Research Area:  Cloud Computing
Cloud infrastructure provides resources needed for tasks for resource scheduling. This work uses a genetic algorithm based on encoded chromosome (GEC-DRP) to manage dynamic resource scheduling. However, the existing scheduling algorithm estimates the number of required physical machines (PM) needed for the client in the future. This developed scheduling algorithm schedules the tasks on cloud by calculating the number of virtual machines needed in the near future along with their predicted CPU and memory requirements, which is the main contribution of the work. K-means algorithm clusters the tasks based on CPU and memory usage as parameters. The future arrival of tasks for every cluster is predicted and accordingly, the required number of VMs is created. The incoming requests known as tasks are scheduled on the appropriate VM using the genetic algorithm (GA). Based on the workload prediction results, a cost-optimized resource scheduling strategy in cloud computing environment is proposed aiming at minimizing the total cost of rental virtual machines from the central cloud. Finally, a genetic algorithm is used to solve the resource scheduling strategy. The developed algorithms are evaluated by the workload prediction accuracy, the total cost of the cluster and the algorithm’s consuming time for solving the resource scheduling problems through the experiments. Finally, the effective of workload prediction algorithm based on SES and cost-optimized resource scheduling strategy is verified by simulation.
Keywords:  
Cloud computing
Data clustering
Workload prediction
Task scheduling
Genetic algorithm
Author(s) Name:   K. Lalitha Devi & S. Valli
Journal name:  The Journal of Supercomputing
Conferrence name:  
Publisher name:  Springer
DOI:  10.1007/s11227-020-03606-2
Volume Information:  volume 77, pages 8252–8280 (2021)
Paper Link:   https://link.springer.com/article/10.1007/s11227-020-03606-2