A Comparative Study of Task and Fault Tolerance Clustering Techniques for Scientific Workflow Applications in Cloud Platform

Scientific workflows are characterized by the existence of complex tasks, which involve multifaceted mathematical and scientific calculations and multiple dependencies. These dependencies can be modeled as Directed Acyclic Graphs (DAG). System overheads that are present in these workflows result in the longer execution time of the task that has otherwise a very short runtime. Hence clustering of the tasks is performed where smaller tasks are combined to form a job, thereby reducing the system overhead and increasing the runtime performance of the tasks. This survey paper discusses the clustering mechanisms that are suitable for the scientific workflow along with the fault tolerance mechanisms that help to make the system robust. The analysis of the performance of various clustering algorithms is also discussed in this paper.