Browse Books

Reviewer: Valentin Cristea

Computational grids represent a new approach to using computers. By analogy with the electric power grid, a computational grid includes very fast communication networks that link people, high-performance computers, sensors, and other resources to create a universal source of computing power. Such a grid can provide pervasive, dependable, consistent, and inexpensive access to advanced computational capabilities and databases. Computational grids do not yet exist, but several grid projects are under way in the United States. The essential building blocks for grids are now in place, including advanced optical networks, parallel computers, communication protocols, distributed software, and security mechanisms. Nevertheless, major technical challenges have to be overcome before all these pieces can be integrated into large networks of geographically distributed resources. The authors hope that, as a result of efforts such as this book, people will better understand how grids will be used, how they work, the major milestones for their progress, and how to collaborate to obtain interoperable technologies. The book explains why grids are so important; how they are constructed and how they operate; what the grid applications will be and how they will be developed; the state of the art in relevant research disciplines; and subjects for further research. With these goals in mind, the book is intended for a wide variety of readers: software engineers, computer center managers, technology program managers, chief information officers, scientists, application developers, students, researchers, and research managers. The book contains two introductory chapters and 20 technical chapters grouped into four sections. In chapter 1, the emerging computational grid infrastructure is compared to infrastructures for railroads, telephones, power and light, and banking. The chapter describes how different users in computational science and engineering, experimental and observational science, industrial engineering, and corporate communications could use the grid today to solve nontrivial problems. The support given by several innovative programs, initiatives, and collaborations for the development of computational grids is also presented. Chapter 2 provides a high-level view of the expected purpose, shape, and architecture of the future computational grids and answers six important questions. The chapter explains that the grids are useful for enhancing human creativity and performance; states that the main grid applications are distributed supercomputing, high-throughput, data-intensive, and collaborative applications; describes the main user categories; explains how the applications will be built; describes the grid architecture; and identifies the research challenges. The chapter serves as a road map for the rest of the book and includes pointers to the remaining chapters. Chapters 3 through 6 discuss applications in four important areas: distributed supercomputing, real-time widely distributed instrumentation systems, data-intensive applications, and teleimmersion. Both the new kinds of applications enabled by grids and the new technologies (models, techniques, tools, and policies) involved in their development are presented. As algorithms evolve and the grid infrastructure improves, more applications will be able to exploit distributed supercomputing. New classes of applications that combine remote simulation, real-time interaction, large-scale databases, and teleimmersion can be used as ideal testbeds for the development of grid infrastructures. Chapters 7 through 10 describe programming tools. They cover the development of application-specific tools; compilers, languages, and libraries; object-based approaches; and high-performance commodity computing. Each chapter presents general aspects related to the integration of these tools into the grid environment and gives examples illustrating the actual achievements. The three-tier architecture used in commodity computing and some other commodity technologies are presented in chapter 10. Chapters 11 through 16 are concerned with the services required to support grid applications and tools. Chapter 11 presents the Globus toolkit. The other chapters cover high-performance schedulers, high-throughput resource management, instrumentation and management, performance analysis and visualization, security, accounting, and assurance. These services simplify application development by defining a virtual machine that is easier to manipulate than low-level hardware. The services and their interfaces define the grid architecture. Chapters 17 through 22 cover various aspects of the low-level infrastructure on which grids will be constructed: computing platforms, advanced network protocols, mechanisms to manage network quality of service, operating system and network interfaces, network infrastructure, networking technologies, and metacomputing testbeds. The last chapter notes the lessons taught by these testbeds, and reviews attempts to construct such testbeds. The book can be used by practicing professionals or as a text for a senior undergraduate or graduate course in advanced networking, distributed computing, or computational grids. Advice about using the book in a course is given in the introduction. The book includes a glossary, a bibliography with 589 references, a list of contributors, and an index. The references are representative of the subjects discussed. Recommendations for further reading are included in each chapter, and a related web site (http://www.mkp.com/grids) contains pointers to documents relevant to the subject. The authors have found an equilibrium among concepts, technology, and practice. They present the state of the art in networking and computational engineering. Several issues (including grid architecture and resource management) are approached from different perspectives in different chapters. This reflects the complexity of the problems discussed and the lack of a standard approach to solving them.