بهبود تعادل بار با زمان‌بندي وظايف بر پايۀ هوش مصنوعي در رايانش ابري نرم‌افزار محور

In this research, we aim to improve the efficiency of cloud computing systems by addressing the challenges of task scheduling in these environments. Given the importance of Software-Defined Networking (SDN) in the dynamic an‎d flexible management of network resources, we propose a novel task allocation algorithm based on Q-learning reinforcement learning. The primary objective of this algorithm is to enhance load balancing, reduce response time, an‎d guarantee that tasks meet their deadlines. The proposed algorithm has a three-module architecture, including network monitoring, task prioritization, an‎d decision-making. The network monitoring module collects necessary information from sending devices. The task prioritization module ranks tasks based on their deadlines. Finally, the decision-making module, using the proposed algorithm, determines the best policy for allocating each task to a specific server. The proposed algorithm interacts with the environment (server system an‎d tasks) an‎d receives rewards, gradually learning which decision is best for task allocation. To eva‎luate the performance of the proposed algorithm, a simulation environment has been created using Mininet an‎d the Ryu controller. This simulation environment enables the creation an‎d configuration of virtual software-defined networks, simulating conditions similar to real-world networks. Simulation results show that the proposed algorithm outperforms other methods in terms of load balancing an‎d reducing latency. Additionally, the algorithm can maintain optimal performance under various workload conditions, an‎d most tasks are completed within their specified deadlines. The use of Mininet as a software-defined network simulator enables the creation an‎d configuration of virtual networks, an‎d the Ryu controller, with its modular design, is a suitable choice for managing an‎d controlling these virtual networks. Using these tools, the proposed algorithm can be eva‎luated accurately an‎d controllably. Finally, this research, by providing an efficient proposed algorithm an‎d an accurate simulation environment, offers a comprehensive solution for optimal task allocation in cloud computing environments. This algorithm, with its distributed, adaptable, an‎d high-performance capabilities, can improve the performance of large-scale cloud systems an‎d better manage resources. Overall, this research is a significant step toward improving the efficiency an‎d productivity of cloud computing systems an‎d can serve as a foundation for future research in this field.

5