Reducing energy consumption in the cloud environment using virtual machine migration

Master thesis in computer engineering (software)

Abstract

Reducing energy consumption in the cloud environment using virtual machine migration

Reducing energy consumption is one of the most important issues of the day, especially in the industry sector. In recent years, the ever-increasing human needs for computer systems have caused the creation and expansion of more and more data centers with a large number of computers, which in total consume significant electricity. Obviously, in such a situation, many efforts have been made by experts to reduce electricity consumption in these centers, and now efforts in this field are of particular importance. One of the ways to reduce power consumption in data centers is virtual machine migration.

In this thesis, using virtual machine migration, a software package has been designed and implemented, which based on the workload of each computer and the transfer of load between systems in the form of virtual machine migration creates suitable conditions for reducing power consumption, and by using it, it reduces consumption significantly. 

Chapter One

Introduction

In this chapter, the electricity consumption in the computer will be explained first. Then, energy consumption in data centers and finally virtualization are described. 1-1- Energy consumption in a computer The electricity consumption in a computer can be divided into two parts: Static: It is a part of the computer's energy consumption that is only used to keep the system on and is not related to the amount of work that the system does. This level of energy consumption keeps the system on and ready to work and is consumed from the moment the system is turned on. A large part of this energy is actually wasted in different ways and at different levels of hardware; such as current leakage in integrated circuits[1].

Dynamic: it is a part of the computer's energy consumption that is used to perform system activities and varies according to the amount of load[2] on different parts of a system (such as: processor, memory[3], hard disk[4], graphics card[5], etc.). It does not consume, but contrary to what is thought, a server consumes about 60 to 70% of its maximum power[6] when idle [Barroso, 2007] and [Fan, 2007] and [Lefurgy, 2007]. The maximum power consumption of a computer is when it works with its maximum processing power[7].

1-2- Data centers and their energy consumption

A data center is a building, including a large number of computers (servers) and their required parts such as network switches and backup energy sources [Kumar, 2009]. such as storage servers [8], cooling systems, network equipment and

The noteworthy point in this case is the contribution of servers to the energy consumption of the data center of approximately 50%. In other words, only half of the energy consumed by a data center is spent on processing and responding to requests, and the rest is spent on other things, the most important of which are cooling systems. Figure 1-1, which shows the breakdown of the energy consumption of a data center, is a good illustration of this issue.

Regarding the amount of energy consumption in data centers, statistics show that in addition to the significant amount, it has a growing trend in terms of the amount and share of the total energy consumption of the society [Koomey, 2011]. Figure 2-1 represents this issue. According to research [Barroso, 2007], [Boher, 2002], [Rangan, 2008] and [Siegele, 2008], the average utilization [9] of servers in a data center is less than 30%, and a server is close to maximum utilization only 10% of the time [Armbrust, 2010].

Thus, according to the share of energy consumption of a server in idle mode, it can be seen that a significant share of the energy consumption of data centers is wasted.

1-3-     Virtualization

Virtualization was first introduced in the 1970s for the simultaneous use of a system by several users [Bugnion, 1997].

Today, virtualization, including tools that have been added to it, features such as increasing the security of users, especially in non-collaborative spaces, increasing the productivity of servers, creating a suitable platform for different software components under different operating systems and at the same time, simplifying service and maintaining systems in data centers, creating the possibility of load balancing [10] between different servers and so on. which has caused most of the industry, especially data centers, to move towards the use of this technology, as today almost all data centers in the world use this technology [Armbrust, 2010]. Such environments consisting of a collection of computers that use virtualization technology to provide their services are called "supervirtuals"[11]. In fact, cloud computing is the same data centers that provide their services on the network and in the form of hardware packages formed through virtualization [Armbrust, 2010] and [Armbrust, 2009]. We call these hardware packages together with the operating system inside it "virtual machine"[12].

Virtual machine migration[13] is one of the features that were added to it some time after the advent of virtualization, and in short, it is the transfer of a virtual machine from one server to another. Virtual machine migration can be live [14] in such a way that the end user [15] who receives service from the migrating virtual machine does not notice any disruption in receiving the service, and in other words, does not notice the movement of the server virtual machine at all [Clark, 2005]. Figure 1-3 shows a diagram of virtual machine migration between two active servers.

If we want to examine live virtual machine migration more closely, there will actually be a service interruption, which will be between 60 and 300 milliseconds [Clark, 2005]. Anyway, from the point of view of the user and the response to the requests, it is important that virtual machines can be moved between different servers without problems or paying high time and energy consumption [Liu, 2011]. The third chapter is dedicated to the proposed model for reducing electricity consumption in data centers. In the fourth chapter, we will describe how to implement, the environment and how to perform tests. The summary of the results and the suggestions for the next work are presented in the fifth chapter. Chapter Two Research Background 1 Research Background The huge energy consumption in data centers imposes exorbitant costs and side problems such as global warming and intensifying the energy crisis. In such a situation, trying to save this energy becomes particularly important, especially considering the energy loss that occurs in these centers. Therefore, a lot of work has been done in this field, which we will discuss in this chapter.

2-1- Saving computer energy

Methods to save the energy consumed by a computer are divided into two parts, according to which part of the energy they target to save.

2-1-1.  Saving dynamic energy

Dynamic power is a part of the power consumption that is caused by the alternating current and frequency of the parts. To reduce this part of power consumption, there are methods at the hardware and software level.

At the hardware level, by making changes and improving the efficiency of parts during their design and manufacture, such as concentrating circuits as much as possible, using alloys and conductors with higher conductivity, reducing the threshold of significant voltage, energy consumption can be reduced. Such changes will reduce the overall energy consumption of a system, regardless of the environment and under what conditions the system works.

Creating capabilities in the hardware that enable reducing energy consumption in certain situations and at a higher level. Such as predicting several different performance modes for the main processor [16] and placing the possibility of selecting these modes at the software level so that when the system is working, the operating system can determine the optimal performance mode according to the energy consumption according to the working conditions.