A Docker image built with monitoring and profiling tools for performance measurement in a containerized environment.
These tools are listed below and the main purpose is to measure, monitor and/or profile data on MPI and OpenMP applications.
This environment was successfully tested with some applications of the NAS Parallel Benchmarks set.
The image is available through command line:
docker pull lraraujo/container-tools:1.0
The Dockerfile might also be downloaded and edited to remove any unnecessary tools or to previously download some desired applications.
Before run the container, the users that will utilize tools as Likwid or PCM need to execute sudo modprobe msr. For further information check Setting up access for hardware performance monitoring.
Users that will utilize Linux perf or PCM need to check /proc/sys/kernel/perf_event_paranoid and set the value desired, otherwise, will get an error message trying to execute these tools.
The command below will initialize the container as an interactive process (shell).
docker run -it --privileged --name <name> lraraujo/container-tools:1.0
Thus, users are able to use command line, inside the container, to download, compile and execute applications with the tools provided.
All installed tools are listed below along with its documentation.
A recommended tool to visualize the files provided by Score-P and Scalasca is Cube.