Resource Management in a Cloud Computing Environment using Generative Adversarial Networks (GANs)

Automating any resource allocation technique allows implementation of elastic cloud services by providing available cloud resources on-demand. This is important for decreasing energy consumption and meeting the SLAs and QoS, mostly for services that the QoS is dependent on latency and response time (Web servers, Big data analytics in real time). Elasticity is one of the key feature of the cloud computing platform, where the resources can be managed based on the user demand. Because the user requests/demand varies and fluctuates in a cloud environment, there is the issue of managing resources. This paper presents and evaluates a Generative Adversarial Network (GANs) based predictive resource allocation mechanism based on machine learning techniques. The GAN based mechanism aims at accurately predicting the load of any server and adequately provision the optimal number of resources required to optimise the response time and satisfy the SLA while also mitigating the issue of resource over-provisioning. Two other techniques (Rule-based and Linear Regression Model) are also implemented in this paper,the results show that the proposed model has the capability of achieving an optimal forecasting accuracy than the other two techniques implemented. The results are evaluated using the following metrics; latency, scaling response, idle cycles and task processed by the cloud system