High Performance Grid Computing: getting HPC and HTC all together

1. High Performance Grid Computing: getting HPC and HTC all together COMPCHEM VO is engaged in designing and implementing accurate realistic multi-scale Molecular and Materials applications like GEMS, involving the usage of various HPC and HTC computational packages and allowing to overcome the shortcomings of the present highly unsatisfactory situation whereby none of the available computational platforms is suited for such requests. Front side view of Alamethicin in the bilayer membrane. THE COMPUTATIONAL PROBLEM http://compchem.unipg.it The concrete objective of the HIPEG project (within a collaboration between University of Perugia, INFN-IGI national NGI and CINECA) is to equip a Science Gateway with prototype computational applications and appropriate workflow tools allowing the combined and optimized usage of HTC and HPC platforms paving the way for the development of highly realistic more complex collaborative (in terms of expertise and platforms) scientific applications. For this reason, in order to be able to compose computational applications of different level of complexity to fit the different needs of various scientific communities, the present project focuses its action into the creation of a tool allowing a chaining of a HTC calculation following a HPC one and viceversa. DESCRIPTION OF THE WORK Focus has been put on enabling GriF (the Collaborative Grid Framework designed and developed by COMPCHEM VO -- http://www.hpc.unipg.it/grif) to route computational tasks to HPC machines (provided by CINECA) so as to enhance its Grid submission efficiency. In our case study, we have used the GAMESS-US program that has been launched on both the Grid and the IBM SP6 cluster of CINECA. In order to distribute grid-oriented GAMESS-US jobs on a HPC platform, monitor their evolution and gather the results back to the users, the whole package has been implemented as a ‘Service’ offered by COMPCHEM and CINECA making use of MPI and OpenMP parallel libraries. PROGRESS BEYOND THE STATE-OF-THE-ART To go beyond a single application submission, two prototypical HPC-HTC schemes (called HTPC skeleton) have been considered as a case study to combine in an iterative fashion an HPC and an HTC computing task. • Other partners (like CASPUR) are joining the project with the final goal to define a common layer of HTC/HPC integration. • A basic Workflow engine running entirely on the Grid combining various scientific applications has been already developed and tested on the EGI Grid. • A tight collaboration with European Middleware Initiative (EMI) people and Italian Grid Initiative (IGI) is going forward for defining the final Workflow engine combining HTC and HPC by supporting EMI Execution Services (EMI-ES).