[[alternative]]Enhancement and Analysis of Fairness between TCP Reno and TCP Vegas

碩士資訊工程學系[[abstract]]近年來傳輸控制協定（Transmission Control Protocol, TCP）的壅塞控制（congestion control）吸引了許多網路研究人員投身研究。雖然有許多研究指出TCP Vegas相較於TCP Reno有許多優點，但Reno目前仍為最廣為採用的傳輸控制協定。造成上述的情形主要肇因於Vegas和Reno的相容性。當網路上只有Vegas的傳輸控制協定時，其表現較Reno來得好，不論是在整體網路使用率、穩定性、公平性、產率、封包遺失率以及突發性等皆是如此。然而當有Reno存在網路上時，Vegas的效能會受到限制而表現低落，也因此目前的作業系統較少採用Vegas為預設的傳輸控制協定。 在這篇論文中我們提出一經過改良的Vegas傳輸控制協定，叫做COmpetitive DElay-based TCP (CODE TCP). CODE能夠和Reno相容並且取得其應有的網路資源並維持良好的公平性。和之前相關研究所提出的方法較為不同之處是在於CODE並不是對其某些參數作最佳化，或者是藉由路由器上的佇列管理機制來達到公平性。故CODE直接實作在傳送端上即可。根據我們的模擬及實驗結果顯示CODE在有Reno存在的網路環境下仍有較佳的公平性並維持Vegas原有的良好特性。[[abstract]]In recent years, Transmission Control Protocol (TCP) congestion control has attracted great interests from the networking researchers. Although lots of researches show the superiority of TCP Vegas over TCP Reno, Reno is still the most widely deployed variant of TCP. This predicament is primarily due to the incompatibility of Vegas with Reno while Vegas in homogeneous scenario performs better with respect to overall network utilization, stability, fairness, throughput, packet loss rate, and burstiness. Its performance is generally mediocre in environments where Reno connections exist. Hence, there exists no incentive for operating systems to adopt Vegas as default transport protocol. In this thesis, we proposed a new Vegas variant called COmpetitive DElay-based TCP (CODE TCP) which could be compatible and get the fair share of network resources with Reno. Unlike the solutions proposed in the past, CODE is neither depends on optimizing any critical parameters nor on the queue management scheme at the routers and hence could be implemented solely at the end host. Our simulations and experiments confirm that CODE has better fairness characteristic in network environments where Reno exist without discarding the good features of Vegas