The term petascale computing denotes massively parallel systems capable of executing 10¹⁵--that is, one quadrillion--floating point operations per second. A study published in 2001 predicted the first petascale systems to be available in 2009 [1]. And indeed, in November 2008, the most powerful computing system in the world, according to the TOP500 list, was an IBM cluster with a sustainable performance of 1.1 petaflops running the Linpack benchmark. Only two machines in this list reach the petaflop order of magnitude, but their number is expected to grow soon.

In the 1990s, the previous high-performance computing (HPC) milestone--teraflop computing--allowed for the development of accurate simulations of complex systems and phenomena, such as earthquake prediction, pharmaceutical research, and weather forecasting. By reaching the petascale level, the scientific community can obtain new and significant results in these and other fields. However, nowadays, few existent HPC applications scale well when using more than a few hundred processors. To harness the computing capabilities of these petascale systems, new algorithms and applications should be developed, both to manage the increasing complexity of the events simulated and to efficiently exploit the higher level of concurrency available in these systems.

This book is the first one devoted specifically to the challenges associated with this new computing milestone. The book contains 24 chapters, representing the contributions of almost 100 researchers in the fields of scalable algorithm design, computational science, and engineering applications, methodologies, and tools. Some chapters cover how petascale computing will help a wide range of HPC applications, including multiphysics simulations, cosmological evolution, climate modeling, and molecular dynamics, while others study novel computer architecture and communications issues, benchmark characterization, performance analysis, and new programming models and paradigms. Each chapter provides in-depth information on a specific issue, showing how petascale computing may change the way the problem is addressed, including performance results and many additional references. This book is an inspiring starting point to understanding how these new HPC systems will enable new discoveries in science and engineering in the 21st century.