Energy-latency trade-off analysis for scientific workflow in cloud environments: The role of processor utilization ratio and mean grey wolf optimizer

Cloud computing has demonstrated its effectiveness in handling complex data that requires substantial computational power, immediate responsiveness, and ample storage capacity. However, there are significant challenges when it comes to distributing intricate tasks, such as workflows structured in a Directed Acyclic Graph (DAG), across cloud resources. These challenges include increased energy consumption and delays experienced by end-users due to geographical disparities between users and service providers. To address these issues, this research work introduces an innovative solution that utilizes a multi-objective Grey Wolf Optimizer (GWO) algorithm for optimizing the allocation of complex workflows. The primary objective of this workflow optimization is to achieve the highest possible processor utilization ratio while adhering to latency constraints. This approach consists of a two-step process. In the initial step, it assesses processor utilization by calculating the number of completed transactions within a specified period divided by the average power consumption. This evaluation ensures that processors operate at the most power-efficient utilization levels, resulting in the highest service profit. In the second step, an optimal mapping algorithm assigns workflow tasks to the most efficient destinations. This is accomplished by applying the Mean Grey Wolf Optimization (GWO) algorithm, which is a metaheuristic technique. The goal is to find an optimal pairing of tasks and virtual machines (VMs) that balances multiple objectives. The simulation of this approach was conducted using the Java-based open -source CloudSim simulator version 5.0. The experimental results from a series of trials indicate that the ELSCiW approach leads to various improvements, including a reduction in latency ranging from approximately 5.35% to 12.92%, a decrease in end -to -end delay costs of about 5.03% to 13.80%, a reduction in energy consumption within the range of 4.71% to 11.19%, an enhancement in load balancing of around 14.84% to 37.30%, an improvement in performance degradation of about 12.52% to 35.14%, a decrease in overhead by approximately 17.27% to 26.92%, and an increase in the quality of service of roughly 56.21% to 64.85%. Furthermore, the ELSCiW approach outperforms recent state -of -the -art metaheuristic methods, as confirmed by statistical findings that emphasize the robustness of the proposed algorithm.