Evaluation of Improvement Algorithms for Dynamic Co-allocation with Respect to Parallel Downloading in Grid Computing

Grid technology is the next evolutionary level in the field of distributed systems. This technology issued to create the illusion of a super computer by connecting a number of heterogeneous systems to share various kinds of resources and to solve large scale problems. Grid technology, with SOA architecture has lead to the virtualization of almost every real life activity. Data Grid is one such, which is used to solve data intensive applications by sharing tera and peta bytes of information. Since a single computer cannot handle such huge amount of data, replication and fragmentation is used. It improves the data retrieval performance in a Grid environment by increasing the data availability. But, for parallel downloading, replication alone will not suffice. Downloading a data file from multiple locations in parallel also require multiple data streams between the clients of an enterprise and the Internet. The Dataset is then divided into a number of disjoint blocks of equal size and threads are created, with each thread downloading a single block of data. Next, these threads are executed in parallel across a number of grid nodes on the client side. Algorithms for scheduling this task must ensure the completion of download process in addition to improving the speed. This research paper is based on the practical outcomes of the implementation of a parallel downloading process in a Grid setup with Alchemi. Net as the Grid machinery and runtime environment for windows based systems.