Non-parametric performance analysis and its applications to the design and control of the Internet

非参数性能分析及其在互联网设计与控制中的应用

基本信息

批准号：
15500180
负责人：
SHINODA Shigeo
金额：
$ 1.73万
依托单位：
Chiba University
依托单位国家：
日本
项目类别：
Grant-in-Aid for Scientific Research (C)
财政年份：
2003
资助国家：
日本
起止时间：
2003 至 2006
项目状态：
已结题

项目摘要

(1)Non-parametric performance evaluation・We established the formula that gives the upper bound of work-loss probability based only on the marginal distributions of work and interarrival processes.・We derived bounds of virtual-waiting-time distribution only assuming that sources are stationary, statistically independent from each other, and have deterministic subadditive envelopes without using a specific traffic pattern.・We proposed a semi-parametric estimator of effective bandwidth for superposition of sources in IP networks. Our proposal works based on a small set of measurements of the workload in the buffer of a router.(2)VoIP admission controlWe proposed an admission control for the IP telephony, service in Diffserv-based IP networks. The proposed control is fully distributed ; each edge router autonomously admits or rejects connection-set-up requests based on the stored information of resource availability. In the proposed control, resource availability is expressed in terms of the maximum number of multiplexable IP telephony sources in advance.(3)Traffic matrix estimationWe studied the problem of estimating a source-destination (SD) traffic matrix of an IP-based VPN from measurements of aggregated traffic along all access link, and derived two different estimators by formulating the SD-traffic-matrix estimation as a deterministic or stochastic optimization problem.(4)Bottleneck link bandwidth estimationWe proposed a technique for measuring the bottleneck bandwidth of a network path between end hosts. Our proposal is based on the packet-pair technique and uses a simple statistical procedure for filtering out the cross traffic effect.(5)Efficiency analysis of IP traceback techniquesWe theoretically evaluate the efficiency of typical PPM schemes based on a simple analytical model. In particular, for each scheme, we derive an explicit representation of the average number of packets that the victim should receive until the entire attack paths can be reconstructed.

(1)非参数绩效评估・我们建立了仅基于工作和到达间隔过程的边际分布给出工作损失概率上限的公式。・我们仅假设来源推导了虚拟等待时间分布的界限是静态的，理论上彼此独立，并且具有确定性的次加性包络，无需使用特定的流量模式。・我们提出了一种用于 IP 网络中源叠加的有效带宽的半参数估计器。我们的建议基于一小组测量结果。的(2)VoIP准入控制我们提出了一种针对基于Diffserv的IP网络中的IP电话服务的准入控制。所提出的控制是完全分布式的，每个边缘路由器自主地准入或拒绝连接集。在所提出的控制中，资源可用性预先用可复用IP电话源的最大数量来表示。(3)流量矩阵估计我们研究了估计源-目的地的问题。基于 IP 的 VPN 的（SD）流量矩阵，通过测量所有接入链路上的聚合流量，并通过将 SD 流量矩阵估计公式化为确定性或随机优化问题，得出两个不同的估计器。（4）瓶颈链路带宽估计提出了一种测量终端主机之间网络路径瓶颈带宽的技术。我们的建议基于数据包对技术，并使用简单的统计过程来过滤交叉流量影响。(5)IP 效率分析。回溯技术我们基于简单的分析模型从理论上评估了典型 PPM 方案的效率，特别是，对于每个方案，我们得出了受害者在重建整个攻击路径之前应接收的平均数据包数量的显式表示。