The current Internet offers a diverse set of applications for end-users (e.g., online gaming, IPTV, VoIP, VoD). These applications require a timely delivery of packets to avoid the deterioration of user-perceived performance. Slow web- browsing, momentarily drops of voice in Internet telephony and lagging remote connections are some of the most com- mon problems that affect the usability of Internet services. Ideally, if applications had access to measurements that quan- tify network performance, they could dynamically adapt to network conditions and improve user satisfaction. Most im- portantly, when application service quality falls below ex- pectations, online measurements help users understand who to blame and what to do to bypass the problem. We propose a troubleshooting tool that runs at end-hosts to automatically detect and diagnose performance degrada- tions. The tool has three aspects: detection of performance degradation, identification of the location of the problem and characterization of the potential causes. In this work, we focus on the detection aspect, which requires continuous monitoring at end-hosts.