Multi‐criteria workflow scheduling on clouds under deadline and budget constraints

Abstract

Cloud Computing is increasingly favored by scientists for executing large-scale workflow applications due to its appealing features such as availability of huge infrastructure, low operating costs, easy accessibility, and high scalability. Moreover, the resources in cloud are provisioned as per user's requirement following pay-per-use model. Scheduling workflows in cloud requires orchestration of tasks onto heterogeneous resources while complying with the task dependencies as well as user's Quality of Service (QoS) demands. The problem becomes challenging if there are multiple QoS requirements which conflict with each other. Recent studies attempted to minimize the makespan and execution cost under deadline and/or budget constraints. However, the failure of resources in cloud poses a serious threat to successful execution of these applications. Hence, it becomes essential to consider reliability of resources in addition to other parameters. This paper presents a workflow scheduling policy which minimizes makespan and execution cost while maximizing the reliability of executing workflows under user specified deadline and budget restraints. We have devised a hybrid of Intelligent Water Drops algorithm and Genetic Algorithm (IWD-GA) to accomplish the desired goal. It provides a set of non-dominated solutions with different makespan, cost, and reliability to offer users more flexibility in selecting a solution as per their preferences. The evaluation of the proposed algorithm has been carried based on two performance metrics, ie, hypervolume and two set coverage. The simulation results thus obtained using four diverse scientific workflows confirm that IWD-GA outperforms non-dominated sort genetic algorithm (NSGA-II) and hybrid particle swarm optimization (HPSO) with regard to accuracy and diversity.

Hybrid scheduling algorithms in cloud computing: a review

“…The most common are the budget and deadline constraints. In budget constraints, the user's tasks need to be complete in some predefined threshold of the cost, whereas in deadline constraint, the makespan value is considered [12]. Mathematically, the scheduling problem with constraints can be described using (1) and (2):…”

“…The CyberShake workflow analyzes the disaster modeling and prediction of the particular geographical location by designing the hazard maps [15]. For studying and analyzing the gravitational waveform, the inspiral workflow is used [12]. The montage workflow is used by NASA in astrophysics to create the custom mosaics of the sky by taking multiple input images [16].…”

“…The throughput is calculated as the number of task instructions completed in a unit time [20]. The throughput of 𝑉𝑀 𝑘 is calculated as per (12),…”

A variant of teaching‐learning‐based optimization and its application for minimizing the cost of Workflow Execution in the Cloud Computing

“…Karla and Singh 77 tried to minimize the makespan and cost, including maximization of reliability of workflow execution under deadline and budget constraint specified by the user. The author applied the intelligent water drops and genetic algorithm.…”

Workflow scheduling using particle swarm optimization and gray wolf optimization algorithm in cloud computing

“…The CyberShake workflow is used to analyze the disaster modeling and prediction of the particular geographical location by designing the hazard maps 3 . For studying and analyzing the gravitational waveform, the Inspiral workflow is used 4 . NASA uses the montage workflow in astrophysics to create the sky's custom mosaics by taking multiple input images 5 .…”