Research On Resource Management Of Cloud Platform Based On Workload Prediction

Over the past few decades,cloud computing has become a transformative technology for the information technology industry.Cloud computing is a concept that uses the Internet to provide a variety of computing resources,including servers,storage,databases,networks,software,analytics and intelligence,to enable faster innovation,more flexible resources and more optimized costs.The rise of cloud computing can be attributed to several factors,including the increasing demand for on-demand access to computing resources,the growth of data and the need to store and process it,and the growing popularity of connected devices and the Internet of Things.As a result,industries and organizations of all sizes have begun to adopt cloud computing to improve their operations and competitiveness.However,as the use of cloud computing increases,so does the need to optimize the use of cloud resources.How to effectively utilize cloud resources has become a hot topic of research,so we will conduct an in-depth study on both cloud platform workload prediction and cloud platform resource management.Specifically,the main work of this thesis is as follows:(1)To address the diverse and dynamic nature of cloud workloads as well as their large scale and complexity,this thesis proposes an integrated load prediction method based on gated recurrent units(GRUs)and convolutional neural networks(CNNs).The method first predicts the workload of a cloud platform using a GRU-CNN deep neural network,and then uses a weighted average of the prediction results from a mean predictor to aid in the prediction.The predicted load data can be used to further analyze how the cloud resources are utilized.Comparative results show that the proposed method outperforms common deep learning methods such as long short-term memory network(LSTM),CNN,and generative adversarial network(GAN)in prediction.Experiments show that the method performs better overall in prediction accuracy,and compared with existing load prediction methods,the proposed method improves 7.9% in prediction metric RMSE and 11.9% in MAPE.(2)To address the major difficulty of how to efficiently and quickly utilize resources in the cloud platform to provide services to users,this thesis proposes a proactive resource allocation based on load prediction and an error-aware resource management strategy,which effectively prevents the missing of services due to insufficient resources.Firstly,the workload prediction model is used to predict the future load,and to reasonably allocate and adjust the resources according to the future load demand.Subsequently,the load prediction errors for each time period are statistically collated and learned to analyze the possible error range for the future time period and adjust the allocated service resources in time.This helps reduce downtime,improve service assurance,and enhance the overall reliability and availability of cloud services.The effectiveness of the approach is evaluated through a simulation procedure,and the results show that it outperforms methods such as Short Job First(SJF),First-Come First-Serve(FCFS),Polling Scheduling(RR),and Clustering Scheduling(CS).Through the method proposed in this thesis,cloud services are completed effectively using service resources,while effectively controlling the waste of resources,and there exists an advantage of at least 1.1% in the evaluation index service guarantee rate.The significance of this study is to make a reflection and experiment for resource management in cloud platform,and prove the effectiveness of the proposed method in workload prediction and resource management in cloud platform.Overall,this thesis proposes a resource management method based on load prediction and error-awareness,which improves the quality of cloud computing services,optimizes the utilization of cloud computing resources,and provides a new solution for the resource management of cloud platforms.