This is targeted at professionals more than it is academics. The application of the technology that surrounds this article mostly seems applicable in a infrastructure business setting.
Type of Document
This document is a Journal article. It explores multiple ways on paralellizing legacy code and the pros and cons of each way. It is somewhat an instructional article and also an educational article.
Summary of Article
Companies with legacy code would like to find a way to easily parallelize that code, but it is not such an easy task. In many cases, the most robust way is to rewrite an architecture to interface over the legacy code and make it thread safe from the bottom up. But this of course requires a reconstruction of the system. Another approach is to distribute the most computationally intensive parts of the legacy code. The article goes into some suggested hardware specs for distributing. This article then summarizes the Fine Grained Distributed Processing approach for parallelizing legacy code.
The Fine Grained Distributed Processing approach is a good approach if you want an easy way to parallelize legacy code. However, there are certain constraints to which this approach works. These constraints are the message length and how easy it is to create a message. But if there is relatively simple message passing in legacy code then The Fine Grained Distributed Processing approach would be a good choice for doing so.
[ add comment ] ( 22 views ) | permalink