Workload Prediction and Resource Management for Energy Efficiency in Cloud Data Centres

Given the increasing deployments of Cloud datacentres and the excessive usage of server resources, their associated energy and environmental implications are also increasing at an alarming rate. This necessitates the need for promoting sustainable computing approaches for achieving environment-friendly and energy-efficient datacentre execution. One of the direct causes of excess energy consumption of the datacentre includes running servers at a low level of workloads and over-provisioning of server resources to the arriving workloads during execution. Another cause of excess energy consumption of the datacentre is the lack of an efficient energy-aware workloads scheduling paradigm that is able to allocate optimal execution resources to various workloads.

With this in mind, a novel energy-efficient Cloud resource management approach is proposed in this thesis, which is enabled by highly reliable workload prediction and scheduling models. Firstly, a classification-aware workloads forecasting model named K-RVLBPNN has been developed for predicting the future workload arrival trend, by exploiting the latency sensitivity characteristics of Cloud workloads, based on a combination of improved K-means and BPNN algorithms. This prediction model achieves reliable prediction accuracy at a high level. Furthermore, it is found stragglers within Cloud workload executions are obvious energy consumers and will heavily influence prediction accuracy. To improve the prediction accuracy, the IGRU-SD model is presented in this thesis, which is based on big data analytics and Recurrent Neural Networks. The proposed model exploits an improved GRU neural network integrated with a resource straggler detection module to classify tasks based on their resource intensity, and further predicts the expected level of resource requests. Finally, to further improve the energy efficiency of Cloud datacentres, an energy-efficient Cloud-edge collaboration scheduling framework has been designed to reduce energy consumption while improving the efficiency of workload processing. The collaboration framework encompasses a task classification module to judge the type of incoming workload, a meta-heuristic Cloud scheduler (DSGA) for scheduling Cloud tasks, and a light-weighted edge scheduler (EA-DFPSO) for scheduling edge tasks. Extensive experiments demonstrate the effectiveness of the proposed approach in terms of the prediction results and energy saving for the Cloud datacentres.