The Data Analytics division at Quantum Dynamics needs to run large, interruptible, batch-processing simulations (15-20 GB of data per job) on data stored in Amazon S3, with the output stored in another S3 bucket for analysis; which solution meets these requirements most cost-effectively?