Fault-tolerant scheduling for scientific workflow with task replication method in cloud

Cloud computing has become a revolutionary paradigm by provisioning on-demand and low cost computing resources for customers. As a result, scientific workflow, which is the big data application, is increasingly prone to adopt cloud computing resources. However, internal failure (host fault) is inevitable in such large distributed computing environment. It is also well studied that cloud data center will experience malicious attacks frequently. Hence, external failure (failure by malicious attack) should also be considered when executing scientific workflows in cloud. In this paper, a fault-tolerant scheduling (FTS) algorithm is proposed for scientific workflow in cloud computing environment, the aim of which is to minimize the workflow cost with the deadline constraint even in the presence of internal and external failures. The FTS algorithm, based on tasks replication method, is one of the widely used fault tolerant mechanisms. The experimental results in terms of real-world scientific workflow applications demonstrate the effectiveness and practicality of our proposed algorithm.