and low rate stations transmit together. The method is called grouping. The analysis of the variables effect on multi-rate system are used to utility function. After design of iterative algorithm, we compare this method with other methods in different configurations. Iterative method is the best method which is the closest to optimal performance even if changing client number in the system or changing arrival rate. Also, the performance is higher than other methods by 15% in the best case.