A platform for end-to-end mobile application infrastructure analytics using system log correlation

Abstract

Monitoring and analyzing the performance of mobile applications is challenging when the applications rely on remote services and functioning networks. Root causes of failures and performance anomalies of client-server applications are difficult to pinpoint because of their distributed nature. In this paper, we introduce the Mobile Infrastructure Analytics System (MIAS), which helps efficiently detect and debug application faults in a distributed environment, holistically analyzing application and network activity across client devices, application servers, database servers, etc. MIAS collects hypertext transfer protocol (HTTP) session data and system logs from servers and instrumented mobile applications, automatically correlates HTTP sessions with server activity, detects anomalous behavior using statistical techniques, and extracts a small and relevant set of log entries for manual inspection. We show how faults were detected and root causes pinpointed just by glancing at the evidence, for a real-world bookstore mobile application.