Data-Oriented Scheduling with Dynamic-Clustering Fault-Tolerant Technique for Scientific Workflows in Clouds