A model for enhancing cloud computing resource allocation management using data analytics

Abstract

Traditional information and communication technology (ICT) systems are currently challenged by the huge and growing demand for digital and data services. This has led to the adoption of cloud computing infrastructure systems using the internet and mobile applications to replace most of the in-house data warehouses. Cloud computing provides improved and simplified capabilities to manage and maintain ICT resources. Unfortunately, even the adopted cloud computing systems are faced with rising demand for data services, which causes data traffic spikes and challenges to cloud computing service resource allocation. Ineffective resource allocation in cloud computing can lead to higher expenses, underuse, decreased performance, and scalability issues. Causes include inadequate capacity planning, overprovisioning, under-provisioning, and lack of elasticity. One of the leading South African automotive companies, Company X, has been experiencing a growing demand for data from cloud computing servers, manifesting as issues with resource allocation on its cloud computing servers daily, for instance. These difficulties are currently resolved manually, and it is hard to forecast when or how they may arise. Investigation into Company X's difficulties showed that cloud network computing resources were also in demand throughout the impacted period. Furthermore, the unpredictability of network traffic flow in cloud computing exacerbated the situation. All these challenges led to dissatisfaction among Company X's clients. Customers began to complain about high bandwidth utilisation, application timeouts and delayed system times because of inconsistent traffic flow. One way of resolving cloud computing challenges and reducing inefficiencies in resource allocation to customers is to use a trustworthy technique for predicting cloud computing traffic. The purpose of this article-based thesis was to use data analytics techniques to propose a model that could forecast and predict the availability of cloud computing services and resource allocation that would satisfy customers. This thesis used pragmatic data analytic techniques and data provided by Company X for the purpose of investigating potential solutions, as well as extra data downloaded from online sources. In this regard, the following research techniques were used: bibliometric analysis, systematic literature review (SLR) and ensemble models. The analysis resulted in the compilation of eighteen (18) research articles. five (5) articles were bibliographic reviews for the analysis of cloud computing; four (4) were SLR articles; two (2) articles covered individual techniques for cloud computing resource allocation using data analytics; Six (6) articles dealt with development, proposal, and recommendation of an ensemble method, namely, Stepwise Gaussian Linear Autoregressive (SGLA), and One (1) papers addressed the validation of the SGLA technique. The SGLA is a combination of the linear regression, support vector machines, Gaussian process regression and the autoregressive integrated moving average technique. The SGLA approach outperformed individual models in terms of resource allocation and prediction accuracy; hence it was recommended as an ensemble model for enhancing cloud computing resource allocation management. By using the averaging strategy, SGLA shows a clear advantage in handling resource allocation better despite traffic fluctuations. Overall experimental results indicate that this method performed better than single models in terms of prediction accuracy. After the application of SGLA, the manual intervention is reduced, application timeout is eliminated, and the overall user experience improves. The SGLA performs better than the average model, with over 90% accuracy. Since obtaining 100% correct predictions and allocation of resources is of major concern owing to the constant changes in technology and the rise in traffic, by adopting SGLA, organisations can prepare for uncertainty and take advantage of changing technology for traffic prediction in cloud computing systems.