A Prediction based Cloud Resource Provisioning using SVM

Aim: To develop a prediction model grounded on Machine Learning using Support Vector Machine (SVM). Background: Prediction of workload in a Cloud Environment is one of the primary task in provisioning resources. Forecasting the requirements of future workload lies in the competency of predicting technique which could maximize the usage of resources in a cloud computing environment. Objective: To reduce the training time of SVM model. Methods: K-Means clustering is applied on the training dataset to form ‘n’ clusters firstly. Then, for every tuple in the cluster, the tuple’s class label is compared with the tuple’s cluster label. If the two labels are identical then the tuple is rightly classified and such a tuple would not contribute much during the SVM training process that formulates the separating hyperplane with lowest generalization error. Otherwise the tuple is added to the reduced training dataset. This selective addition of tuples to train SVM is carried for all clusters. The support vectors are a few among the samples in reduced training dataset that determines the optimal separating hyperplane. Results: On Google Cluster Trace dataset, the proposed model incurred a reduction in the training time, Root Mean Square Error and a marginal increase in the R2 Score than the traditional SVM. The model has also been tested on Los Alamos National Laboratory’s Mustang and Trinity cluster traces. Conclusion: The Cloudsim’s CPU utilization (VM and Cloudlet utilization) was measured and it was found to increase upon running the same set of tasks through our proposed model.

A New Resources Provisioning Method Based on QoS Differentiation and VM Resizing in IaaS

In order to improve the host energy efficiency in IaaS, we proposed an adaptive host resource provisioning method, CoST, which is based on QoS differentiation and VM resizing. The control model can adaptively adjust control parameters according to real time application performance, in order to cope with changes in load. CoST takes advantage of the fact that different types of applications have different sensitivity degrees to performance and cost. It places two different types of VMs on the same host and dynamically adjusts their sizes based on the load forecasting and QoS feedback. It not only guarantees the performance defined in SLA, but also keeps the host running in energy-efficient state. Real Google cluster trace and host power data are used to evaluate the proposed method. Experimental results show that CoST can provide performance-sensitive application with a steady QoS and simultaneously speed up the overall processing of performance-tolerant application by 20~66%. The host energy efficiency is significantly improved by 7~23%.