World-wide Web Traffic Characterization with Application in In-network Caching and Prefetching

最初的万维网（WWW）体系结构——它需要从客户机到可能的远程服务器的网络连接来检索每个文档——是不可伸缩的。服务器和网络的带宽有限，无法跟上Web使用的持续增长，特别是随着带宽密集型媒体类型（如视频）的普及而增长。该项目通过描述WWW流量特征，并将该特征应用于设计和评估网络内缓存和预取文档的算法，来解决Web可伸缩性问题。该项目有五个目标：1)收集来自四个不同交通来源的痕迹；2)通过扩展使用Chitra跟踪分析工具收集、可视化、分析和建模Web流量的方法，准备处理收集到的大量跟踪；3)利用目标二的技术，根据目标一收集到的痕迹来表征Web流量；4)在大学和社区网络的背景下，开发客户端、服务器和Web流量的预测模型和模拟，允许设计和评估网络内文档缓存算法，以减少Web服务器和网络的带宽需求。仿真将使用工作负载表征的结果。该模型将允许网络管理员在现有网络中放置和配置网络缓存（例如代理服务器）；5)通过仿真设计和评估预取文档的算法，从而部分隐藏延迟。

The original World-Wide Web (WWW) architecture--which requires a network connection from a client to a possible distant server for each document retrieved--is not scalable. Servers and network have limited bandwidth that cannot keep pace with continuing growth in Web usage, particularly as the popularity of bandwidth-intensive media types, such as video, grows. This project addresses Web scalability by characterizing WWW traffic and applying that characterization to design and evaluating algorithms to cache and prefetch documents within networks. The project has five objectives: to 1) collect traces from four disparate traffic sources; 2) prepare to handle the large volume of collected traces by extending methods to collect, visualize, analyze, and model Web traffic using the Chitra trace analysis tool; 3) use techniques from objective two to characterize Web traffic based on the traces collected for objective one; 4) develop a predictive model and simulation of clients, servers, and Web traffic in the context of universities and community networks that permits the design and evaluation of algorithms for in-network document caching to reduce the bandwidth requirements of Web servers and networks. The simulation will use results from workload characterization. The model will allow a network manager to place and configure in network caches (e.g. proxy servers) within an existing network; and 5) design and evaluate, through simulation, algorithms to prefetch documents, thereby partially hiding latency.