Up, Down and Around the Stack:

Abstract

Broadband characterization has recently attracted much at tention from the research community and the general public. Given this interest and the important business and policy implica tions of residential Internet service characterization, recent years have brought a variety of approaches to profiling Internet servic es, ranging from Web-based platforms to dedicated infrastructure i nside home networks. We have previously argued that network-inte nsive applications provide an almost ideal vantage point for broa dband service characterization at sufficient scale, nearly conti nuously and from end users. While we have shown that the approach is indee d effective at characterization and can enable performance c omparisons between service providers and geographic regions, a k ey unanswered question is how well the performance characteri stics captured by these network-intensive applications can pred ict the overall user experience with other applications. In this paper, using BitTorrent as an example network-inten sive application, we present initial results that demonstrate h ow to obtain estimates of bandwidth and latency of a network conne ction by leveraging passive monitoring and limited active measur ements from network intensive applications. We then analyze user e xperienced web performance under a variety of network conditio ns and show how estimated metrics from this network intensive application can serve as good web performance predictors. Categories and Subject Descriptors C.2.2 [ Communication Systems Organization ]: Computer Communication Networks— Network Protocols ; C.2.5 [ Communication Networks ]: Local and Wide-Area Networks— Internet ; C.4 [ Performance of Systems ]: Performance Attributes General Terms Experimentation, Performance, Measurement