Energy-aware dynamical hosts and tasks assignment for cloud computing

• Published in: The Journal of systems and software 2016-05, Vol.115, p.144-156
• Main Author: Wen, Yean-Fu
• Format: Article
• Language: English
• Subjects: Cloud computing; Computer programming; Dynamical systems; Dynamics; Energy consumption; Energy efficiency; Energy management; Load balance; Loads (forces); Mathematical models; Schedules; Scheduling; Studies; Tasks; Threshold
• ISSN: 0164-1212; 1873-1228
• DOI: 10.1016/j.jss.2016.01.032

•Explore the network, task, routing, power consumption, and response time models.•Propose a pair of power-on and suspend thresholds to achieve energy efficiency.•Propose a dynamical active host and task assignment scheme.•Balance the process and transmission power consumption to avoid ping-pong problem.•Evaluate the proposed schemes from several aspects of controlled variables. One feature of MapReduce is to split user request into multiple tasks and then process around multiple datacenters for cloud computing. This study addresses an energy efficiency problem of dynamic cloud hosts (CHs) and task assignments as well as a subset of CH power-on or suspended schedules by controlling the range between the power-on and suspended thresholds for high-energy efficiency. A dynamical CHs and tasks assignment scheme is proposed to reduce the overall system energy consumption. The main concept of the proposed scheme entails setting the thresholds to satisfy the constant and variable traffic loads, nodal load balance, migration overhead, basic required power, and processing power. The reason is the established energy consumption required for initialing power-on and variable rates to keep working. This work evaluates the proposed scheme and compares it with the CHs and tasks assignment schemes to show how the proposed scheme achieves energy efficiency. The simulation results show that the proposed scheme obtains the lowest energy consumption under the tolerable responding time constraints even though the request traffic load is varying. The average improvement rate is 16.3% to balance the number of active hosts and migration overhead as well as 4.8% for task schedule.