A Comparison of Techniques for Distributing File-based Tasks for Public-Resource Computing

Public-resource computing (PRC) projects use volunteered computational resources in order to accomplish some goal [1]. Because the projects are so computationally intensive and only a small percentage of the public participates in them, improvements in efficiency are very important. One way to improve the efficiency of PRC projects is to decrease the amount of time wasted duplicating work unnecessarily. One cause of this waste is distributing tasks to clients in a suboptimal manner. This work focuses on file-based tasks that require clients to download data files to perform the tasks. We propose a method of distributing tasks and compare the waste it produces to the waste created by the other currently used techniques.