International Conference on Computer and Knowledge Engineering

Home / 14th International Conference on Computer and Knowledge Engineering

A Cloud Broker with Gap Analysis Perspective for Scheduling Multi-Workflows Across On-Demand and Reserved Resources

Authors :

Negin Shafinezhad¹ Hamidreza Abrishami² Saeid Abrishami³

1- Department of Computer Engineering Ferdowsi University of Mashhad Mashhad, Iran 2- Department of Computer Engineering Ferdowsi University of Mashhad Mashhad, Iran 3- Department of Computer Engineering Ferdowsi University of Mashhad Mashhad, Iran

Keywords :

Multi-workflow hybrid scheduling،workflow as a service broker،gap analysis،reserved resources،deadline constraint

Abstract :

The rapid growth in cloud services has led to the accessibility of Workflow-as-a-Service (WaaS) as the compatible platform for executing multiple workflows. Users desire to execute independent workflows with separate defined constraints in a shorter time and in a cost-efficient manner, while cloud service providers want to increase resource utilization during scheduling phases. Controlling gap spaces can help to achieve these aims, as gaps occur as idle time slots between prior scheduled tasks and the current task on paid resources, creating opportunities for executing more tasks. In this paper, we propose a WaaS broker with a gap analysis viewpoint as a third-party entity to mediate between users and service providers. This broker utilizes a hybrid scheduling method with static planning and dynamic scheduling for task assignments on reserved and on-demand resources under specified deadlines. It aims to reduce the execution cost by maximizing the utilization of reserved instances, which have lower costs. The static planner creates a cost-efficient primary scheduled map with minimum gap spaces, using a defined gap threshold parameter to control task assignment and estimate the number of on-demand resources required. The dynamic scheduler then uses the static scheduled map and considers uncertainties in task execution to perform real-time resource allocation and acquire new on-demand instances as needed. To evaluate the performance, we simulate numerous scenarios of scientific multi-workflows with different deadline constraints and workflow structures. The results show that the proposed broker can execute the maximum number of multi-workflows at minimum cost compared to state-of-the-art approaches, while maintaining high resource utilization.

Papers List

List of archived papers

Improving Machine Learning Classification of Heart Disease Using the Graph-Based Techniques

Abolfazl Dibaji - Sadegh Sulaimany

Using Deep Learning for Classification of Lung Cancer on CT Images in Ardabil Province

Mohammad Ali Javadzadeh Barzaki - Jafar Abdollahi - Mohammad Negaresh - Maryam Salimi - Hadi Zolfeghari - Mohsen Mohammadi - Asma Salmani - Rona Jannati - Firouz Amani