Complex systems are defined as systems with many interdependent parts that give rise to nonlinear and emergent properties determining their high-level functioning and behavior. Due to the interdependence of their constituent elements and other characteristics of complex systems, it is difficult to predict system behavior based on the “sum of their parts” alone. Examples of complex systems include human economies and societies, nervous systems, molecular interaction networks, cells and other living things, such as bees and their hives, and ecosystems, as well as modern energy and telecommunication infrastructures. Arguably, one of the most striking properties of complex systems is that conventional experimental and engineering approaches are inadequate to capture and predict the behavior of such systems. A relatively recent and more holistic approach employs computational techniques to model and simulate complex natural phenomena and complex man-made artifacts. Complex system simulations typically require considerable computing and storage resources (processing units and primary and secondary memory) as well as high-speed communication links. Supercomputers are the technology of choice to satisfy these requirements. Because supercomputers are expensive to acquire and maintain, there has been a trend to exploit distributed computing and other large-scale computing technologies to facilitate complex system simulations. Grid computing, service-oriented architectures, programmable logic arrays, and graphic processors are examples of such technologies.

The purpose of this volume is to present a representative overview of contemporary large-scale computing technologies in the context of complex system simulation applications. The book is intended to serve simultaneously as design blueprint, user guide, research agenda, and communication platform. As a design blueprint, the book is intended for researchers and technology and application developers, managers, and other professionals who are tasked with the development or deployment of large-scale computer technology to facilitate complex system applications. As a user guide, the volume addresses the requirements of modelers and scientists to gain an overview and a basic understanding of key concepts, methodologies, technologies, and tools. For this audience, we seek to explain the key concepts and assumptions of the various large-scale computer techniques and technologies, their conceptual and computational merits and limitations. We aim at providing the users with a clear understanding and practical know-how of the relevant technologies in the context of complex system modeling and simulation and the large-scale computing technologies employed to meet the requirements of such applications. As research agenda, the book is intended for computer and complex systems students, teachers, and researchers who seek to understand the state of the art of the large-scale computing technologies involved as well as their limitations and emerging and future developments. As a communication platform, the book is intended to bridge the cultural, conceptual, and technological gap among the key disciplines of complex system modeling and simulation and large-scale computing. To support this goal, we have asked the contributors to adopt an approach that appeals to audiences from different backgrounds.

Clearly, we cannot expect to do full justice to all of these goals in a single book. However, we do believe that this book has the potential to go a long way in fostering the understanding, development, and deployment of large-scale computer technology and its application to the modeling and simulation of complex systems. Thus, we hope this volume will contribute to increased communication and collaboration across various modeling, simulation, and computer science disciplines and will help to improve the complex natural and engineering systems.

This volume comprises nine chapters, which introduce the key concepts and challenges and the lessons learned from developing and deploying large-scale computing technologies in the context of complex system applications. Next, we briefly summarize the contents of the nine chapters.

Chapter 1 is concerned with an overview of some large-scale computing technologies. It discusses how in the last three decades the demand for computer-aided simulation of processes and systems has increased. In the same time period, simulation models have become increasingly complex in order to capture the details of the systems and processes being modeled. Both trends have instigated the development of new concepts aimed at a more efficient sharing of computational resources. Typically, grid and cloud computing techniques are employed to meet the computing and storage demands of complex applications in research, development, and other areas. This chapter provides an overview of grid and cloud computing, which are key elements of many modern large-scale computing environments.

Chapter 2 adopts the view of an e-infrastructure ecosystem. It focuses on scientific collaborations and how these are increasingly relying on the capability of combining computational and data resources supplied by several resource providers into seamless e-infrastructures. This chapter presents the rationale for building an e-infrastructure ecosystem that comprises national, regional, and international e-infrastructures. It discusses operational and usage models and highlights how e-infrastructures can be used in building complex applications.

Chapter 3 presents multiscale physical and astrophysical simulations on new many-core accelerator hardware. The chosen algorithms are deployed on parallel clusters using a large number of graphical processing units (GPUs) on the petaflop scale. The applications are particle-based astrophysical many-body simulations with self-gravity, as well as particle and mesh-based simulations on fluid flows, from astrophysics and physics. Strong and soft scaling are demonstrated using some of the fastest GPU clusters in China and hardware resources of cooperating teams in Germany and the United States.

Chapter 4 presents an overview of the SimWorld Agent-Based Grid Experimentation System (SWAGES). SWAGES has been used extensively for various kinds of agent-based modeling and is designed to scale to very large and complex grid environments while maintaining a very simple user interface for integrating models with the system. This chapter focuses on SWAGES’ unique features for parallel simulation experimentation (such as novel spatial scheduling algorithms) and on its methodologies for utilizing large-scale computational resources (such as the distributed server architecture designed to offset the ever-growing computational demands of administering large simulation experiments).

Chapter 5 revolves around agent-based modeling and simulation (ABMS) technologies. In the last decade, ABMS has been successfully applied to a variety of domains, demonstrating the potential of this approach to advance science, engineering, and other domains. However, realizing the full potential of ABMS to generate breakthrough research results requires far greater computing capability than is available through current ABMS tools. The Repast for High Performance Computing (Repast HPC) project addresses this need by developing a next-generation ABMS system explicitly focusing on larger-scale distributed computing platforms. This chapter’s contribution is its detailed presentation of the implementation of Repast HPC, a complete ABMS platform developed explicitly for large-scale distributed computing systems.

Chapter 6 presents an environment for the development and execution of multiscale simulations composed from high-level architecture (HLA) components. Advanced HLA mechanisms are particularly useful for multiscale simulations as they provide, among others, time management functions that enable the construction of integrated simulations from modules with different individual timescales. Using the proposed solution simplifies the use of HLA services and allows components to be steered by users; this is not possible in raw HLA. This solution separates the roles of simulation module developers from those of users and enables collaborative work. The environment is accessible via a scripting API, which enables the steering of distributed components using straightforward source code.

Chapter 7 is concerned with the data dimensions of large-scale computing. Data-intensive computing is the study of the tools and techniques required to manage and explore digital data. This chapter briefly discusses the many issues arising from the huge increase in stored digital data that we are now confronted with globally. In order to make sense of this data and to transform it into useful information that can inform our knowledge of the world around us, many new techniques in data handling, data exploration, and information creation are needed. The Advanced Data Mining and Integration Research for Europe (ADMIRE) project, which this chapter discusses in some detail, is studying how some of these challenges can be addressed through the creation of advanced, automated data mining techniques that can be applied to large-scale distributed data sets.

Chapter 8 describes a topology-aware evolutionary algorithm that is able to automatically adapt itself to different configurations of distributed computing resources. An important component of the algorithm is the use of QosCosGrid-OpenMPI, which enables the algorithm to run across computing resources hundreds of kilometers distant from one another. The authors use the evolutionary algorithm to compare models of a biological gene regulatory network which have been reverse engineered using three different systems of mathematical equations.

Chapter 9 presents a number of technologies that have been successfully integrated into a supercomputing-like e-science infrastructure called QosCosGrid (QCG). The solutions provide services for simulations such as complex systems, multiphysics, hybrid models, and parallel applications. The key aim in providing these solutions was to support the dynamic and guaranteed use of distributed computational clusters and supercomputers managed efficiently by a hierarchical scheduling structure involving a metascheduler layer and underlying local queuing or batch systems.

Werner Dubitzky

Krzysztof Kurowski

Bernhard Schott

Coleraine, Frankfurt, Poznan

May 2011