2011-10-27
MPI Job Reference
Let's say you want to run a MPI job, but don't know how. Well here are all the possible parameters that can be used with MPI jobs:
:type (required) - The type of job to run. For MPI jobs, this should be set to :mpi.
:keyname (optional) - The keyname associated with the AppScale cloud that this job should be run over. Defaults to 'appscale' if not provided.
:nodes_to_use (required) - The number of machines to run the MPI job over.
:procs_to_use (optional) - The total number of processes to run (not the number of processes per node). Must be at least as large as :nodes_to_use.
:code (required) - The location in the underlying storage system of the code to run.
:output (required) - The location in the underlying storage system that the result of the job (the standard out and standard error) should be saved to.
:storage (optional) - The storage backend to use. Defaults to 'appdb' (whatever database AppScale is running), and can also be 's3', 'gstorage', or 'walrus', for Amazon S3, Google Storage, or Eucalyptus Walrus, respectively.
:EC2_ACCESS_KEY (optional) - The EC2 access key to use when using S3, Google Storage, or Eucalyptus Walrus as the storage backend.
:EC2_SECRET_KEY (optional) - The EC2 secret key to use when using S3, Google Storage, or Eucalyptus Walrus as the storage backend.
:S3_URL (optional) - The S3 URL to use when using S3, Google Storage, or Eucalyptus Walrus as the storage backend.
