Abstract

We will assert that the key observation in the software engineering of scientific systems1 in recent years is that such systems are being constructed and used in highly distributed environments–scientists across the world are working together with their colleagues in search of answers to previously unimaginable scientific problems (eg, accurate and precise early detection of cancer, mapping out of the human genome, earthquake simulation, high energy physics computations, and the like). The principal enabling technology of these systems has been grid-computing technologies. Grid computing connects dynamic collections of individuals, institutions, and resources to create virtual organizations which support sharing, discovery, transformation, and distribution of data and computational resources. Distributed workflow, massive parallel computation, and knowledge discovery are only some of the applications of the grid.
In the last few years, our research group at USC has studied a number of grid computing technologies by examining their as-implemented architectures, and comparing and contrasting them with their as-intended architecture, a five-layer grid “reference architecture” by Kesselman et al. 2, shown in Figure 1. We published the results of one such study3 in 2005, and an addendum to the study in 20094. In these studies, we examined eighteen off-the-shelf, open source grid-computing technologies, including a major data-grid technology called OODT, developed by NASA, and the pervasive computational grid technology, Globus. The results of our study yielded three critical conclusions:(1) the requirements for grid systems are very …

Metadata

publication

year

2013

publication date

2013

authors

C Kesselman

link

https://citeseerx.ist.psu.edu/document?repid=rep1&type=pdf&doi=7af79be118be69358f14e4fe4d25e9cbbaf4a745

resource_link

https://citeseerx.ist.psu.edu/document?repid=rep1&type=pdf&doi=7af79be118be69358f14e4fe4d25e9cbbaf4a745