Energy-efficient resource allocation scheme based on enhanced flower pollination algorithm for cloud computing data center

Abstract

Cloud Computing (CC) has rapidly emerged as a successful paradigm for providing ICT infrastructure. Efficient and environmental-friendly resource allocation mechanisms, responsible for allocatinpg Cloud data center resources to execute user applications in the form of requests are undoubtedly required. One of the promising Nature-Inspired techniques for addressing virtualization, consolidation and energyaware problems is the Flower Pollination Algorithm (FPA). However, FPA suffers from entrapment and its static control parameters cannot maintain a balance between local and global search which could also lead to high energy consumption and inadequate resource utilization. This research developed an enhanced FPA-based energy efficient resource allocation scheme for Cloud data center which provides efficient resource utilization and energy efficiency with less probable Service Level Agreement (SLA) violations. Firstly, an Enhanced Flower Pollination Algorithm for Energy-Efficient Virtual Machine Placement (EFPA-EEVMP) was developed. In this algorithm, a Dynamic Switching Probability (DSP) strategy was adopted to balance the local and global search space in FPA used to minimize the energy consumption and maximize resource utilization. Secondly, Multi-Objective Hybrid Flower Pollination Resource Consolidation (MOH-FPRC) algorithm was developed. In this algorithm, Local Neighborhood Search (LNS) and Pareto optimisation strategies were combined with Clustering algorithm to avoid local trapping and address Cloud service providers conflicting objectives such as energy consumption and SLA violation. Lastly, Energy-Aware Multi-Cloud Flower Pollination Optimization (EAM-FPO) scheme was developed for distributed Multi-Cloud data center environment. In this scheme, Power Usage Effectiveness (PUE) and migration controller were utilised to obtain the optimal solution in a larger search space of the CC environment. The scheme was tested on MultiRecCloudSim simulator. Results of the simulation were compared with OEMACS, ACS-VMC, and EA-DP. The scheme produced outstanding performance improvement rate on the data center energy consumption by 20.5%, resource utilization by 23.9%, and SLA violation by 13.5%. The combined algorithms have reduced entrapment and maintaned balance between local and global search. Therefore, based on the findings the developed scheme has proven to be efficient in minimizing energy consumption while at the same time improving the data center resource allocation with minimum SLA violation.