差異性服務之智慧型控制

Intelligent Control in Differentiated Services

蕭博謙(Po-Chien Hsiao)

指導教授 ： 陳朝烈

崑山科技大學/工程學院/電子工程研究所/碩士(2004年)

https://doi.org/10.6828/KSU.2004.00007

摘要

在這一篇論文之中，我們提出了一個結合模糊控制理論與差異性服務理論的方法，來提高網路之間資料傳輸所應該對應的速度，並且提高網路頻寬、資源的使用率，進而提升網路之間傳輸的服務品質。由於網路的變化日新月異，資料的傳遞已不僅侷限在有線的實體線路部份，越來越多無線傳輸的環境被使用，並且網路上許多型態的資料傳輸，如：檔案的傳輸、訊息的傳遞、即時應用軟體…等，有越來越多被大量使用的趨勢。但是因為不同的傳遞資料，它的環境需求，如：頻寬(Bandwidth)大小、封包延遲(Delay)的長短、封包延遲變化率(Jitter)大小、無線通道傳輸的等待時間…等等，也不盡相同。在這樣的需求之下，差異性服務(Differentiated Service, 簡稱DiffServ)便引起廣大的研究，並提出了許多解決方案，然而這些方案都有其限制，原因是對於瞬息萬變的網際網路而言，任何節點之間都是彼此互相影響的，況且，許多網路狀態或效能對於環境參數的變化，多是非線性甚或是不確定性的，因此使用分析的方式(Analytic Methodology)或者是機率統計的方法都無法保證能夠達到穩定且有效的控制，基於這個原因，常被用來對於不確定性系統進行控制的軟式計算(Soft Computing)就是解決方案的最佳抉擇。
所以我們使用模糊控制理論，可以針對不同型態的資料，做出區隔的動作，然後在依照各種資料的型態，利用模糊控制的方法分別給予不同等級的傳輸控制，以達到頻寬的有效利用，進而提升資料傳輸的傳輸速度或傳輸率，以提高網路之間的服務的品質。本論文使用軟式計算中模糊控制的方法來達到這樣的目標。其控制方法來自於模糊自動機理論(Fuzzy Automata Theory)，其穩定性已經被證明，為了提高控制的有效性，本論文並提出使用內?法使得控制器可以快速達到設定的目標，以降低其穩定時間(Settling Time)。
在控制時，本論文針對差異性服務同時提出全域性以及區域性的的考量，其中對於全域性的頻寬保留，我們使用模糊控制器觀察封包佇列的變化，對有線的骨幹頻寬進行配置，使得Edge Router到Core Router的頻寬得以達到設定的目標，對於區域性的頻寬配置上，採用分散式控制的方法，對於802.11中的分散式協調功能(Distributed Coordinated Function, 簡稱DCF)中的競爭視窗參數進行模糊控制，使得移動式節點(Mobile Node, MN)的頻寬得以實現在DiffServ機制下設定的頻寬。
在模擬的結果之中我們可以發現，在有線網路部份：可以有效的利用佇列之間封包微量的變化，降低封包遺失的可能性，並且可以控制頻寬維持在一定的範圍之中。在無線網路部份：採用分散式控制的方法，並調整各個資料流中競爭視窗(Contention Window)的CWmin參數來控制傳輸的流量，改變封包在傳輸時通道的使用率，進而提高無線網路之中的服務品質。

關鍵字

模糊控制 ； 無線網路 ； 分散式協調功能 ； 差異性服務 ； 服務品質

並列摘要

In this thesis, we introduce a way, which combines fuzzy control theory and differentiated service to raise the quality of information communication between networks including utilization of bandwidth and other network resources. Since network communication technology is developing incredibly quickly. The information amount is large and is exchanged among not only wireline networks but also wireless ones. Different data transmission types including file transfereing, message exchanging, and real-time multimedia applications are integrated in the information networks. However, different classes of transmissions have different requirements provided by the network environment. The requirements introduce different transmission classes constraints including bandwidth, packet delays, variations of delays (jitters), defer time duration of wireless nodes, lost rates, and so on. We attempt to classify the different transmission types so that the network environment can provide corresponding services. Under this requirement, Differentiated Service, abbreviated DiffServ is widely studied. Many scientists proposed solutions. However, These solutions have limits in network applications. The limits come from mercurial states of Internet and nodes mutual influences. Many parameters, states, and performances factors are nonlinearly and uncertainly affect each other. Therefore, many proposed analytic and statistic methods cannot make sure whether the network controls are as well stable and efficient in different applications. Due to this reason, the commonly used soft computing methodology for tackling uncertain and onolinear problems become best candidate solution.
The proposed fuzzy control algorithms reserves bandwidths and control the delay in the networks so that bandwidth and other network resources can meet the constraints of different transmission classes. Therefore, after this control the total data rates are promoted and network achieves a satisfactory service quality. The proposed fuzzy control comes from generalized fuzzy automata (GFA) theory. The stability property is proven. Moreover, to enhance the efficiency of control, this thesis also proposes a control command interpolation approach. Therefore, the reference target can be quickly tracked and the settling time is novelly decresed.
In the controller, global and local service considerations under DiffServ are proposed. In the global consideration, bandwidth reservation principles are realized with dynamic bandwidth allocations according to the variation of packet queues along backbone wirelines. The bandwidth and flows between an edge router and a core router are controled tend to constant as specified in service agreement. Therefore, the bandwidth allocations for cells are acheved. In the local consideration, we adopt distributed control mechanism. The mechanism performs fuzzy controls the contention window parameters for distributed coordinated function (DCF) of the standard IEEE 802.11. Therefore, bandwidth specifications in the service agreement of DiffServ for mobile nodes are satisfied even when the specifications are dynamic and the network background traffic is bursted.
In the simulations of wireline network, we show that dynamics of packet queues are used to monitor the network status so that adequate controls are provided to reduce packet lost rates and maintain the bandwidth variations within a small range. In the wireless network simulations, we show that distributed controls of data flows over the CWmin parameter of the contention window effectively and efficiently enhance the channel utilization when transmitting packets. Integrated the global wireline and local wireless fuzzy control algorithms, the network provides high service quality.

並列關鍵字

Diffserv ； IEEE 802.11 ； Fuzzy Control ； DCF ； QoS

參考文獻

[4] Runtong Zhang, Jian Ma, "Fuzzy QoS Management in Diff-Serv Networks" , IEEE, 2000, p.3752-3757

[7] J. Antonio Garcfa-Macfas, Franck Rousseau, Gilles Berger-Sabbatel, Leyla Toumi, Andrzej Duda, “Quality of Service and Mobility for the Wireless Internet” , ACM Press Conf, 2001

[10] Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications, IEEE Standard 802.11, June 1999.

[18] Chao-Lieh Chen, "Programmable fuzzy logic device for sequential fuzzy logic synthesis", The 10th IEEE International Conference on Fuzzy Systems, Vol. 1, Dec. 2001, pp.107 - 110

[20] Sheldon S. L. Chang and L. A. Zadeh, "On fuzzy mapping and control", IEEE Tran. Syst., Man, and Cybernet., Vol. 2, No. 1, Jan. 1972, pp. 30-34.