DISTRIBUTED RESEARCH RESOURCES: Instrumenting and Measuring Distributed Internet Applications: Trace Collection and Dissemination

本项目涉及采用双管齐下的方法对分布式Internet应用程序、仪器应用程序进行测量，然后使用这些应用程序收集数据。研究人员调查应用程序，如多人游戏和点对点系统，由于缺乏跟踪工作负载而受到阻碍。尽管web服务器跟踪广泛可用，但这些应用程序的跟踪却不是。为了解决这些限制，该项目旨在部署一些流行的应用程序，收集其行为的详细痕迹，并将这些痕迹传播给研究社区。这项工作包括两个协同的组成部分：跟踪收集，其中现代Internet应用程序将被用来收集其行为的应用程序、网络和操作系统级别的跟踪，以及跟踪传播，其中基于web的存储库将被设计用于向研究社区广泛传播跟踪。跟踪传播涉及存储库管理，以及诸如隐私之类的附加考虑。这项工作，集中于测量和测量分布式应用程序的行为，主要集中在：测量和表征网络多人游戏测量和表征点对点（P2P）应用程序测量和表征低级应用程序资源使用更广泛的影响：该跟踪存储库将作为全国研究人员的资源，并将启用或帮助各种研究领域，如Internet应用程序的工作负载建模、当前Internet应用程序的设计和性能调优、协议设计以及服务器操作系统和Internet托管平台的设计。此外，跟踪工作量也将用于分布式系统课程，以帮助学生了解测量数据在实验计算机科学研究中的使用。

This project, involving a two-pronged approach to the measurement of distributed Internet applications, instruments applications and then uses these applications to collect data. Researchers investigating applications, such as multiplayer games and peer-to-peer systems, have been hampered by the lack of trace workloads. Although web server traces are extensively available, traces of these applications are not. To address these limitations, this project aims at deploying some popular applications, gather detailed traces of their behavior, and disseminate these traces to the research community. The effort involves two synergistic components: Trace collection, where modern Internet applications will be instrumented to gather application-, network-, and OS-level traces of their behavior, and Trace dissemination, where a web-based repository will be designed to disseminate the traces broadly to the research community.Trace dissemination involves repository management, and additional considerations such as privacy. The work, concentrating on instrumenting and measuring the behavior of distributed applications, focuses on: Measuring and Characterizing Networked Multiplayer Games Measuring and Characterizing Peer-to-Peers (P2P) Applications Measuring Low-Level Application Resource UsageBroader impact: This trace repository will serve as a resource to researchers nationwide and will enable or aid a variety of research areas such as workload modeling of Internet applications, design and performance tuning of current Internet applications, protocol design, and the design of server operating systems and Internet hosting platforms. Moreover, the trace workload will also be used in a distributed system course to help students learn about the use of measurement data in experimental computer science research.