智能与分布计算实验室
  一种面向QoS的拥塞管理机制及仿真研究
姓名 范小丹
论文答辩日期 2005.04.11
论文提交日期 2005.05.15
论文级别 硕士
中文题名 一种面向QoS的拥塞管理机制及仿真研究
英文题名 A Method of Congestion Management for QoS and the Research in Simulating
导师1 卢正鼎
导师2
中文关键词 服务质量;拥塞;集中拥塞管理;仿真
英文关键词 Quality of Service;Congestion;Aggregating congestion management;Simulator
中文文摘 QoS(quality of service,服务质量)是用带宽、分组延迟和分组丢失率等参数描述的关于分组传输的质量。传统的Internet只提供非实时的、单一的尽力而为(best-effort)型服务,而当今高速网络中的多媒体应用则需要提供端到端的QoS控制和保证。IETF(Internet Engineering Task Force, Internet工程任务组)等标准化组织提出了各种不同的Internet QoS业务体系结构:综合服务(integrated services)、区分服务(differentiated services)等,它们通过服务区分、制定优先级、资源预留等机制提供QoS控制和保证。其技术的核心都需要在恰当的层次和粒度上对流量进行必要的管理,其中包括接纳控制、流量成形、队列管理、调度和拥塞控制等诸多方面,但最基本和最核心的应该依旧是拥塞控制,因为很难想象一个时常有可能出现严重拥塞且无法及时加以恢复的网络能够实现良好的QoS保证。实施拥塞控制应该是其它QoS机制正常工作的必要前提。 目前,IETF新近提出的综合服务和区分服务业务体系结构仍处于研究阶段。作者从服务区分的角度出发,在对Internet网络拥塞控制领域已有研究成果进行系统、全面研究总结的基础上,提出了对源端的多个TCP数据流进行集中拥塞管理的机制。在网络资源有限的情况下,该机制可以根据各流(flow)的性能需求调整其所占用带宽,从而满足对QoS要求较高的流(如视频、声音等)的需要,提供QoS控制和保证。通过对该机制进行网络仿真和性能分析,可以验证该机制具备可扩充性、公平性和最大利用率等优点。
英文文摘 QoS(quality of service) is used to descript the quality of the packet traffic with bandwidth, packet delay and probability of packet losing etc. The traditionary Internet provides only the best-effort service which is non-realtime and simple. However,today’s multimedia applications running in high-speed network need the end-to-end control and guarantee of QoS. IETF(Internet Engineering Task Force) has constituted a serial of standards, such as integrated services and differentiated services, which provide the QoS-control and QoS-guarantee through differential-service, premium-service and resource-reserving etc.The key technique of these is to manage the traffic in felicitous level and granularity, including admission control,flow shaping,queue management and congestion management etc.And the congestion management is the most elementary, it’s unimaginable that the network can get control and guarantee of QoS when the serious congestion frequently happen and can’t come back in time.Congestion control should be the precondition of those diverse mechanisms of QoS. Today, integrated services and differentiated services are still in research status. Author summarizes the research of current congestion-control management and presents the method aggregating congestion management of multi-TCP flows. This method can adjust the bandwidth of each flow according its performance requirement and thus meet the higher bandwidth need of those flows (video, audio etc). At last, through simulating this method and making performance analysis, the fairness, utilization and scalability are all confirmed.