Research Of Deadline-Constrained Scheduling Methods For Large File Transfers In Optical Networks

The high-speed development of distributed computing systems and grid systems needs lots of files with large scale data aggregating from local computers to remote data processing center for computing, storing or displaying. These large files are expected to be aggregated without data loss and reach their destination before given deadlines. All the files should be aggregated to the final data processing center before their deadlines. However, a few number of files transmitted over their deadlines are also allowed in high load conditions. But we want to decrease the difference between the real finish time and the deadline for the file which is finished over its deadline. This problem is named as the Deadline-constrained Large File Transfer Problem (DLFTP).In this paper, we introduce the recent research on large file transfers firstly, and then give the analysis and conclusions. Secondly, we give the solution of this problem in static and dynamic conditions.DLFTP is a problem combining routing and scheduling. We have proposed a new dynamic routing algorithm to solve the routing problem. Then a novel algorithm to combine dynamic routing and scheduling in optical networks has also been proposed, which has two mainly processes. One is dynamic routing process by which we can find the"best"route for the file transfer and dynamically switch light-paths between the primate routing path and the alternate routing path to make more files get light-paths with enough bandwidth. The other process is file scheduling process. In this process we schedule the files of the waiting queue according to the network traffic distribution to make more files transmission can be finished before their deadlines. We proposed many scheduling strategies in this paper.In order to compare the performance of our algorithm, a Java-based simulator is also proposed. By simulation, we can find that by using our dynamic routing we can get better results than traditional routing algorithms in case of using the same scheduling strategy. For scheduling strategies proposed in this paper, we can find that Earliest Deadline First Scheduled strategy performs best.