Beowulf is a phenomenon, perhaps even a unique phenomenon. When the original Beowulf project was initiated, its goals were modest, its anticipated lifetime limited. It was not even known if effective scientific computation could be performed on clusters of personal computers. But the answer was ''yes" and the timing proved auspicious. Many researchers at diverse locations and organizations were considering similar approaches; the idea was in the air. The Beowulf project provided some of the earliest demonstrations and quantitative test results as well as supplying some of the necessary software. But today, with hundreds of sites, Beowulf-class computing no longer belongs to any one group or project. It is a tool of the computational science community, providing one alternative approach for high end computing. Its attraction to many is its flexibility, low cost, and user control. A number of efforts are underway across the country and around the world to develop new software tools that will enhance its utility and improve its ease of use. While many of these tools are under development for a broad range of system classes, Beowulfs are among the intended targets. Other such tools are being developed expressly to assist in using Beowulf clusters. A community of Beowulf users and developers has emerged, sharing results and tools, and contributing to the general understanding of the means and value of this model. The future of Beowulf-class computing is in the hands of its users. As new roles are discovered and explored, the world of Beowulf will continue to grow. But if the many challenges still confronting useful and robust application of this technology are not met in the near future with innovative tools and techniques, then the Beowulf revolution will stall. This book has been offered as one step in advancing the objectives of Beowulf-class computing and facilitating their exploitation by new users for an increasing range of applications.
