18-845: Internet Services Project Proposal Abstract Problem Detection and Diagnosis for Large-Scale Mobile Wireless Video Streaming Applications Nathan D. Mickulicz For project 2, I would like to propose a topic related to my research into the diagnosis of performance problems in Internet services over large-scale wireless networks. Overall, my research attempts to characterize the performance of a particular technology where resources must be transferred to and from an Internet service (usually a "cloud" server or cluster) to multiple end-users, where the final hop to the end-user is a wireless link (Wi-Fi, 3G/4G, etc.). Currently, the "particular technology" here is video streaming to smartphones, but could be a number of other Internet services. For the course project, I would like to develop a small portion of this research related to understanding the performance characteristics of the system, which is a prerequisite for diagnosing the root cause of performance problems and then ultimately mitigating these problems. Specifically, I would like to enumerate and evaluate the instrumentation points available in the system, develop the necessary protocols and other infrastructure for reliably transporting the data from these instrumentation points to a backend service for analysis, develop the tools to analyze this data and compute performance statistics and find anomalies, and then finally test the system with injected performance problems and determine detection accuracy (all w.r.t. some fault model, which will probably be 2-3 faults at most due to time constraints).