Design And Development Of A Distributed Computing Platform For The Mesh Network Connectivity Problem

We often come across some intricate problems in our researches, such as Mesh network connectivity problem in network theory. These problems are hard to be run out on a single PC in a shorter time. But on the other hand, there are hundreds of millions of idle PCs underused on the Internet. We can fully harness these computational resources by distributed computing.Based on this idea, we design and implement a distributed computing platform, and apply it to the Mesh network connectivity problem. The platform can decompose a large task (such as the Mesh network connectivity problem) to small tasks, which would be allocated to client machines and run during the idle CPU time of clients; when the subtask is completed, the client will return the result to server, where the server program will gather the results from each client and produce the final result. We bring the conception of task tree in the task decomposition process, and build the task tree by trace-back algorithm in the database. So the entire structure of the platform is typical three-tier Client/Server model, including a Server program, a Client program, and a back database. Now, the platform has been implemented under Windows. The Server and Client program was developed by Visual C++.NET, employing CS-SDK of Mithral as the communication API between the Client and Server; and the back database was designed by Oracle8.1.6.This platform has been successfully applied to solve the Mesh network connectivity problem, and will be used in many other distributed computing problems in the future.Finally, the paper sums up our work during the design and development, and discusses how to improve the platform in the future.