项目名称: 应用服务保障完成时限的网络传输机制研究
项目编号: No.61462007
项目类型: 地区科学基金项目
立项/批准年度: 2015
项目学科: 自动化技术、计算机技术
项目作者: 叶进
作者单位: 广西大学
项目金额: 45万元
中文摘要: 课题旨在进行保障应用服务完成时限的网络传输机制设计,围绕保数据流调度、传输控制协议设计、并发拥塞控制等不同层面的问题进行系统研究,包括针对应用服务的典型应用特征进行任务时延建模和瓶颈分析,以获得包括优先级描述的QoS指标;通过在中间结点实施不同类型优先级的流调度,分析推理所有流在进出口结构的平均时延最小化调度方案;通过端到端传输控制协议设计,使得低优先级应用容忍高延迟和丢包,以及相应的自适应拥塞控制算法,从而使获得服务的用户数最大化;针对最后期限感知的TCP协议进行优先级同步流的并发拥塞控制,以获得大量不同的优先级数据流共存时网络吞吐量最大化。通过以上四个方面的研究形成一套传输控制的整体方案,实现应用服务保障任务完成期限的最优化设计的目标。研究工作可望突破数据中心网络服务质量控制的一些基础性理论问题,大幅度改进和优化应用服务器的传输性能,为工程实践奠定坚实的理论和技术基础。
中文关键词: 服务质量;网络资源分配;数据中心网络;拥塞控制;网络优化
英文摘要: To fulfill the delay requirements of the application services in datacenter network, the project is to porpose a set of network transmission mechanisms, such as flow scheduling, transmission control protocol and the concurrent congestion control. Firstly, we plan to construct the delay distribution model for the typical applications and analyze the transmmision bottleneck to meet the QoS priority requirements. Secondly, to minimizing the average flow complete time in the I/O structure, different flow scheduling is designed on the relay nodes based on the application priority. Thirdly, to maximize the number of deadline aware users, the enhanced end-to-end transmission control protocol is to be proposed. As the result, the applications with low priority are able to tolerate high latency and packet loss. Fourthly, to achieve the higher throughput when the priority synchronization problem occurs in the different priority flows, a special congestion control mechanism is to be designed. Based on the above schemes, the integrated solution could achieve the target that more applications can finish their transfer before their deadlines. Our researches are expected to make a breakthrough in basic theory problems in QoS control of datacenter networks. Research results could largely improve the the transmission performance for delay sensitive applications, and lay a firm foundation in theory and technology for network engineering practice.
英文关键词: Qaulity of Service;Resource Allocation;Datacenter Network;Congestion Control;Network Optimization