Right now, users have to specify a mix of globus, condor and pegasus profiles depending on the environment there job is running in. We should harmonize and come up with single set of Pegasus profiles that the planner should automatically translate to the appropriate values.

some pegasus profiles we should support are

runtime - runtime of jobs in seconds
memory - maximum memory a job requires in MB
diskspace - diskspace a job requires in MB
cores - the number of processors a job requires
nodes - the number of nodes a job should run on.
queue - the queue
project - the project to be charged to.

involves,

• changing code generator styles
• aggregation in clustered jobs
• pmc handling
• checkpointing
• metadata generation