Energy and Performance Centric Resource Allocation Framework in Virtual Machine Consolidation Using Reinforcement Learning Approach

Title:Energy and Performance Centric Resource Allocation Framework in Virtual Machine Consolidation Using Reinforcement Learning Approach

Volume: 17

Author(s): Madala Guru Brahmam and Vijay Anand R*

Affiliation:

• School of Computer Science Engineering & Information Systems, Vellore Institute of Technology, Vellore, India

Abstract: Virtual machines are used to reduce cloud platform application performance, management costs, and access irregularities. Virtual machines are frequently vulnerable to delays, overburdening workloads, and other obstacles while consolidating and migrating servers. To significantly disperse loads among virtual machines, dynamic consolidation techniques are implemented to control energy dissipation, monitor overloading, and address underloading problems.

Background: The process of consolidation involves more calculations and resources in order to transfer services between virtual machines, provided that Service Level Agreements are observed.

Methods: The suggested approach promotes the use of cutting-edge architecture to combine virtual machines, and, therefore, strike a balance between performance and energy requirements. The main design considerations for the suggested Dynamic Weightage algorithm, which includes the clustering approach in relation to reinforcement learning approaches, are overall resource needs and Performance to Power Ratio (PPR). A cluster of ideal virtual machines is created, and resources are distributed according to performance and energy requirements. Virtual machine resource requests are converted into a matching relationship factor, which represents the individual hosts while taking PPR into account. The overall workload associated with virtual machine consolidation is also provided by these estimations. It is noted that there is little energy trade-off and that performance is maintained at a nominal level across the cluster. The architecture is put into practice throughout offline platforms, which are dispersed ecosystems that allow for increased system performance and scaling.

Results: The CloudSim simulator is used to validate the system using datasets that are obtained from PlanetLab. According to the data, energy saving has produced yields of up to 47% and promising quality of service attributes.

Conclusion: The validation of the system is performed using the CloudSim simulator with datasets from PlanetLab. The results indicate significant energy conservation, up to 47%, along with promising quality of service parameters. The proposed architecture is compared with other state-of-the-art algorithms for distributed architectures and heterogeneous environments, showcasing its efficiency. The conclusion emphasizes the prioritization of VM consolidation and energy efficiency in the proposed architecture, which has been tested on a Proliant G7- based data center using a variety of hosts. Notably, the CloudSim Toolkit is highlighted as outperforming OpenStack-based techniques in simulation results.

Cite this article as:

Brahmam Guru Madala and R Anand Vijay*, Energy and Performance Centric Resource Allocation Framework in Virtual Machine Consolidation Using Reinforcement Learning Approach, Recent Advances in Computer Science and Communications 2024; 17 () : e150224227004 . https://dx.doi.org/10.2174/0126662558289911240206071447