AI-Driven Resource Provisioning: Enhancing Elasticity and Efficiency with Hybrid RNN-LSTM Models

Abstract

Automatic resource provisioning techniques that dynamically modify resources in response to service demand are essential to implementing elasticity services. For systems with strict latency or reaction time requirements, such as Enterprise Resource Planning (ERP) systems under high traffic loads, this flexibility is crucial for lowering power consumption and guaranteeing quality of service (QoS). Determining the best point at which to scale resources is still a difficult task. In this research, we provide an AI-integrated system that adjusts resources according to anticipated demand. To predict load requests, the system uses an LSTM model and a Hybrid Recurrent Neural Network (RNN). This strategy seeks to minimize overprovisioning, which will lower the cost of infrastructure and energy usage. The deep learning model aims to estimate the resources required to improve service response time and satisfy customer requests, as well as to anticipate with high accuracy the processing load of distributed servers. Proactive provisioning decisions are made for the servers based on the anticipated load. Our tests on a common server request dataset show that the suggested RNN+LSTM model performs better than traditional deep learning models in terms of efficient resource management and prediction accuracy.