Parallel high-performance grid computing: capabilities and opportunities of a novel demanding service and business class allowing highest resource efficiency

The hardware and software requirements for parallel applications depend on the problem size, type and the number particles / parameters, the degree of parallelization possible, the load balancing over different processors / memory, the calculation type and the input / output and storage amount. These determine the suitability of a hardware and software solution optimizing the hardware use. Computational time is the meta-parameter for optimization. The interaction with users, the management of jobs, accounting, and billing, are getting increasingly important and make further optimizations necessary. Thus, parallel high-performance grid computing has to combine the challenges of optimizing a dedicated problem within a heterogeneous hardware setting with the right management. By combination of small grid jobs with huge parallel jobs, CPU/core usage efficiencies > 99.9% can be reached. In the life-science and health-care sectors demanding IT and especially large scale grid resources are needed to analyse huge and complex systems. We showed here with MD simulations of different sizes, how the presence of large grid clusters within grid infrastructures allows now parallel high-performance grid computing efficiently and thus combines the benefits of dedicated super- computing centres and grid infrastructures. Therefore, accounting and billing can be used to motivate the users to use the hardware in a most efficient way, which not allways results in shortest computation times. Thus, we combined classic with parallel high-performance grid usage, while increasing the efficiency of IT resource providers. This is beyond the mere “yes-we-can” a great opportunity for the life-science and health-care sectors and allows reaching the maximum resource efficiency.