CAREER: Reassessing the Foundations of Internet Transport

在过去的十年里，互联网在几个方面发展迅速。然而，其运行中使用的核心传输机制并没有太大变化。很难相信十年前网络设计所基于的假设即使在今天仍然有效。不幸的是，由于过去收集足够的统计数据的不可行性，与当今互联网中传输协议性能相关的几个基本问题的答案是未知的：“今天的互联网中是否存在拥塞？”过去的传输机制是否适用于当前的应用程序？如果不能，如何重新设计？”本项目通过：(1)广泛的测量和分析来解决这些问题，以建立对网络传输性能的基本理解，并验证遗留假设；(2)调查无效假设对现有运输机制的设计、分析和评估的影响；(3)基于研究结果设计新的机制和分析技术。一旦完成，该项目将有助于科学地重新评估过去在互联网传输协议设计中使用的原则和假设，并将使几个社区受益。首先，测量和分析方面的经验对于参与大型数据集信息挖掘的联邦、商业和学术机构来说是无价的。其次，在计算机科学教育中使用测量工具将有助于培养研究生和本科生的重要技能，并有助于开发新的教学技术。第三，随着技术进步而扩展的传输协议的设计将影响互联网在科学研究和分布式分析中的使用。

The Internet has grown rapidly along several dimensions over the past decade. Yet, the core transport mechanisms used in its operation have not changed much. It is difficult to believe that the assumptions on which network designs were based a decade ago are valid even today. Unfortunately, due to the in-feasibility of collecting sufficient statistics in the past, answers to several fundamental questions related to the transport protocols performance in today's Internet are unknown: "Is there any congestion in today's Internet? Are transport mechanisms of the past working well for current applications? If not, how can they be redesigned?" This project addresses these questions through: (1) extensive measurements and analysis to develop a fundamental understanding of transport performance of networks and validate legacy assumptions; (2) investigation of the impact of invalid assumptions on the design, analysis, and evaluation of existing transport mechanisms; and (3) design of new mechanisms and analysis techniques based on the findings.Once completed, this project will help scientifically re-assess the principle and assumptions used in the design of Internet transport protocols in the past and will also benefit several communities. First, experience in measurements and analysis is invaluable to federal, commercial, and academic institutions that are involved in mining for information in large data-sets. Second, the use of measurement tools in Computer Science education will help train graduate and undergraduate students with important skill set and help develop new pedagogical techniques. Third, the design of transport protocols that scale with technological advancement will impact the use of the Internet for scientific research and distributed analysis.