AI-BASED OPTIMIZATION OF AUTO SCALING TECHNIQUES FOR LOAD BALANCING IN CLOUD COMPUTING ENVIRONMENTS

Abstract

An efficient way for task scheduling in cloud computing environment to allocates relevant Virtual Machines in Server, based on the demand is more important .Due to the complicated structure of jobs and resources in cloud data centers, this task scheduling problem is known to be NP-complete, making it a difficult issue to solve. Schedule length (or "make span") minimization is the primary focus of task scheduling. They can make better use of cloud resources and shorten the duration of jobs' execution, response, and waiting times if our lower the span of execution. Load balancing, in which incoming work is dispersed among available resources, comprises one of the primary uses of task scheduling. In order to solve the Unrelated Parallel Machine Scheduling Problem with sequence-dependent setup times, the Courtship Learning-Improved Firefly Algorithm has been presented. In order to enhance cooperation among the Fireflies and prevent them from settling for suboptimal solutions, this approach makes use of a Cauchy's value density value. The primary goal of Firefly is to dynamically distribute work among available machines in order to achieve maximum efficiency. The basic goal of Firefly is to dynamically balance workload across computers to enhance performance. A sequential UPMSP numerical solution is provided by objective fitness value using an upgraded Firefly algorithm with engagement learning. Dynamic Q-Learning aims to address the complexity and overhead of handling structured and unstructured data formats, the demand for power and energy efficiency in modern processor design, and the need for resource utilization and allocation in processing complex tasks using neural networks. Additionally, the focuses on task scheduling algorithms in the cloud to improve processor productivity and utilization while considering system bandwidth. Reinforcement learning can be applied in cloud

iv

load balancing situations to optimize the distribution of network traffic across servers in a cloud environment. By using reinforcement learning, load balancing algorithms can continuously learn and adapt to changing conditions. The reinforcement learning agent can perceive and interpret the current state of the system, take actions such as redirecting traffic to different servers, and receive feedback in the form of rewards or penalties based on the performance of those actions. Through trial and error, the reinforcement learning agent can learn the optimal load balancing strategy for a given environment. It can learn to adjust traffic distribution based on variables such as server capacity, network latency, and the current workload. This adaptive learning capability can lead to more efficient and effective load balancing, improving performance and resource utilization. Overall, applying reinforcement learning in cloud load balancing enables the development of intelligent and adaptive load balancing algorithms that can optimize the distribution of traffic and improve the performance of cloud-based applications.