A new temporal locality-based workload prediction approach for SaaS services in a cloud environment

As the paradigm shift toward Software as a Service (SaaS) continues to gain the interest of companies and the scientific community, performances must be optimal. Indeed, cloud providers must provide an optimal quality of service (QoS) for their users in order to survive in such a competitive cloud market. Workload forecasting techniques have been proposed in order to improve capacity planning, ensure efficient management of resources and, hence, maintain SLA contracts with end users. In this context, we propose a new approach to predict the number of requests arriving at a SaaS service in order to prepare the virtualized resources necessary to respond to user requests. The method will be implemented in order to simultaneously achieve a twofold benefit: obtain precise forecast results while optimizing response time. In this regard, we have chosen to control the computation time by dynamizing the size of the sliding window associated to the recent history to be analyzed, since the larger the size of the entry in the prediction model, the more the algorithmic complexity increases. Then, the prediction will be established based on the temporal locality principle and the dynamic assignment of weights to different data points in recent history. Moreover, the proposed method can be extended to cover other uses in prediction. Experiments were carried out to assess the performance of the proposed method using two real workload traces and compared to state-of-the-art methods. The proposed method offers a compromise between the execution time and the accuracy of the prediction. (c) 2021 The Authors. Published by Elsevier B.V. on behalf of King Saud University. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).