Machine learning techniques analysis to Efficient resource provisioning for elastic cloud services

By automatically allocating resources in response to changes in service demand, automated resource provisioning solutions enable elastic services. In order to decrease power consumption without sacrificing quality of service or service level agreements, these standards must be followed by services with strict latency or reaction time QoS requirements, such as continuously streaming data, heavily used web servers, or real-time big data analytics. Cloud computing and virtualised data centres frequently employ auto-scaling technologies to incorporate elasticity. These automatically determine resource allocation based on the significance of specific indicators monitoring the performance of infrastructure and/or services. By combining machine learning methods from queueing theory and time series forecasting, this study introduces and assesses a new predictive auto-scaling mechanism. The goal of the new mechanism is to optimise service response time, service level agreement fulfilment, energy consumption, and infrastructure costs by properly predicting the processing load of a distributed server and estimating the amount of resources that should be deployed. The suggested model outperforms competing classical models in terms of resource allocation and forecasting accuracy, according to the results.