Scheduling Scientific Workflows to Meet Soft Deadlines in the Absence of Failure Models

Abstract

Highly distributed systems such as Clouds and Grids are used to execute complex scientific workflow applications by researchers from various areas of science. While scientists rightfully expect efficient and reliable execution of their applications, current systems often cannot deliver the required Quality of Service. We propose a dynamic execution and scheduling heuristic able to schedule workflow applications with a high degree of fault tolerance, while taking into account soft deadlines. Experimental results show that our method meets soft deadlines in volatile highly distributed systems in the absence of historic failure trace data or complex failure models of the target system.

A GPU-based Parallel Ant Colony Algorithm for Scientific Workflow Scheduling

“…Kernel function global_update updates the global pheromone according to formula (10) and (11). This function is executed only once during one iteration and updates pheromone with the optimal solution.…”

“…Benedict et al solved the scheduling of scientific workflows using Niched Pareto Genetic Algorithm (NPGA) [10]. Jia Yu et al presented a genetic algorithm approach to address scheduling optimization problems in workflow applications, based on two QoS constraints, deadline and budget [11][12] scientific workflows for the grid [13]. Plankensteiner et al proposed a dynamic execution and scheduling heuristic capable of scheduling workflow applications with a high degree of fault tolerance [14].…”

Admission and routing of soft real-time jobs to multiclusters: Design and comparison of index policies

“…The prime motivation for the study of such a system model comes from certain time-sensitive e-service applications, where it is desirable but not mandatory to meet given transaction deadlines, and which increasingly rely on heterogeneous multicluster web-server architectures. Examples of such soft real-time applications include e-commerce (see Bertini et al (2010)), real-time databases, e.g., for online stock trading (see Kao and García-Molina (1996) and Kang et al (2007)), and scientific computing (see Zhu (2001) and Plankensteiner et al (2010)). See also He et al (2006), which considers a similar model to the present one, yet focusing on the pure-routing case, i.e., without admission control.…”