NBD: Dipzoom: A Global Ecosystem for Internet Measurements

互联网测量对于加强和管理互联网基础设施以及扎根的互联网性能研究至关重要。然而，获得高质量的代表性测量仍然是一项艰巨的任务，需要专业知识和专业关系，而相对较少的人拥有这些知识和专业关系。这项研究的目标是探索一种方法，以提供多样化、代表性和易于访问的按需互联网测量。具体地说，这个项目中的方法包括设计、原型设计和研究一个名为DipZoom(For？Deep Internet Performance Zoom-in？)的系统，该系统基于两个关键思想。首先，DipZoom认识到，任何一家提供商在建立一个代表互联网规模和多样性的平台时都会面临困难，因此它实施了一项撮合服务，使用P2P概念将需要测量的实验者与外部测量提供商联系在一起。因此，DipZoom利用实验者自己部署了大量且种类繁多的测量点。其次，DipZoom提供了整个测量提供商网络的统一视图，可以从参与者S本地计算机访问和控制该视图。通过消除招募个人测量探头并与之互动的需要，DipZoom应该会大幅降低进行高质量互联网性能研究的准入门槛。该项目的智力价值在于上述思想，以及这些思想所涉及的一些新概念和技术，如测量提供商的排名、测量主机的选择性校准、针对给定测量的测量主机的选择、识别潜在攻击和开发对策等。更广泛的影响包括使广大技术社区能够访问高质量的网络测量，并通过允许网络课程纳入复杂的网络测量实验来加强计算机网络本科教育。

Internet measurements are crucial for enhancing and administering the Internet infrastructure as well as for grounded Internet performance research. Yet obtaining high quality representative measurements remains a daunting task, requiring expertise and professional connections that relatively few possess. The goal of this research is to explore an approach for providing diverse, representative, and easily accessible on-demand Internet measurements. Specifically, the approach in this project involves designing, prototyping, and studying a system named DipZoom (for ?Deep Internet Performance Zoom-in?), which is based on two key ideas. First, recognizing the difficulty any single provider would face in building a platform representative of the scale and diversity of the Internet, DipZoom implements a matchmaking service instead, using P2P concepts to bring together experimenters in need of measurements with external measurement providers. Thus, DipZoom leverages experimenters themselves in deploying a large number and variety of measuring points. Second, DipZoom provides a unified view over the entire network of measurement providers, which is accessible and controllable from a participant?s local machine. By removing the need to recruit and interact with individual measuring probes, DipZoom should drastically lower the barrier of entry for conducting quality Internet performance research. The intellectual merit of this project lies in the above ideas and in a number of new concepts and techniques these ideas entail, such as ranking of measurement providers, selective calibration of measuring hosts, selection of measuring hosts for a given measurement, identification of potential attacks and development of countermeasures, and so on. The broader impact includes making high-quality network measurements accessible to a wide technical community, and enhancing undergraduate education in computer networks by allowing networking courses to incorporate complex network measurement experiments.