The next level, and biggest payback, is often referred to as "fine grain" parallelization. This is more difficult than coarse grain because it requires that the application itself be modified.
This challenge is especially significant for applications that drive critical decision-making. Examples include R&D strategies in the pharmaceutical and biotech industry, risk and portfolio balancing decisions in the investment industry, and exploration choices in energy development.
In all cases, these models tend to be intensely complex and algorithm-specific. Simply splitting the inputs is not enough to meet elapsed time reduction goals. At this level of complexity, chunking the application may significantly reduce the elapsed time -- and may even give different results.
To take advantage of grid computing, these complex applications require more sophisticated analysis and an implementation toolkit. This is especially true for Monte Carlo-based algorithms that use very specific random number generators; non-linear (quasi-Newtonian calculations) such as non-linear mixed effect models; and linear algebra-based calculations, or geometric calculations used for molecular dynamics. For these, the answer lies in a new concept: the application grid, the latest evolution on the road to more effectively exploiting the capacity made available by the processor grid.
The application grid is independent of the processor grid; it not only parallelizes the computation, enabling it to run across multiple CPUs -- it also retains the mathematical integrity and provides the necessary granularity to achieve the reduced elapsed time results and scalability.
By enabling even the most complex, mission-critical business applications to leverage the resources made available by the processor grid, while avoiding any costly and time-consuming "application surgery" and retaining all algorithms and underlying logic, the application grid offers companies the best of both worlds -- faster and better decisions.
IT Manager’s Journal | Grid computing can aid complex applications
- السفيرة, at 5:47 AM
Great informative Article on Grid computing. I came to know about the Computing technologies by the conference Cloudslam 2010.By Harinath, at 1:49 PM