非同步傳輸模式網路中服務品質為基之擁塞控制

Abstract

本論文中所討論的，為非同步傳輸模式(ATM) 網路中之擁塞控制方法。非同步傳輸 模式目前已被當作寬頻服務數位網路的一種交換與多工的傳輸技術。由於非同步傳 輸模式網路中，不同型態的資料，其傳輸的特徵及傳輸品質的要求有很大的差異， 並且資料之傳輸速率相當高；所以我們在ATM 上作擁塞控制時相當不易。本論文所 提出者乃屬一種服務品質驅動的控制技術，包含了許可控制(Admission Control) 及流量強制技術(Traffic Enforcement Mechanism) 。此處所指的服務品質主要是 以最大延遲變化量及最大資料遺失率來衡量。 在本論文所提出的許可控制技術中，當要求一個新的連線時，我們根據此連線所能 忍受的最大延遲變化量，在網路上的交換節點中，給與適當的邏輯緩衝空間，並求 出對應各連線的資料遺失率。如果各連線能接受其對應的資料遺失率，則我們將接 受這個新的連線要求。一旦新的連線要求被接受了，我們必須控制這個連線輸入網 路的資料流量，以防止過多的資料量造成網路擁塞而癱瘓。如果一個連線要求是在 網路負載很輕時被接受的，我們就放寬限制，讓資料以其最高速率進入網路；當網 路負載變重時，我們就恢復對流量的限制，讓資料以其平均速率進入網路。這樣一 來，我們就可提高網路的使用率，並同時維持網路所承諾的服務品質。 In this thesis, a QOS driven congestion control method is proposed. As we already know, Asynchronous Transfer Mode (ATM) is expected to be adopted by the CCITT as the switching and multiplexing method for BISDN. Congestion control in ATM networks is difficult because of the high link speed, diverse service requirements, and various characteristics of the traffic ATM is expected to support. In our method, the QOS is mainly regarding the maximum delay variance and maximum cell loss rate. The proposed congestion control method includes admission control and policing techniques. In the proposed admission control method, each call with different QOS requirements will obtain different logical buffer size in the switch nodes, and therefore, will experience different cell loss rate. A new coming call request will be admitted by a switch node if each existed connection plus the new coming one can tolerate the cell loss rate in the output buffer of the switch node. Once a call request is admitted, we must monitor the traffic source to avoid its output overwhelming the network (policing). In the proposed policing method, we use a "sleeping leaky bucket" to control the traffic flow, where cells will enter the network at peak rate if the traffic load along the designated route is light, and at mean rate if the traffic load becomes heavy. The simulation results show that our proposed congestion control method has an expected performance.