Highthroughput and Trivially Parallel
An example could be that you have 10 different models you wish to process. You could submit this as one job which does each in order, but assuming that order is not important (they are independent) then you could submit this as 10 jobs each doing one model. The system might then be able to run all 10 concurrently and so you would get results back in one tenth the time.
One caveat to the above is if you are using a package with license restrictions then running 10 simulations concurrently might not be possible, and the rate of running will be throttled back by the number of licenses available at any instant in time.
More information, including implementation and running, can be found on our external HPC website using the below link: