WebProfiler: Cooperative diagnosis of Web failures

Despite tremendous growth in the importance and reach of the Web, users have little recourse when a Web page fails to load. Web browsers provide little feedback on such failures, typically offering only generic suggestions such as re-checking the URL or the machine´s network settings. Hence, users are often unable to diagnose Web access problems, and resort to haphazardly modifying their settings or simply trying again later. We advocate a client-based collaborative approach for diagnosing Web browsing failures. Our system, WebProfiler, leverages end-host cooperation to pool together observations on the success or failure of Web accesses from multiple vantage points. These are fed into a simple, collaborative blame attribution algorithm. Our evaluation on a controlled testbed shows WebProfiler can accurately diagnose 3.6 times as many failures than possible from a single client´s perspective. We present the design and prototype implementation of WebProfiler for an enterprise network.