Transparent adaptation of e-science applications for parallel and cycle-sharing infrastructures

Grid computing is a concept usually associated with institution-driven networks assembled with a clear purpose, namely to address complex calculation problems or when heterogeneity and users' geographical dispersion is a key factor. However, regular home users willing to take advantage of distributed processing cannot regard this a viable option. Even if Grid access was open to the general public, a home user would not be able to express task decomposition without clearly understanding the program internals. In this work, distributed computation, and cycle-sharing in particular, are addressed in a different manner. Users share idle resources with other users provided that such resources (namely, CPU cycles) are mostly employed to execute already installed applications (e.g., popular commodity applications targeting video compression/ transcoding, image processing, ray tracing). Users need not to modify an application they already use and trust. Instead, they require only access to an available format description of the application input/output, in order to allow transparent and automatic decomposition of a job in smaller tasks that may be distributed and executed in cycle-sharing machines.