CAREER: Beyond Akamai and BitTorrent: Information and Decision Dynamics in Content Distribution Networks

最近的研究估计，大部分互联网流量是数字内容分发，其主要组成部分是需要严格时间期限的应用程序，如内容流和实时服务。始终可用的基于服务器的内容分发网络（cdn）的过高成本，以及低成本点对点（P2P）网络中缺乏服务保证，导致人们追求两种系统的灵活混合，以寻求充分利用各自的优点。这种混合需要跨节点、跨信息和跨节点可用的共享资源进行协调，这构成了本项目的三个研究重点。该项目的第一个目标是利用博弈论、学习和随机网络的思想，开发一个系统的分析框架来实现这种协调。其目的是利用分析模型捕捉大规模内容分发系统的基本属性，并通过使用该框架将诸如文件共享、流媒体、基于云的服务和设备对设备无线网络等应用程序绑定在一起。第二种方法是，通过披露该体系对参与主体的价值，为确保经济可行性开出处方。最后，所开发的基本分析方法将普遍适用于内容分发的核心问题之外。分析的见解将通过基于智能手机的服务器辅助P2P内容流网络进行验证。更广泛的影响：该项目得到了教育和推广议程的支持，旨在使学生熟悉所开发的分析结果，以及如何最好地将其应用于现实世界。核心部分是为高年级学生和一年级毕业生开设的一学期课程。在这门课程中，通过AM-Droiders智能手机实验室的应用程序开发，课堂教学得到了研究经验的补充。推广活动包括每年为高中生举办的为期一周的夏令营的演示和说明性项目，以及来自多所大学（包括HBCU的Prairie View A&amp；M）的学生可以互动参与的远程研讨会。通过与工业界的紧密合作，该项目得到进一步加强。

Recent studies estimate that a majority of Internet traffic is digital content distribution, whose dominant component is applications that require tight time deadlines, such as content streaming and realtime services. Excessive costs of always-available server-based content distribution networks (CDNs), and a lack of service guarantees in low-cost peer-to-peer (P2P) networks have led the pursuit of agile hybrids of both systems that seek to leverage the best of each. Such a hybrid requires coordination across peers, across information, and across shared resources available to the peers, which form the three research thrusts of this project. The first objective of this project is the development of a systematic analytical framework to achieve this coordination, using ideas drawn from game theory, learning and stochastic networks. The aim is to capture the fundamental properties of large-scale content distribution systems using analytical models, and bind together applications such as file sharing, streaming, cloud-based services, and device-to-device wireless networks by employing this framework. The second is to generate prescriptions for ensuring economic viability by using revelations on the value of the system to participating agents. Finally, the basic analytical methodology developed will be applicable generally beyond the core problem of content distribution. The analytical insights will be validated via a smart-phone-based server-assisted P2P content streaming network. Broader Impact: The project is supported by an education and outreach agenda that aims at familiarizing students with analytical results developed and how best to apply them in a real-world setting. A core component is a semester-long course for seniors and first-year grads in which classroom teaching is complemented by research experience through app development at the AM-Droiders smart-phone laboratory. Outreach includes demonstrations and illustrative projects at a yearly week-long summer camp for high school students, as well as tele-seminars in which students from multiple universities (including Prairie View A&M, an HBCU) can participate interactively. The project is further enhanced by strong collaborations with industry.