This paper describes an integrated performance monitoring environment for parallel systems. It consists of:
1) A distributed monitoring system that collects performance data from instrumented applications and sends it to analysis tools.
2) Graphical and command-line profiling and visualization tools that analyze the performance data to identify bottlenecks.
3) A common graphical interface that provides a consistent way to instrument applications, start tool runs, and view performance results across different tools.
The environment aims to handle large amounts of performance data from massively parallel applications and provide insights at both the application and system level. It is initially targeted for the Intel Paragon but is designed to support different programming models.