ANALYSIS OF CLOUD COMPUTING

IN OPEN SOURCE PLATFORM

Fui Fui Wong

1, 2

and Cheng Xiang Tan

Department of Computer Science and Technology, TongJi University, Shanghai, 201804, China

School of Information Technology, Kuala Lumpur Infrastructure University College, Kajang, 43000, Malaysia

Keywords: Open source cloud computing, Cloud computing, Eucalyptus, OpenNebula, Nimbus, abiCloud, openQRM,

Xen cloud platform.

Abstract: There is an increasing interest in the use of open source cloud computing infrastructure platform by larger

companies, academic researchers, scientists and service providers that can tailor the platform for specific

users. The open source cloud computing has joined the mainstream where it is starting to offer companies a

lot of flexibility and cost savings to combine public and private clouds. These tools could let companies to

build and customize their own computing clouds to work alongside with more powerful commercial

solutions. To understand these issues, this paper outlines an analysis study of open source cloud computing

platform using six of the most advanced platforms (Eucalyptus, OpenNebula, Nimbus, abiCloud, openQRM

and Xen Cloud Platform) in order to gauge whether its potential opportunities could outweigh their possible

drawbacks.

1 INTRODUCTION

Cloud computing provides a valuable way for many

business and government organizations to

implement low cost, low power and high efficiency

systems to deliver scalable infrastructure. It means

that there will be no upfront investment in servers or

software licensing. Instead of spending millions of

dollars and months in building their own

infrastructures, a paradigm shift is taking place to let

smaller organizations tap on the infrastructure of

giant organizations According to a recent Gartner

research report, cloud computing is expected to be

the top priorities for CIOs in the year 2010 as the

businesses shift from a traditional in-house data

center to a hosted solution.

However, there are growing concerns over the

control ceded to large commercial cloud vendors

come with drawbacks, including the lack of

information privacy, security, and reliability

concerns. To understand these issues, this article

aims to take a closer look at the open source cloud

computing platforms in order to gauge whether their

potential opportunities outweigh their possible

drawbacks.

2 EMERGING OPEN SOURCE

CLOUD COMPUTING

INFRASTRUCTURE

PLATFORMS

Lately, the open source cloud computing has joined

the mainstream where it is starting to offer

companies a lot of flexibility and cost savings to

combine public and private clouds . It has become

very clear that open source cloud platform occupies

a significant portion of cloud market. Thus, it is

important to examine the characteristics and

suitability of open source cloud computing platforms

for the public understandings.

2.1 Eucalyptus Cloud Computing

Eucalyptus (2010e) is a Linux-based open source

software developed by the Computer Science

Department at the University of California, Santa

Barbara as the research project to support the high

performance distributed computing environment

(Nurmi et al., 2009). Eucalyptus aims to implement

on-premise and private clouds within an enterprise’s

existing IT infrastructure that is commonly referred

as Infrastructure as a Service (IaaS). The most

158

Fui Wong F. and Xiang Tan C..

ANALYSIS OF CLOUD COMPUTING IN OPEN SOURCE PLATFORM.

DOI: 10.5220/0003382701580165

In Proceedings of the 1st International Conference on Cloud Computing and Services Science (CLOSER-2011), pages 158-165

ISBN: 978-989-8425-52-2

 2011 SCITEPRESS (Science and Technology Publications, Lda.)

remarkable feature of Eucalyptus is emulated many

Amazon Elastic Compute Cloud (Amazon EC2)

functionalities has attracted many researchers and IT

professionals to use this free version of EC2. For

example, Eucalyptus supports many EC2

functionalities like the EC2 interface, the Amazon

Simple Storage Service (Amazon S3) interface, the

Elastic Block Store (EBS), Zone, Elastic IP

addresses, and security groups. The supported

virtualization techniques are: Xen and Kernel-based

Virtual Machine (KVM).

2.1.1 Eucalyptus Design

The Eucalyptus infrastructure consists of five main

components, the Cloud Controller (CLC), the

Cluster Controller (CC), the Node Controller (NC),

the Storage Controller (SC) and Walrus (Nurmi et

al., 2008). The deployment of the Eucalyptus is

engineered to mimic the Amazon Web Services

(AWS) as a result of supporting AWS Application

Programming Interfaces (APIs) from Eucalyptus

without modification. Figure 1 shows the Eucalyptus

architecture.

