Minimum Power Performance-Based Virtual Machine Consolidation Technique for Green Cloud Datacenters

Minimum Power Performance-Based Virtual Machine Consolidation Technique for Green Cloud Datacenters

T.R.V. Anandharajan (College of Engineering, Guindy, Anna University, Chennai, Tamil Nadu, India) and M.A. Bhagyaveni (College of Engineering, Guindy, Anna University, Chennai, Tamil Nadu, India)
Copyright: © 2014 |Pages: 20
DOI: 10.4018/ijgc.2014010103


Infrastructure as a Service is an important component in the cloud building block. The authors present a Cloud Simulation experience with an objective to handle the performance and energy tradeoff in an Infrastructure as a Service (IaaS) environment. The authors present the orchestration of statistical, machine learning and energy model based Minimum Power Performance (MPP) algorithm to validate simulation using PlanetLab VMs real world traces from real systems. Their proposed algorithm consolidates virtual machines (VMs) in the Processing Elements (PE or Host or Server) and is 39% better than the legacy algorithms.
Article Preview

1. Introduction

Computing as a Resource is the buzzword now in Cloud Computing. Organizations business (or) institution has abundant unambiguous resources on the Cloud. Management of such systems seems to be too complex. Therefore researchers use different methodologies and tools to analyze, evaluate and predict their proposed algorithms and methods. Cloud Computing deals with handling virtualized instances of machines in different geographical locations. A Cloud environment is shown in Figure 1.

Figure 1.

Cloud environment

This has lead to capital expenditure over the data centers and the running expenditure like maintain monitoring of the over-provisioned resources. Underutilized servers consume 70% of the power which adds to the running expenditure (Fan, Weber, & Barroso, 2007). We in this paper concentrate on the IaaS system where the under-used servers are put into consolidation and care is taken on the overloaded and normal servers. Objective is a small-memory-footprint, embarrassingly parallel or loosely coupled applications requiring little to no inter-processor communication, given the current interconnect bandwidth and topologies are utilized in the commercial offerings (JAYANT, 2012). While providing the appropriate resources (bandwidth, latency, memory, etc.) the number of physical resources is shared among many users, and the resources may be heterogeneous in a Cloud Datacenter.

1.1. Contributions

  • We analytically show the cost function in terms of SLA and VM migration and analyse the energy cost involved for the simulation;

  • We define Minimum Power Performance (MPP) strategy and analyse other legacy algorithms by energy heuristics modelling;

  • We predict the cloud workload by machine learning technique like Composite Quantile Regression (CQR);

  • We have simulated our statistical real world traces (PlanetLab Trace) comparing our proposed algorithms to provide a better QoS satisfying the trade-off.

The services to be initiated to form an orchestrated simulation have been concentrated in our paper. Machine Learning concepts using regression was implemented on Real-time workload traces and CPU utilization was predicted and proposed algorithms have been addressed. Efficient Energy results were obtained in a simulated Datacenter. We then provide a survey of the relevant literature in Section 2, and present our model in Section 5. We discuss results from our model in Section 11, and conclude in Section 12.

Complete Article List

Search this Journal:
Open Access Articles: Forthcoming
Volume 10: 2 Issues (2019): Forthcoming, Available for Pre-Order
Volume 9: 2 Issues (2018): Forthcoming, Available for Pre-Order
Volume 8: 2 Issues (2017): 1 Released, 1 Forthcoming
Volume 7: 1 Issue (2016)
Volume 6: 2 Issues (2015)
Volume 5: 2 Issues (2014)
Volume 4: 2 Issues (2013)
Volume 3: 2 Issues (2012)
Volume 2: 2 Issues (2011)
Volume 1: 2 Issues (2010)
View Complete Journal Contents Listing