DISTRIBUTED STATISTICAL APPLICATION PROFILER

A method for profiling an application using partial profile data from multiple instances of the application deployed for various users is presented herein. The method enables reconstruction of a complete and accurate application profile by distributing instrumentation across many instances of an application to sample only parts of an application by each instance. The method allows for highly detailed and accurate profiling of real world use cases for an application without affecting performance of the application because the sampling is distributed across the user base. The method may allow for tuning of the detail of sampling in order to collect as much detail as possible without affecting the performance of the application. Tuning may also allow for selective sampling of particular classes of users and devices