Software applications provide a wide area of features and configurations, or process large datasets. Hence, certain performance problems only occur on specific constellations which are especially revealed when the application is used by customers. Application Performance Management (APM) was introduced to monitor and analyze performance data at runtime. In this thesis, an APM system was created to visualize the performance data of a Java application. It detects and alerts unusual performance behavior while the application is executed on several machines with several configurations. Therefore, a lot of the application’s metadata is collected and compared. In the field of automobile industry, Electronic Control Units (ECUs) and their software have become more and more complex. Configuration tools are used to handle this complexity and to generate the related source code. The processing of the large datasets leads to long duration. A configuration tool called DaVinci Configurator Pro of the company Vector Informatik is used for configuration, validation and generation of the necessary basic software and runtime environment of the ECUs. This tool provides a case study for the described APM system in this thesis. The implementation of the APM system has shown that performance regressions of heterogeneous environments, which are caused by specific metadata can be detected by the reduction to a point anomaly detection problem and single comparisons for the several parameters of the metadata. The quality or reliability of the detection depends on the selection of the captured metadata. Regarding algorithms, the box plot rule-based algorithm as well as the nearest neighbor-based approach can be used for this task.