The complexity of developing efficient parallel applications for modern supercomputers has increased significantly due to available concurrency. Many applications utilize message passing for parallelization, which includes three main communication paradigms: point-to-point, collective, and one-sided communication. The one-sided paradigm allows a single process to define complete communication, decoupling communication from synchronization. This feature is crucial for runtime systems of new programming paradigms and advanced dynamic load-balancing strategies. In process interactions, wait states can occur, where a process idles while waiting for another to proceed. To address these wait states, developers need tools to detect, quantify, and identify their root causes. Unfortunately, support for identifying complex wait states in one-sided communication is limited. This thesis presents innovative methods for the scalable detection and quantification of wait states in one-sided communication, along with automatic root cause identification and optimization assessment. The proposed methods build on existing techniques by employing a parallel post-mortem traversal of process-local event traces to model runtime behavior. Performance-relevant data is exchanged on recorded communication paths, but the nature of one-sided communication complicates the availability of this information across all involved processes. A novel high-level mes
