Researchers need to perform a multitude of tasks in order to get their computational work done on HPC systems and get the results back on their workstations. They need to log in to the supercomputer or cluster, transfer input data, submit a batch job, wait for it to finish, then transfer results back to the workstation. If they divide their computational work over different clusters or supercomputer, they need to do this sequence of tasks for each system separately and combine the results from each system on their workstation - an inefficient use of researchers' time.
KAUST IT Research Computing automated these tasks by implementing an interface between a research application running on a local workstation and the remote HPC systems. This interface automatically executes the submission of the application to HPC systems and the retrieval of the results. These interfaces are called KAUST HPC add-ons, as they add an HPC capabiity to the research application running on the workstations. These HPC add-ons have been implemented for the following research applications: MATLAB, VASP/MedeA, ADF, ParaView and Materials Studio (Biovia). The method of implementation of the HPC add-ons is tailored for each application.
The new system gives researchers reduced time-to-solution due to automation of the workflow, easier access to the HPC systems - Noor clusters and Shaheen supercomputer - and less time spent mastering unfamiliar Linux commands. "Usage of the clusters was made much easier with the MATLAB HPC Add-on", stated post-doc Maged Serag of the Biological and Environmental Sciences and Engineering Division of KAUST. "I managed to reduce my turnaround time from 24 hours to 15 minutes."