Turnaround Time Minimization-Based Static Scheduling Model Using Task Duplication for Fine-Grained Parallel Applications onto Hybrid Cloud Environment

ABSTRACT

Cloud computing is a multifarious computing paradigm incorporating the benefits of service-oriented architecture and utility computing through virtualization. Hybrid cloud, an amalgamation of two or more public and/or private clouds, is gaining high popularity between users due to various reasons involving improved performance, flexible business operations, capacity expansion, optimized costs, and enhanced security. The efficient execution of fine-grained parallel applications onto hybrid cloud system becomes limited due to a number of factors. From the application point of view, it ranges from the dynamicity of the applications to their precedence and communication constraints while for the computational resources, it includes heterogeneity of processors and participating clouds with their interconnection topology. This work proposes a compile time hybrid cloud-based task duplication strategy to execute the fine-grained applications represented as directed acyclic graph (DAG) onto the hybrid cloud environment. The proposed strategy schedules the tasks based on a degree relative to the critical path in the DAG and tries to achieve lower bound of the DAG. Furthermore, it makes an effort to avoid redundant duplication by duplicating only the required parent tasks considering the available idle slots to minimize the execution time of the application. The experimental study reveals that the proposed strategy performs better than its peers in terms of achieving the lower bound more efficiently with lesser degree of duplication for fine-grained jobs. The strategy is highly useful for cloud environment as it results in lower cost of usage of resources with enhanced system utilization.

Keywords:

• Directed acyclic graph (DAG)
• hybrid cloud computing
• scheduling
• task duplication
• turnaround time (TAT)

Acknowledgements

The authors would like to acknowledge UPE-II, Jawaharlal Nehru University for the support.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Notes on contributors

Mohammad Sajid

Mohammad Sajid received his MTech in 2012 after completion of MCA degree in 2010 from Jawaharlal Nehru University, India. He is currently pursuing PhD from the School of Computer and Systems Sciences, Jawaharlal Nehru University, India. His research interests include high performance computing, scheduling/stochastic scheduling, evolutionary algorithms and multi-objective evolutionary algorithms.

E-mail: [email protected]

Zahid Raza

Zahid Raza is a faculty in the School of Computer and Systems Sciences, Jawaharlal Nehru University, India. He is MSc in electronics and MTech in computer science. He did his PhD in computer science from Jawaharlal Nehru University, India. Prior to joining JNU, he served as a lecturer in Banasthali Vidyapith University, Rajasthan, India. His research interests include parallel and distributed systems, evolutionary algorithms and multi-objective evolutionary algorithms. He has published many peer-reviewed articles and has proposed various scheduling models for computational grid, cloud and cluster systems.

E-mail: [email protected]