Autonomic Solutions for Scalable, Reliable Computing

The trend toward larger high-performance computing (HPC) systems is leading to an increased demand for scalable software. However, HPC systems are also becoming more complex making the development and use of scalable software increasingly difficult. As a result, very few applications run effectively (or at all) at extreme scales. Furthermore, many application problems are manifested only at large scales, and tools that can diagnose effectively functional and performance complications in these applications are also lacking. With the availability of hundred thousand processor systems and the advent of million processor systems, the lack of scalable software systems is increasingly problematic. This problem is particularly concerning for capability computing where the most powerful computational resources are used to solve large, demanding problems.