As more enterprises move HPC workloads to the cloud universe, they are facing new challenges, such as how to avoid lock-in risk and how to enhance operation agility to react for new business opportunities. Cloud technologies are continuing to advance together with conventional compute facilities and modeling tools. Distance matters less and less, and computing speed has reached the exaflops scale of performance. Distributed object computing across multiple clouds can help many enterprises manage the challenges while helping the cloud community achieve economies of scale.
This research focuses on leveraging open-source software stacks of distributed object computing such as those available at the Distributed Object Computing (DOC) Group at Vanderbilt University to modularizing some HPC workloads in order to realize the real-time computing of conventional HPC applications across multiple operating systems over multiple places and operated by multiple cloud provides/vendors.