Research and Infrastructure Development for Public-Resource Scientific Computing

University Of California-Berkeley, Berkeley CA

Investigators

Abstract

The SETI@home project has pioneered public-resource computing, using millions of computers in homes and offices around the world to perform scientific computing. This approach, though it presents some difficulties, has provided unprecedented computing power and has led to a unique public involvement in science. The researchers propose research in which anew software infrastructure for public-resource computing will be developed. The researchers call this the Berkeley Infrastructure for Distributed Computing (BIDC). Among the design goals of BIDC, two are of special importance: BIDC will provide mechanisms for efficiently and economically distributing data from project servers to participant hosts, which typically have commercial ISP connections. The mechanisms will use noncommercial networks like Abilene and organizational LANs where possible, and will use idle link capacity, similarly to the use of idle CPU cycles, with little or no impact on regular network usage. BIDC will be open-source, and will have an open architecture that will allow multiple independent projects to share a single participant base. Participants have fine-grain control over how much of their resources are used, and how they are divided among the projects. New projects can be started easily, and function even if existing projects fail. The researchers will use BIDC for upcoming SETI projects, including new searches using the Parkes observatory in Australia, the future multibeam receiver at Arecibo, and wider frequency bands from the current receiver at Arecibo. In addition the researchers will encourage and support the use of BIDC by other distributed projects; there are potential users in numerous areas including climatology, oceanography, physics, mathematics, and ecological science. There will be social and educational components as well. Public-resource computing projects succeed only if they attract and retain participants. In SETI@home, the researchers have developed a number of schemes for encouraging and harnessing friendly competition among users, and for keeping users informed and interested about the contributions of their computers to research. BIDC will include variants of these schemes that can be used within and across the projects based on it. The proposed research will advance knowledge in Internet-scale public-resource computing, which has emerged as an important direction both in Computer Science and in numerous application areas. In particular, the research will expand the range of applications for which public-resource computing is effective, including applications that involve large data. It will allow projects to develop and deploy these applications easily, and will allow many such projects to efficiently share a single resource pool. By carrying out the SETI@home project the researchers have demonstrated their qualification to do this research, and are in a unique position to connect new projects with their user base of 3,500,000 people worldwide. The research will have a broad impact. It will enable computational techniques that were previously impractical because of their resource requirements, and will therefore make possible new types of research in many scientific areas. It will provide a framework for experimental research in other types of Internetscale systems (storage, semantic-based searching, security and communication). By enabling new scientific computing projects, it will help educate and inform the worldwide public in multiple scientific areas, it will directly involve the public in science, and will give the public a voice in the directions of science research.