= Tasks distribution over different computational resources =  

Tasks distribution is implemented using [http://www.gridgain.com/javadoc/org/gridgain/grid/GridTaskAdapter.html GridGain GridTaskAdapter]. In this scenario, a cloud should have several segments. Workers in the cloud can belong to one or more segments (Worker [http://www.molgenis.org/wiki/ComputeStartExample8 settings example]). Every script should contain information about execution environment or cloud segment, where it should be executed. First, We found all Worker nodes from this segment. Then, a script is sent to one of them  

{{{
public Map<? extends GridJob, GridNode> map(List<GridNode> subgrid, Script script) throws GridException
{
    Map<ExecuteCommandLineJob, GridNode> jobs = new HashMap<ExecuteCommandLineJob, GridNode>(1);

    //get the id of the execution environment
    String eeID = script.getExecutionEnvironmentID();

    //list of worker nodes
    List<GridNode> workerNodes = new ArrayList<GridNode>();

    //find worker nodes for the job execution
    for (Iterator it = subgrid.iterator(); it.hasNext();)
    {
       GridNode node = (GridNode) it.next();
       String segment = node.getAttribute(eeID);

       if (segment != null)
       {
          if (segment.equals("true"))
          {
             workerNodes.add(node);
          }
       }
    }
 
    if(workerNodes.size() == 0)
    {
        ...
        return null;
    }

    ...
    ExecuteCommandLineJob myJob = new ExecuteCommandLineJob(script);

    //the worker can be pick-up randomly
    //int i = rand.nextInt(workerNodes.size());

    ... 
    jobs.put(myJob, workerNodes.get(i));
    return jobs;
}

}}}


Back to [http://www.molgenis.org/wiki/ComputeStartExamples Examples]