Figure 1: Eucalyptus environment.

The cloud controller (CLC) is a top-level

component that aggregate resources from multiple

clusters to accommodate the incoming user requests

and authentication, resource allocation, persistent

user and system data, manage and monitor the

underlying virtualized resources. The Walrus is also

a top level component in a cloud architecture that

written in Java program to implement bucket-based

storage through Amazon S3-compatible SOAP and

REST interfaces.

The Cluster Controller (CC) is a cluster-level

component that executes on a cluster front-end

machine to handle cluster-level scheduling and

network control. Each cluster requests a cluster

controller (CC) to communicate with both the nodes

running Node Controllers (NC) and to the machine

installing the CLC. The Storage Controller (SC) is

one of the cluster-level components that implement

EBS-style block-accessed network storage. The SC

emulates the functionality of Amazon Elastic Block

Storage. It manages dynamic block devices (e.g.

EBS) that Virtual Machines (VM) can use for

persistent storage as well as supporting various

storage systems (e.g. NFS, iSCSI, etc.)

Each physical machine in the cluster with a

hypervisor will need a Node Controller (NC). The

NCs control the execution, monitor the state

information and termination of the VM instances on

that particular machine in response to enquiries and

control requests from the cluster controller (CC).

2.2 OpenNebula Cloud Computing

OpenNebula (ONE)(2010g) is an open source toolkit

for cloud computing to build any type of Cloud

deployment (e.g. Private, Hybrid and Public

Clouds). Its origins are as a research project at

Complutense University of Madrid in 2005. The

aim of the OpenNebula is to produce a Virtual

Infrastructure (VI) management solution with the

flexible architecture, interfaces and components that

fits into any existing data center and external clouds

(Sotomayor et al., 2009) .

2.2.1 OpenNebula Design

The OpenNebula private cloud implements a

classical cluster-based architecture, as shown in

figure 2. It consists of a front-end cluster which also

known as administration node, several cluster nodes

where virtual machines (VM) will be installed, and

an image repository that holds the base images of the

VMs. Communications between front-end and

cluster nodes use SSH.

The administration node executes on a front-end

cluster machine, interconnect with a LAN to the

cluster nodes that serve the user requests. As shown

in the figure 2, the front-end machine is where

OpenNebula is installed. OpenNebula is a Virtual

Infrastructure (VI) manager that is executed and

configured by a Front-End cluster. The OpenNebula

internal architecture (see Figure 3) can be classified

into three layers, the Tools, the Core and the

Drivers dedicated to different aspects

of VI management (2010f).

ANALYSIS OF CLOUD COMPUTING IN OPEN SOURCE PLATFORM

159

Figure 2: OpenNebula design.

Figure 3: OpenNebula internal architecture.

To summarize, the tools layer consists of 3

components, Scheduler, Command Line Interface

(CLI) and other third party tools developed using the

interface given by the OpenNebula Core layer. The

Command Line Interface allows the infrastructure

administrators and users to submit, control and

monitor the virtual machines; to add, delete and

monitor the hosts, virtual networks and users

respectively. The scheduler component is an

independent entity in the OpenNebula architecture

that is responsible for the assignment between the

pending Virtual Machines and it is known as Hosts.

The Haizea Lease Manager(2009a) can be used as a

drop-in replacement for OpenNebula’s scheduling

daemon to support advanced reservation and

preemption of resources and queuing of best effort

requests (Sotomayor et al., 2008).

The Core layer encompasses several components

to control and monitor virtual machines, virtual

networks, storage and hosts through pluggable

drivers. To perform VM’s life cycle, the

OpenNebula core orchestrates six different

management areas: the request manager for

answering client requests; the virtual machine

manager for the management and monitoring of

VMs; the transfer manager for transferring VM

images among the cluster node, image repository

and cluster front-end; the virtual network manager

for creating of virtual network and handling of IP

and MAC addresses; the host manager for managing

and monitoring the physical hosts; the database for

storing the persistent data needed in the management

VMs.

Drivers are separate processes that communicate

with the OpenNebula core.

