利用頻寬適應性技術之雲端資料中心網路壅塞控制機制

Abstract

在過去十年來，雲端資料中心迅速崛起並成為運算上重要的一塊，並且成為一個重要的研究議題。然而，最近的研究者著重於解決網路設計，而網路通訊管理方面的議題仍值得深入。如何為網路流選擇正確的路徑並不是個可以簡單回答的問題，尤其當需要保證分佈平均以及負載平衡，還要能快速的識別以及回應故障時。而更加困難的是，當這個問題還面臨著資料中心裡不確定的流量負載。 這份碩士論文意旨在研究雲端資料中心的壅塞問題。這份研究的目標在於提出一個壅塞控制機制，並評估目前的多路徑繞徑技術的效能。本文所提出的方法稱為「頻寬適應性技術之網路壅塞控制」，該方法是基於雲端資料中心的網路流量的特性與行為為討論基礎。我們透過兩個模型來檢驗所提出的方法的有效性，分別是m-port n-tree fat-tree 拓撲和英業達資料中心拓撲。模擬的情境是透過NS-2 模擬器來實作並完成測試。其效能結果顯示，透過「頻寬適應性技術」優於雲端資料中心使用傳統TCP在吞吐量、封包錯誤率與延遲等問題的表現。

During the last decade, datacenters have risen to dominate the computing landscape and become an important research topic. Among four main components of a datacenter networking architecture: physical topology, routing over topology, selection between multiple paths supplied by routing, congestion control technique on the high speed, low latency network, the last three objects in the list haven't investigated worthy. In this thesis, we propose a congestion control mechanism in cloud datacenter by bandwidth adaptation and evaluates the performance of multipath routing in its present in cloud datacenter. Both fat tree and experimental topologies are considered in the testing of performance. Simulations are conducted to compare the e?effectiveness of the proposed mechanism. The performance results demonstrate that the proposed congestion control in cloud datacenter by bandwidth adaptation can outperform the throughput, packet error rate and delay issues compare with traditional Transmission Congestion Protocol (TCP) in cloud datacenter environments.