Cloud Running Cost Optimization using Inactive VM Identification by Data Mining Technique

Authors: Vaidehi Bakshi, Mohit Jain

Certificate: View Certificate

Abstract

Cloud is a huge infrastructure; it is composed of networks, data centers and the brokers. Among broker is third party entity which is associated with a number of cloud infrastructures. Additionally different data centers are associated with other infrastructures. That mess is created to offer scalability on computational resources. The service providers borrow the resources from other infrastructure when the self resources are in high workload, and the resources are ideal when the fewer loads arise for computation. Both the conditions increase the cloud infrastructure running cost. In this context this issue is observed when the resources are not properly managed. Thus the cloud computing is focused on appropriate utilization of computational resources for achieving the higher performance from the cloud infrastructures. The management of resources can be possible by adopting the scheduling strategy or by managing the virtual machines. The proposed work is focused on management of virtual machines in cloud servers. In this context the proposed work offers a data mining technique that is helping us to identify the ideal or inactive VMs in infrastructure. Basically the inactive VMs are degrading the services of cloud data centers because these machines engage the cloud resources but it not functioning as requirements. Thus the inactive VMs are increases the server running cost, to maintain the performance it is necessary to identify and repair the VMs. The proposed technique usages the k-means and SVR (support vector regression) to classify the cloud server log for identifying the target types of VMs. The implementation of the presented technique is performed using JAVA technology and it is observed the proposed technique works more efficiently as compared to the similar available techniques..

Introduction

The cloud computing enhancing the computational experience, it offers the computational as well as the storage resources. Due to its characteristics it offers shared resources, scalable computational and storage resources, and more. In order to achieving this, different resource management techniques are applied. Among them the virtualization, load balancing and resource scheduling techniques are much popular. In this work virtual machine management is simulated for recovering performance on existing cloud infrastructure. Therefore, the aim of the cloud is to optimize the resource utilization for maintaining the service quality. The proposed work is involved for resource management of cloud servers. The data centers are basically configured with the multiple VMs for effective resource utilization, but the less number of active VM not provide effective productivity. Thus identification and repairing of VMs are required. In this context for classifying the VMs the data mining technique is used. That technique help to analyze the server log and provide the outcomes based on classification of active and inactive VMs. The proposed work involves the study of data mining learning methods supervised and unsupervised respectively. The data mining techniques are employable over the data in different manner for recovering the target patterns. This target patterns are help to recognize the properties of VM which are not working or damaged. However the proposed work involves the supervised learning approach for efficient and accurate classification of inactive VMs.

Conclusion

The cloud computing offers scalable computing, plague and play resources. Additionally according to the use of application that is able to arrange the required resources. But to maintain the service quality a significant amount of background efforts are required. The cloud computing is a huge infrastructure, using this large computational problems are solved. In this context the efficient resource management techniques are required. Additionally unused resources are adding penalty over the cloud service providers. Thus the virtualization techniques are used for obtaining higher computational throughput. In this context multiple VMs are created over a single host or data center. But when the VMs are not working properly or damaged due to some reasons it is necessary to repair these deficiencies. Thus recognition of inactive VMs is required to improve the performance of cloud infrastructure. The proposed work involves the data mining techniques for this task. The data mining techniques are able to analyze the generated huge log files and can successfully identify the target VMs.