2.3 Nimbus Cloud Computing

The Nimbus toolkit is an open source cloud

computing that help to turn your existing

infrastructure into an Infrastructure-as-a-service

(IaaS) cloud with cloud-like interfaces. Its primary

objective is to provide an IaaS-style cloud platform

to the scientific community. The Nimbus v2.4

version eliminates the need for a separate Globus

container installation and added significant amount

of new improvements and features. It also leverages

Amazon EC2 WSDLs, Amazon EC2 Query API and

Grid community WSRF interfaces and offers self-

configuring virtual cluster support. Virtualization

implementation is based on Xen and KVM

2.3.1 Nimbus Design

The architecture of the Nimbus system is simple and

flexible with a list of components to allow a user to

dynamically deploy and manage VMs. The software

framework is a highly modular cooperative set of

web services that interoperate using standard

communication protocols. Each Nimbus component

is lightweight and self-contained that can be sanely

replaced by other module in a variety of ways. The

Nimbus 2.4 version contains a piece of software

called Nimbus service will install on one service

node per site and another piece of software known as

Control agent to setup on Virtual machine monitor

(VMM) nodes as shown in Figure 4. Clients interact

with the service using X.509 credentials via multiple

protocols. User can choose to launch VM via

different clients interface.

The Nimbus service internal architecture (see

Figure 5) encompasses several components

specialized in different aspects of creating,

controlling, and monitoring virtualized resources on

an infrastructure-as-a-service cloud. The workspace

service is a site manager which allows remote

protocol frontends (for example, web service

resource framework (WSRF), EC2 WSDL)) to

deploy and manage VMs. The workspace resource

manager executes VM deployment. The workspace

CLOSER 2011 - International Conference on Cloud Computing and Services Science

160

pilot implements the integration with existing VMs

with the job assigned.

Figure 4: Nimbus design.

Figure 5: Nimbus service internal architecture.

2.4 abiCloud

Abicloud (2010a) has been developed as an open

source cloud solution for implementing the creation

and integral management of public and private cloud

in heterogeneous environment. It was developed and

maintained by Abiquo(2010c). AbiCloud is mainly

active in Spain which contains three different

versions (such as community, enterprise and xSP

versions) to target the different needs of commercial

organizations.

2.4.1 abiCloud Design

The AbiCloud design is primarily motivated by the

concept of resource management where it applies the

total separation of the physical infrastructure from

the virtual application infrastructure. Therefore,

AbiCloud is as scalability as a result of its loosely

coupled modular system. The Abicloud (Figure 6) is

composed of three main components (2010d). The

abiCloud-Server is the core management software

written in java that is running in the centralized

servers store of the global business logic of the cloud

platform. It acts as a front-end to communicate with

the database, the rest of the modules in AbiCloud

platform and extensible by adding more servers

components and load balancer. The abicloud WS

(Web Service) is responsible for the management

and monitoring of Virtual Machines. The plug-in

manager works with WMware ESX and ESXi,

Microsoft Hyper-V, Virtual Box, Xen, KVM. The

abiCloud VMS (virtual monitor system) is in charge

of the virtual infrastructure for the correct

deployment of virtual machines by keeping track of

the events or states in the cloud.

VMWare

abiCloud_Server

(JAVA)

Third Parties apps/

Cloud platforms

abiCloud-client

Open API’s

Remote Objects

connector

Web services interface

abiCloud_WS

(JAVA)

abiCloud_VMS

(JAVA)

abiCloud

Appliance

Manager

(JAVA)

abiCloud

Storage

Management

Plugin

Manager

Plugin

Manager

Database

VBox XEN KVM

…….xVM

Figure 6: Platform architecture of abiCloud (2010d).

2.5 openQRM Cloud Computing

The openQRM (2010b) (open Qlusters Resource

Manager) was initially developed by the Qlusters

Company back in year 2001 as the commercial

license software to manage the virtual machine

deployment in data center. The openQRM 4.6

supports the Amazon EC2 adapter which allows for

the creation of hybrid Cloud Computing, migration

to other cloud provider such as Ecuclyptus or UEC

and vice versa(2010h). The supported virtualization

techniques are: VMware, Xen, KVM and Linux

VServer.

