Modeling Traffic Dynamics for Internet Traffic

研究目标是开发互联网的流量和性能模型。 互联网应用呈现出增长趋势、短期和长期流量可变性，并对网络提出不同的服务质量要求。这种业务环境要求网络迁移到动态路由和资源分配范例，以有效地共享网络资源。 基线流量模型和技术， 验证和自适应地更新这些模型必须在网络中适当地支持动态网络功能。 这项工作提出了测量驱动的时间序列建模方法，用于捕获流量特性和其他季节性和外源性因素，影响互联网上的流量变化。该模型将被开发，以捕捉短时间尺度上的到达过程可以是固定的，以及较长的时间尺度的非平稳功能。实时流量监控，测量和建模算法，可以自适应地更新和预测流量描述符将被设计用于在网络节点上实现。 初步研究计划旨在为上述建议的长期研究工作确定可行的交通测量策略。流量测量将在互联网上的一小组接入路由器上进行。 将概述基线流量抽样和定性方法。 分组语音，可变比特率视频和传统的互联网流量将被认为是在流的水平。增加粒度的效果 在交通测量和相对增加的信息内容有用的资源分配将审查。 这项初步工作将确定相关的时间尺度流量监测和测量和更好地了解类型的建模工具，必须在互联网节点成功地实现自适应控制的服务和路由规则。 本研究计划的结果将是PI的长期研究兴趣，在设计一个自适应的双模式连接和流级流量charac-terization方案，用于接纳控制，端到端的资源预留和自适应路由决策在宽带分组网络的整体。 请参阅http://morse.uml.edu/~kchandra/traffic_rsrch.html， 正在进行的研究活动。

The research objective is to develop traffic and performance models for the Internet. Internet applications exhibit growth trends, short and long term traffic variability and pose dif- ferent quality of service demands on the network. This traffic environment requires that networks migrate to dynamic routing and resource allocation paradigms for efficient sharing of network resources. Baseline traffic models and techniques to validate and adaptively update these models must be in place in the net- work to support dynamic network functions. This work proposes measurement driven time-series modeling approaches for capturing traffic characteristics and other seasonal and exogenous factors that influence traffic variability on the Internet. The models will be developed to capture both the short time-scales over which the arrival processes can be stationary as well the longer time-scale non-stationary features. Real-time traffic monitoring, measurement and modeling algorithms that can adaptively update and predict traffic descriptors will be designed for implementa- tion at network nodes. The preliminary research plan is aimed at determining feasible traffic measurement strategies for the longer term research work proposed above. Traffic measurements will be taken at a small set of access routers on the Internet. A baseline traffic sampling and characterization methodology will be outlined. Packet voice, variable bit rate video and traditional Internet traffic will be considered at the flow level. The effect of increasing granularity in traffic measurements and relative increase in information content useful for resource allocation will be examined. This preliminary work will determine the relevant time scales for traffic monitoring and measurements and a better understanding of the types of modeling tools that must be in place at Internet nodes to successfully implement adaptively controlled service a nd routing disciplines. The results of this research plan will be integral for the PI's long term research interests in designing an adaptive dual mode connection and flow level traffic charac- terization scheme to be used in admission control, end-to-end resource reservation and adaptive routing decisions in broadband packet networks. Refer to http://morse.uml.edu/~kchandra/traffic_rsrch.html for ongoing research activities.