Optimizing Task Scheduling and Resource Utilization in Cloud Environment: A Novel Approach Combining Pattern Search With Artificial Rabbit Optimization

The increasing demand for Cloud service with sudden resource requirements of Virtual Machines (VMs) with different resource types and sizes may create an unbalanced state in the Cloud datacenters. In turn, it will lead to low resource utilization and slow down the server’s performance. This research article proposes an enhanced version of the Artificial Rabbit Optimization (ARO) called Improved Artificial Rabbit Optimization based on Pattern Search (IARO-PS), where ARO has been utilized to schedule the dynamically independent requests (tasks) for overcoming the challenges discussed above and a Pattern Search (PS) method has been hybridized to address the shortcomings of ARO and to provide better exploration-exploitation balance. The initial step in the proposed approach is to employ a load balancing strategy by dividing the workloads (user requests) across the available VMs. The next step utilizes the IARO-PS method to map the workloads (user requests) onto the optimal VMs for the scheduling process to carry out across the diverse resources. A standard benchmark function (CEC2017) is used to assess the IARO-PS technique’s efficacy. A comprehensive evaluation has been carried out by taking an available real-world dataset having different specifications of tasks in the CloudSim to evaluate the performance of the methodology. Additionally, a simulation-based comparison is carried out with various metaheuristic-based workload scheduling methods like Genetic Algorithm (GA), Bird Swarm Optimization (BSO), Modified Particle Swarm Optimization based on Q-learning (QMPSO), and Multi-Objectives Grey Wolf Optimizer (MGWO). Based on the simulations, the IARO-PS algorithm performed better than the previously mentioned algorithms, reducing makespan by 10.45% (GA), 2.31% (QMPSO), 4.35% (MGWO), 15.35% (BSO), and 4.17% (GA), 1.03% (QMPSO), 1.44% (MGWO), 7.33% (BSO), in both homogeneous and heterogeneous surroundings, respectively, and improving resource utilization by 36.74% (GA), 14.31% (QMPSO), 19.75% (MGWO), 45.23% (BSO) and 12.17% (GA), 6.02% (QMPSO), 9.10% (MGWO), 19.39% (BSO). Furthermore, statistical evaluation through Friedman’s test and Holm’s test has also been carried out showcasing the decrease in makespan and an increase in VM utilization, which are the outcomes of the simulated experimental study.