2.5.1 openQRM Design

The platform design of the openQRM is primarily

motivated by two engineering goals: high-

availability and automatism. openQRM is based on a

strictly pluggable architecture so that they can be

selected and composed in a variety of ways. It has

no feature of itself but all the features of openQRM

are provided by its plugins. Indeed, the openQRM

server is designed to just manage the plugins



ANALYSIS OF CLOUD COMPUTING IN OPEN SOURCE PLATFORM

161

through its automated and generic interfaces without

changing a single line of code in the base server

(Rechenburg, 2010). The proven open-source and

commercial third party plugins in openQRM

perform the actual implementation which integrated

into the openQRM server framework to

communicate with the internal server-object to carry

out the required action. The plugins tools are ranging

from Server-management, Storage-management,

deployment, provisioning, remote administration,

local server integration, high availability, real-time

monitoring and etc (figure 7).

Figure 7: Platform architecture of openQRM server.

2.6 Xen Cloud Platform

Xen Cloud Platform (XCP) (2010i), a project

initiated by Xen.org community. The most striking

feature of Xen Cloud Platform is its infrastructure

supports other popular open source projects, for

example, Eucalyptus, OpenNebula, Nimbus,

Convirture, OpenXenCenter and Xen VNC Proxy to

leverage the Xen hypervisor on cloud computing

environment (2009d). The release version of XCP

0.1 leverages other Xen tools, such as Latest Xen

3.4.1, Linux 2.6.27 Kernel, Window PV Drivers,

Citrix XAPI enterprise-class management tool stack,

basic SR-IOV Support. It has the particularity to

address the need of ISVs and service provider by

supporting a complete virtual infrastructure cloud

service for multi-tenancy, SLA guarantees, security

and network virtualization technologies (2010i).

2.6.1 Xen Cloud Platform Design

The Xen Cloud Platform is a server virtualization

platform that consists of four main components as

shown in figure 8 (2009b) that is dedicated entirely

to the task of hosting VMs. The Xen hypervisor is

installed between the server’s hardware and

operating system which provides an abstraction

layer to run one or more guest operating systems.

The control domain is a virtual machine that is

responsible for controlling host hardware device and

creating more guest domains. Storage manager

(SMAPI) and XAPI management stack are

components that running inside the control domain.

To be a part of resource pool, a collection of Xen

Cloud Platform host that group together into a single

unit of administration which can host virtual

machines across a cluster of machines (2009c).

Figure 8: Xen cloud platform architecture.

3 COMPARISON OF THE 6

CLOUD INFRASTRUCTURE

PLATFORMS

In this section we analyze six major open source

cloud computing infrastructure players in this arena

against 5 levers: scalability, high availability,

interoperability, virtualization management and

deployment management.

Scalability refers to the ability of a computer

system to be scaled up easily to meet the increased

demand in order to maintain performance levels as

the number of concurrent requests raises. For

instance, Eucalyptus does not support for auto

scaling; OpenNebula allows dynamic resizing the

physical infrastructure which makes it easily to

expand or shrink according to the demand although

it is not fully automated; Nimbus permits the client

to specify the resource allocation during

deployment; Abicloud allows to automatically

modify soft limits to scale the infrastructure which

let the users to control the provision of servers,

storage, networks, virtual network devices and

applications. OpenORM supports growth and auto-

scaling through a power management plug-in.

XenCloud anticipates the future expansion by

CLOSER 2011 - International Conference on Cloud Computing and Services Science

162

manually adjusting the capacity through the

workload balancing deployment.

Availability concerns may stop some companies

from using cloud computing for transaction oriented

applications due to the cloud computing has not

always provided round-the-clock reliability. Looking

at the current providers, Eucalyptus, Nimbus and

AbiCloud are not supported for fault tolerance

whereas no documentation can be found about this

feature; OpenNebula using a persistent backend

database to store their hosts and VM information

through its repository; openQRM maintains a stand-

by server for fault tolerance that allows groups of

servers to use this single backup server to achieve

the N to 1 failover concept; XenCloud allows

administrator to restart the failed VMs on another

host if high availability is activated on the resource

pool.

Interoperability creates an open infrastructure /

framework that let applications and data to

seamlessly work together for the purpose for wider

industry adoption of cloud computing technology.

