Advances in Dynamic Virtual Machine Management for Cloud Data Centers

Advances in Dynamic Virtual Machine Management for Cloud Data Centers

Rashmi Rai (Birla Institute of Technology, India) and G. Sahoo (Birla Institute of Technology, India)
DOI: 10.4018/978-1-5225-2013-9.ch004
OnDemand PDF Download:
$30.00
List Price: $37.50

Abstract

The ever-rising demand for computing services and the humongous amount of data generated everyday has led to the mushrooming of power craving data centers across the globe. These large-scale data centers consume huge amount of power and emit considerable amount of CO2.There have been significant work towards reducing energy consumption and carbon footprints using several heuristics for dynamic virtual machine consolidation problem. Here we have tried to solve this problem a bit differently by making use of utility functions, which are widely used in economic modeling for representing user preferences. Our approach also uses Meta heuristic genetic algorithm and the fitness is evaluated with the utility function to consolidate virtual machine migration within cloud environment. The initial results as compared with existing state of art shows marginal but significant improvement in energy consumption as well as overall SLA violations.
Chapter Preview
Top

Introduction

With the advancements in computing power and the rapid increase in computing services being delivered as a utility to the consumer, there is a paradigm shift towards cloud computing technologies (Armbrust et al., 2010; Bilal et al., 2013; Foster et al., 2008). Big and small organizations, businesses as well as individual users have started relying on cloud services instead of building and managing their own data centers for fetching the required services. As a result of which there has been sharp increase in the number of large scale data centers across globe. For example, there are more than 454,000 servers for Amazon EC2 and it is steadily increasing every year (Amazon data centre size, 2012).

The number of datacenters is growing at a steady pace as per the recent forecast by CISCO and the cloud workload will be almost triple from 2013 to 2018.However the work load in traditional data centers will decline as per the survey due to increasing virtualization in cloud environment (Cisco Global Cloud Index, 2013).

These large-scale data centers consume humongous amount of energy leading to huge operating costs and also contribute significantly towards the global CO2 emission. According to (Koomey, 2011), the total energy consumption by data centers will persist to grow at a past pace until unless some sophisticated energy efficient measures are deployed. Thus, reducing the energy usage in cloud data centers have become a prime concern across the globe, both for the sake of cloud providers benefit and the greener environment.

Reduction in high energy usage involves eliminating the energy waste that happens at various levels of cloud data center environment. At the hardware level the energy usage can be minimized by improving the physical infrastructure of data centers while at software level optimizing various resource allocations and scheduling algorithms can reduce the energy wastage. Latest and advanced designed data centers have resulted in drastic increase in infrastructure efficiency. For example, the social media giant Facebook’s data center located at Oregon has successfully achieved a PUE (Power Usage Effectiveness) of 1.08, this ratio (Open Compute Project, 2015) clearly indicates that roughly 91% of the data center’s energy is used by the computing resources. Another very recent announcement made by Facebook is the newest datacenter at Fort Worth, Texas which is expected to be one of the most advanced and efficient datacenter that will use 100% renewable energy.

At software level the inefficient usage of computing resources accounts for the maximum amount of energy waste. As per the analysis of more than 5000 servers for a period of six months it was noted that the server utilization barely ever reached full potential although servers were not kept idle. These servers mostly operated at almost 10-50% of their complete capacity which caused added cost towards maintenance and management of over-provisioned servers leading to increased TCO i.e. Total Cost of Ownership (Barroso and Holzle, 2007). Hence underutilized servers are detrimental both from cost as well as energy usage perspective.

Our work mainly focuses on improving the energy usage at software level specifically the virtual machine migration level. The main contributions of this chapter are the following.

  • 1.

    A Utility function based Meta heuristic for the efficient management of virtual machine migration which significantly reduces the energy consumption in a cloud environment.

  • 2.

    An elaborate simulation and performance evaluation through a different set of experiments conducted for 10 simulation days.

Complete Chapter List

Search this Book:
Reset