Where does all the time go when performing calculations in a grid?

In any grid computing environment, you have data nodes, and you have computation nodes. Performance is important, so is it better to send the executables to the data? Or send the data to the executables? Or both? The questions are even more important when your grid is distributed across a number of geographically dispersed sites. This article examines the challenges we all face in making these design and implementation choices, in this four-part series of articles.