Workload prediction for cloud services by using a hybrid neural network model

Rawat, Preeti (2022) Workload prediction for cloud services by using a hybrid neural network model. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (1MB) | Preview
Preview	PDF (Configuration manual) Download (1MB) | Preview

Abstract

Cloud Service Providers hold vast volumes of computing resources in the cloud data centers and with advancements in the field of computing, it is possible to provide resources to the users on-demand. Automatic resource allocation shields the customers from the concerns of infrastructure needs. But there are certain issues in cloud computing related to resource management. It is important to have a resource allocation strategy to avoid resource over or under-provisioning, which in turn ensures that quality of service (QoS) is maintained as per the Service Level Agreement (SLA) between the cloud service provider and customer. Otherwise, cloud service providers have to face heavy penalties for not abiding by the SLA. To avoid these concerns a workload prediction method is proposed in this paper, which uses machine learning algorithms like Support Vector Regression (SVR) and Long-Short Term Memory (LSTM). Python language is used for the implementation of the proposed model. For conducting the experiments, a time series dataset is generated, which includes pseudo randomness. Savitzky-Golay filter is used to remove the outliers and noise from the input signal. After that, 1 scale Wavelet Transformation is used to divide the input signal into high and low-frequency components. Low-frequency components are passed to SVR for training whereas high-frequency components are passed to LSTM for predicting computational load for the next time slot, after that by using Inverse Wavelet Transformation (IWT) output from both algorithms are combined to generate the original series. The proposed model is evaluated by measuring Accuracy(R2), RMSE, and MSE. The metrics are calculated for two models, which are the LSTM only and the Proposed model (LSTM and SVR) and it was found that the proposed model outperformed another model.

Item Type:	Thesis (Masters)
Subjects:	Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Cloud computing Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
Divisions:	School of Computing > Master of Science in Cloud Computing
Depositing User:	Tamara Malone
Date Deposited:	16 Dec 2022 10:46
Last Modified:	07 Mar 2023 17:42
URI:	https://norma.ncirl.ie/id/eprint/5989

Actions (login required)

View Item