Data and Task Scheduling in Distributed Computing Environments

摘要

Data-aware scheduling in today's large-scale heterogeneous environments has become a major research and engineering issue. Data Grids (DGs), Data Clouds (DCs) and Data Centers are designed for supporting the processing and analysis of massive data, which can be generated by distributed users, devices and computing centers. Data scheduling must be considered jointly with the application scheduling process. It generates a wide family of global optimization problems with the new scheduling criteria including data transmission time, data access and processing times, reliability of the data servers, security in the data processing and data access processes. In this paper, a new version of the Expected Time to Compute Matrix (ETC Matrix) model is defined for independent batch scheduling in physical network in DG and DC environments. In this model, the completion times of the computing nodes are estimated based on the standard ETC Matrix and data transmission times. The proposed model has been empirically evaluated on the static grid scheduling benchmark by using the simple genetic-based schedulers. A simple comparison of the achieved results for two basic scheduling metrics, namely makespan and average flowtime, with the results generated in the case of ignoring the data scheduling phase show the significant impact of the data processing model on the schedule execution times.