Energy-efficient resource provisioning in cloud data centers
In recent years, energy efficiency has become a challenging issue in large data centers. Energy-efficient resource provisioning manages the usage of each computing resource and shares virtual machines (VMs) so that energy consumption can be reduced. Thus, to maintain energy consumption, the proposed...
Saved in:
Published in | Sustainable computing informatics and systems Vol. 47; p. 101167 |
---|---|
Main Authors | , |
Format | Journal Article |
Language | English |
Published |
Elsevier Inc
01.09.2025
|
Subjects | |
Online Access | Get full text |
Cover
Loading…
Summary: | In recent years, energy efficiency has become a challenging issue in large data centers. Energy-efficient resource provisioning manages the usage of each computing resource and shares virtual machines (VMs) so that energy consumption can be reduced. Thus, to maintain energy consumption, the proposed study aims to develop an effective resource provisioning mechanism (Eff-RPM) in cloud data centers. The proposed framework contains the following stages: workload pre-processing, workload clustering, and optimized resource provisioning. Initially, the incoming workloads are pre-processed to remove the noisy and invalid requests. The pre-processing stage allocates a specific ID number for each request, which helps to make the Service Level Agreement (SLA) table for each workload. Then, the pre-processed workloads are clustered through the Enhanced Genetic K-means (EGKM) algorithm. Meanwhile, the enhanced genetic algorithm is hybridized with K-means clustering to categorize the workloads in clouds depending on the quality of service (QoS) requirements. The proposed clustering stage helps to reduce the makespan and computation time. Finally, the new Hybrid Optimal Convolutional Neural Network (HOCNN) model is proposed for provisioning suitable resources to the workloads. Here, the workloads are predicted using a Convolutional Neural Network (CNN), and the Cuckoo Search Optimization (CSO) algorithm is employed for optimally provisioning the resources by regulating parameters like response time, makespan, resource utility, and energy consumption. The proposed Eff-RPM is implemented in the CloudSim platform, and the performances are evaluated against various evaluation criteria and compared to existing methods. On the evaluations, the proposed approach resulted in an overall energy consumption of 76.504 kWh, total cost of 236,001 C$, SLA violation rate of 6.541 %, delay of 6.197 s, resource utilization of 88.922 %, response time of 278.84 s, and makespan of 111.531 s, respectively. As a result, the proposed Eff-RPM has achieved superior performance for resource provisioning against the existing methods.
•To develop an optimized deep learning model for generating an energy-efficient resource provisioning in cloud environment.•To perform an effective pre-processing stage to eliminate the available noises and unwanted requests.•To minimize the makespan and computational time, a robust clustering process is performed.•Proposing a CNN-based deep learning model for predicting future requirements of CPU utilization and long-term workloads.•To introduce the CSO approach for optimally allocating the resources by satisfying four different parameters. |
---|---|
ISSN: | 2210-5379 |
DOI: | 10.1016/j.suscom.2025.101167 |