Special projects between Bavaria and Georgia
Angewandte Informatik II
Universität Bayreuth
Computer Science Dept.
Georgia State University
Parallel Scientific Computing on Public Computing Platforms
Public computing is a type of grid computing architectures composed of autonomous volunteer workstations coordinated by a central server complex over the Internet. Exploiting these distributed resources comes at the price of unpredictable availability, fluctuating performance, and heterogeneous participant nodes. Central to addressing these problems is an efficient and accurate scheduling mechanism for the public computing architecture. In this research, we will study grid computing in general and a public computing platform called BOINC in particular. Several major issues and challenges in grid computing research will be identified. We will also introduce new scheduling schemes based on intelligent algorithms such as the ant colony algorithm. We plan to design a new peer-to-peer architecture for solving the bottleneck problem in the client-server architecture used in the BOINC system. The new architecture will be a completely distributed system and new scheduling schemes will be designed and studied.
Final report
Public computing is a type of grid or cloud computing architectures composed of autonomous volunteer workstations coordinated by a central server complex over the Internet. Exploiting these distributed resources comes at the price of unpredictable availability, fluctuating performance, and heterogeneous participant nodes. Central to addressing these problems is an efficient and accurate scheduling mechanism for the public computing architecture. In this research project, we have started to address the challenges arising from program development for public computing platforms. In particular, we have started to address the problem of transforming a given application, coming, e.g., from the area of scientific computing, to an application that can run on a public computing platform. The approach is based on the transformation of compute-intensive program parts into web services that can be executed on powerful cluster machines. This has been successfully demonstrated for a ray tracing program from computer graphics. This research will be continued by Marvin Ferber as part of his PhD thesis.