A Statistical Learning Framework for Investigating Scalability and Performance of Measurement-based Network Monitoring

用于研究基于测量的网络监控的可扩展性和性能的统计学习框架

• 批准号: 0300605
• 负责人: Chuanyi Ji
• 金额: $ 27万
• 依托单位: Georgia Tech Research Corporation
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2003
• 资助国家: 美国
• 起止时间: 2003-05-01 至 2006-04-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=0300605&HistoricalAwards=false
• 关键词: Statistical; Learning; Framework; Investigating; Scalability

Measurement-based network monitoring and surveillance is an important area for providing reliability and security to IP network infrastructure. A challenge is how to develop scalable monitoring approaches.This Project will investigate the scalability and performance of measurement-based network monitoring. At the high level scalability relates to the growth rate of monitoring resource with respect to the size of a network. The performance characterizes the accuracy of network monitoring. The scalability and performance together impact the feasibility of such approaches to large networks.The PI will use a statistical learning framework to investigate the issues of scalability and performance. She will formally define and analyze scalability, and provide understanding on when and why scalability may be achieved. She will then investigate the trade-off between scalability and performance, and provide guidelines for developing good measurement-based approaches for large networks.Although there have been many efforts to develop measurement-based methods for network monitoring, the scalability and performance issues have been investigated little. Meanwhile, though statistical learning theory has been applied to many areas, it has not been widely used in measurement-based network monitoring. The proposed investigation hopes to benefit measurement-based network monitoring, and to enrich the statistical learning theory also. This research will be done in Collaboration with Anwar Elwalid at Bell-Labs Lucent Technologies.

基于测量的网络监控是为IP网络基础设施提供可靠性和安全性的重要领域。一个挑战是如何开发可扩展的监控方法。该项目将调查基于测量的网络监控的可扩展性和性能。从高层次上讲，可伸缩性与监控资源相对于网络大小的增长率有关。该性能表征了网络监控的准确性。可伸缩性和性能共同影响这种方法在大型网络中的可行性。PI将使用统计学习框架来研究可伸缩性和性能问题。她将正式定义和分析可伸缩性，并就何时以及为什么可以实现可伸缩性提供理解。然后，她将研究可扩展性和性能之间的权衡，并为为大型网络开发良好的基于测量的方法提供指导。尽管已有许多努力为网络监控开发基于测量的方法，但对可扩展性和性能问题的研究很少。同时，统计学习理论虽然已经在很多领域得到了应用，但在基于测量的网络监控中还没有得到广泛的应用。本文的研究希望对基于测量的网络监控有所裨益，同时也希望丰富统计学习理论。这项研究将与贝尔实验室朗讯技术公司的安瓦尔·埃尔瓦利德合作完成。