bscppc/class-pycompss-rotterdam repository overview

⁠COMPSs and Rotterdam

Example of integration between COMPSs and Rotterdam. It executes a sample benchmark demo (Java matrix multiplication).

⁠Prerequisites

Rotterdam endpoint 192.168.7.28

Prometheus endpoint 192.168.7.28

⁠Installation and demo execution

1. Pull the docker image

   docker pull bscppc/class-pycompss-rotterdam
   

   Copy

2. Create the container. The port number to expose is not relevant, but it is recommended to use the same port number in the host and the container.

   docker run -d -it --name class-compss-rotterdam -p 25000:25000  bscppc/class-pycompss-rotterdam
   

   Copy

3. Access to the container.

   docker exec -it class-compss-rotterdam /bin/bash
   

   Copy

   Both the matmul example and the configuration files are located in /root/matmul.

   cd  /root/matmul
   

   Copy

4. Execute the matmul demo

   runcompss --master_name=192.168.7.28 \
   	  --master_port=25000 \
   	  --project=project.xml \
   	  --resources=resources.xml \
   	  --scheduler="es.bsc.compss.scheduler.rtheuristics.RTHeuristicsScheduler" \
   	  --task_monitor=es.bsc.compss.types.DeadlineMonitor \
   	  --scheduler_config_file=config.file.iterative \
   		matmul.iterative.Matmul 3 3 300
   

   Copy

   The --master_name corresponds to the IP of the machine that executes the container. It must be connected to the Rotterdam endpoint.

   The --master_port correspods to the port of the host machine defined in the binding of the step 2.

   The last parameter 300 corresponds to the number of iteration of the simulated control loop of the application. It can be modified: as this number increases, the execution time of the application, increases as well.