For example, Eucalyptus and openQRM support the

Amazon EC2 API; openNebula has the capacity for

integrating with other cloud computing tools like

Eucalyptus, Amazon, ElasticHosts and Nimbus and

supports open standard such as OGF's Open Cloud

Compute Interface specification and Open

Virtualization Format (OVF); Nimbus has an open

API which enables the integration of other third

party applications such as Amazon EC2 WSDLs,

Amazon EC2 Query API and Grid community

WSRF; abiCloud recommends the usage of open

virtualization format between datacenters and

integrate with the Sun’s Open Cloud API; XenCloud

will interact and integrate with other tools and

products, like Eucalyptus, Convirture, OpenNebula,

OpenXenCenter, Xen VNC Proxy, and Nimbus. It is

also compatible with the Distributed Management

Task Force (DMTF), CIM and Open Virtualization

Format (OVF) standards.

Virtualization technology is a vital part of

establishing platform leadership. It would be

impressive if live migration of running virtual

machines are able to be carried out from one

physical server to another with zero downtime and

undetectable to the users. However, this feature is

yet to be a part of the cloud framework. For

instance, Eucalyptus, Nimbus and abiCloud are not

supported for managing machines on a cluster with

features as live migration; OpenNebula supports live

migration but only can be implemented with Shared

FS; openQRM with remarkable solution for various

migration scenarios such as Physical to Virtual,

Virtual to Physical, Virtual to Virtual; XenCloud

supports live migrations through its complete list of

plug-in.

One aspect of the deployment management is the

user interface. Graphical user interface to access the

cloud can ease the deployment of cloud compared to

the command line that acts as a barrier to entry. For

instance, Eucalyptus has well integrated with third

party API, with the use of graphical tools from

RightScale’s and amazon web interface, Nagios,

Ganglia, Elasticfox and Hybridfox to ease the

management of virtual machine instances;

OpenNebula relies on the command line interface

(CLI) for the deployment of cloud. Nimbus comes

with a friendly web interface to manage

administrative and user functions. AbiCloud has a

straightforward and comprehensive one stop

graphical user interface build with Adobe Flex that

decreases the effort of the platform deployment.

OpenQRM provides a user-friendly web interface to

monitor the status and performance of its managed

subsystems. XenCloud has a web interface that used

to manage the VMs and perform live migrations via

a web browser.

Table 1: Summary of comparisons among cloud

infrastructure platforms.

Criterion Eucalyptus OpenNebula

Auto Scaling No No

Fault Tolerance No Yes

(use persistent database

backend to store hosts,

networks and virtual

machines)

Interoperability Amazon

EC2/S3/EB,

Rackspace,

GoGrid.

Amazon EC2,

ElasticHost,

GoGrid,

Rackspace,

Terremark,

RimuHosting

Eucalyptus, Nimbus

Live migration No Yes

User interface CLI Unix-like command line

interface

Different strategies are employed by the open

source infrastructure platform vendors. Clearly there

is no single vendor which is good in all identified

criterions. We cannot say one vendor is better than

the others. However, enterprises need to know their

key concerns when moving their data center into the

cloud.

ANALYSIS OF CLOUD COMPUTING IN OPEN SOURCE PLATFORM

163

Table 2: Summary of comparisons among cloud

infrastructure platforms (contd.).

Criterion Nimbus Abiquo

Auto Scaling No No

Fault Tolerance No No

Interoperability Amazon EC2 Sun’s Open

Cloud API

Live migration No No

User interface web interface Graphical User

Interface

Table 3: Summary of comparisons among cloud

infrastructure platforms (contd.).

Criterion openQRM Xen Cloud

Platform

Auto Scaling Yes

(via its power

management

plug-in)

Fault Tolerance Yes

(with fault-

tolerance and

fail-over

capabilities to all

applications)

Yes

(with the ability

to restart the

failed VMs on

other machine

automatically)

Interoperability Amazon EC2,

Eucalyptus, UEC

Eucalyptus,

Convirture,

OpenNebula,

OpenXenCenter,

Xen VNC Proxy,

Nimbus

Live migration Yes(P2V, V2P,

V2V)

Yes

User interface Web Interface Web Interface

4 CONCLUSIONS

Along the way, the open source movement has

touched almost every sphere of cloud computing that

we know today. It is good to see open source

becomes one of the key drivers competing in the

cloud computing world to provide more flexible and

rich capabilities to cloud providers. The use of the

open source platform as a viable alternative to

proprietary software is of particular interest to the

organizations today. However, presently open source

cloud computing is relatively new and mostly still in

the continuous development stage and yet to reach

the maturity level. The work presented in this article

will need updates as new releases are constantly

emerging. Looking forward to see more innovative

features in the open source platforms that eventually

would benefits the human societies in the same way

as what is Linux has done to the world since it has

been developed. We believe that cloud computing

will change and evolve our daily life.

