Predicting Multi-Attribute Host Resource Utilization Using Support Vector Regression Technique

Labeb Abdullah,Huixi Li,Shamsan Al-Jamali,Chang Ruan,Abdulrahman Al-Badwi

doi:10.1109/access.2020.2984056

Abstract

Studies on the resource workload demand in cloud computing environment aim at reducing resource wastage by optimizing the resource utilization in a cloud data center. Based on this goal, most of the existing approaches rely on resource management mechanisms such as resource allocation and Virtual Machine (VM) consolidation to reach an ideal solution for reducing resource wastage. Because of instability and high variability of the cloud resource usage and workloads, there is a demand for cloud providers to apply the prediction methods for forecasting the future cloud resource utilization. This paper employs a supervised statistical learning method, i.e., Support Vector Regression Technique (SVRT), to forecast the future usage of multi-attribute host resource. The method is particularly suitable to handle a non-linear cloud resource workload. To improve the prediction accuracy of SVRT, we decide Radial Basis Function as the kernel function of SVRT and apply Sequential Minimal Optimization Algorithm (SMOA) for the training and regression estimation of the prediction method. Besides, compared with the existing work, we consider the multi-attribute cloud resources other than the single resource. The method is employed under eight sets of real-world workloads, which are collected from BitBrain (BB), PlanetLab (PL) and Google Cluster Workload Traces (GCWT). Series of experiments conducted on the workload dataset show the effectiveness of our approach. Based on evaluation metrics, the final results show that the accuracy was enhanced by approximately 4%-16% and the error percentage was reduced by approximately 8%-60% compared with the state-of-the-art methods.

Highlights

With the great development of Internet and data center networking technologies [1], [2], there is a rapid increasing of tenant requests for cloud computing services, Infrastructureas-a-Service, Network-as-a-Service, and physical resources to host their applications [3]. this significant development has attracted a growing number of tenants towards the services provided by cloud computing, meeting the resource demands of the services for tenants faces multiple obstacles, such as fluctuating in the application tasks, unpredictable demands andThe associate editor coordinating the review of this manuscript and approving it for publication was Tao Zhou .the high variability of the resource usage
- We propose that leveraging multi-attribute resource utilization, to perform the prediction method on the host resource usage, is better than the single resource utilization, and suggest the Support Vector Regression Technique (SVRT) method to predict multi-attribute resource utilization per host in a cloud environment
Based on the predicted values produced by this method, various metrics including Mean Absolute Percentage Error (MAPE), Mean Absolute Error (MAE), Mean Square Error (MSE), R2, Normalized Mean Square Error (NMSE), and PRED are used to predict CPU usage as a single workload data

Summary

INTRODUCTION

With the great development of Internet and data center networking technologies [1], [2], there is a rapid increasing of tenant requests for cloud computing services, Infrastructureas-a-Service, Network-as-a-Service, and physical resources to host their applications [3]. Our approach aims to improve the cloud resource utilization by forecasting the Physical Machine (PM) workload of multi-attribute resources include processors, RAM, disk I/O, and network I/O in the cloud. The SVRT is an effective method to handle the fluctuation of resource utilization For this method, determining the suitable kernel function significantly improves the non-linear modeling ability so that the prediction accuracy is enhanced. Most of the recent methods rely on single resource (i.e., CPU) to predict a host utilization in the future rather than multi-attribute resources as our approach presented. - We analyzed the prediction results based on eight sets of resource workload data include CPU utilization, memory utilization, disk I/O throughput, network I/O throughput

PAPER ORGANIZATION The rest of the paper is coordinated as follows

BACKGROUND

PROBLEM FORMULATION

PERFORMANCE EVALUATION

CONCLUSION