Contributions

The primary contributions of this work are the following:

• a flexible framework, based on a data-flow programming model, that allows the application programmer and the dataset provider to deploy a network of application-specific and dataset-specific functionality across the grid;

• an algorithm to restructure a data-flow application graph to improve data flow across a wide-area network, based on programmer- and user-assigned properties that describe the behavior of the nodes within the graph;

• a hierarchical graph-partitioning scheme that leverages existing software to decide where to place individual application components in a way that benefits the application and abides by allocation policies set by individual administrative domains; and

• an evaluation of the I/O performance of a variety of applications using Armada.

The ideas developed in the Armada project should provide important insights in the emerging area of Grid computing, enabling computational scientists to be able to work with large remote data sets more efficiently. Their applications, ranging from weather modeling to oil exploration to nuclear weapons, have a broad effect on society.