REFERENCES

History (Online). Available: http://www.opennebula.org/

about:history (Accessed 20/4/2010).

Nimbus (Online). Available: http://www.nimbusproject.

org/ (Accessed 24/4/2010).

Nimbus 2.4 Documentation (Online). Available: http://

www.nimbusproject.org/docs/2.4/index.html (Ac-

cessed 22/4/2010).

2009a. Haizea (Online). Available: http://haizea.cs.

uchicago.edu/index.html (Accessed 24/4/10 2010).

2009b. Xen Cloud Platform 0.1 User Security. 1.0 ed.:

Xen.org.

2009c. Xen Cloud Platform Administrator's Guide.

Release 0.1. 1.0 ed.: Xen.org.

2009d. Xen in the Cloud (Online). Xen.org. Available:

http://www.xen.org/files/Marketing/XeninCloud.pdf

(Accessed 1/5/2010).

2010a. abicloud Home (Online). Available: http://

community.abiquo.com/display/ABI15/Home (Ac-

cessed 2/5/2010).

2010b. About openQRM (Online). Available: http://

www.openqrm.com/ (Accessed 5/5/2010).

2010c. About Us (Online). Available: http://

www.abiquo.com/company/about-us.php (Accessed

7/5/2010).

2010d. Architecture Overview (Online). Available: http://

abicloud.org/display/admin/Architecture+Overview

(Accessed 2/5/2010).

2010e. Open Source Home (Online). Available: http://

open.eucalyptus.com/ (Accessed 5/5/2010).

2010f. OpenNebula 1.4 Architecture (Online). Available:

http://www.opennebula.org/documentation:rel1.4:arch

itecture (Accessed 24/4/10 2010).

2010g. OpenNebula Home (Online). Available: http://

www.opennebula.org/ (Accessed 23/4/10 2010).

2010h. openQRM 4.6 release brings streamlined interface

design and new professional cloud management tools

(Online). Available: http://www.openqrm-

enterprise.com/about-openqrm/release-notes.html (Ac-

cessed 5/5/2010).

2010i. Xen Cloud Platform - Advanced Virtualization

Infrastructure for the Clouds (Online). Available:

http://xen.org/products/cloudxen.html (Accessed

1/5/2010).

Nurmi, D., Wolski, R., Grzegorczyk, C., Obertelli, G.,

Soman, S., Youseff, L. & Zagorodnov, D. 2009. The

Eucalyptus Open-Source Cloud-Computing System.

Proceedings of the 2009 9th IEEE/ACM International

Symposium on Cluster Computing and the Grid. IEEE

Computer Society.

Nurmi, D., Wolski, R., Obertelli, C. G. G., Soman, S.,

Youseff, L. & Zagorodnov, D. 2008. Eucalyptus : A

Technical Report on an Elastic Utility Computing

Archietcture Linking Your Programs to Useful

CLOSER 2011 - International Conference on Cloud Computing and Services Science

164

Systems. UCSB Computer Science Technical Report

Number 2008-10 (Online). Available:

http://open.eucalyptus.com/documents/nurmi_et_al-

eucalyptus_tech_report-august_2008.pdf.

Rechenburg, M. 2010. The OpenQRM Datacenter

Management and Cloud Computing Platform.

Germany: openQRM Enterprise GmbH.

Sotomayor, B., Montero, R. S., Llorente, I. M. & Foster, I.

Year. Capacity Leasing in Cloud Systems using the

OpenNebula Engine. In: Workshop on Cloud

Computing and its Applications 2008 (CCA08),

October 22-23, 2008 2008 Chicago, Illinois, USA.

Sotomayor, B., Montero, R. S., Llorente, I. M. & Foster, I.

2009. Virtual Infrastructure Management in Private

and Hybrid Clouds. Internet Computing, IEEE, 13, 14-

22.

ANALYSIS OF CLOUD COMPUTING IN OPEN SOURCE PLATFORM

165