2012 12th IEEE/ACM International Symposium on Cluster, Cloud and Grid Computing

A Compartive Study of Cloud Computing middleware

Chaker El Amrani, Kaoutar Bahri Filali, Kaoutar Tarek El-Ghazawi

Ben Ahmed, Amadou Tidiane Diallo, Stéphano Department of Electrical and Computer Engineering
Telolahy The George Washington University
Department of Computer Engineering Washington DC, USA
Abdelmalek Essaadi University organization [email protected]
Tangier, Morocco
[email protected]

Abstract—Cloud computing is an emerging IT technology that Platform as a Service (PaaS) or Infrastructure as a Service
is being used increasingly in industry, government and (IaaS), as shown in “Fig. 1”.
academia. There are several Cloud Computing middleware
solutions available in the market. This paper proposes an
approach and set of characteristics and metrics for comparing
Cloud computing middleware based on functionality. Three
popular open source middleware: Nimbus, Eucalyptus and
OpenNebula, are also analyzed and evaluated on the basis of
the proposed parameters. Future work will include more
systems and an experimental benchmarking to study the
relative performance.

Keywords-component; cloud middleware; functional

characteristics; conceptual metrics; Nimbus; Eucalyptus;
OpenNebula.
Figure 1. Cloud Computing Overview
I. INTRODUCTION IaaS provides machine instances, PaaS provides a
The development of Cloud Computing systems has gone programming environment that abstracts machine instances
through enormous advances in recent years. Cloud and other technical details, and SaaS, which is the most
Computing is a model for enabling ubiquitous, on-demand popular Cloud Computing services, consists on applications
network access to a shared pool of configurable computing offering online resources and storage.
resources. Consumers have access to large amounts of data The main techniques used in Cloud Computing are [4]:
and computational infrastructures through resource • Virtualization: provides on-demand flexible and
management middleware. The users pay based on their usage scalable platforms.
[1, 2]. • Workflow orchestration: Cloud should be able to
Several commercial and open source Cloud computing automatically orchestrate services from various
middleware are available and comparing and contrasting sources and of different types to form a workflow for
them is not easy. In order to better choose the appropriate users.
solution, a methodology is based on a set of functional • Web service and SOA: make Cloud available on
characteristics has been developed, and a detailed various distributed platforms and can be accessed
comparison between three middleware systems has been through networks.
presented in this paper. • Web 2.0: is related to the development of Web-based
communities and business models.
II. CLOUD COMPUTING ARCHITECTURE • Large distributed storage system: Cloud should
A Cloud Platform is composed of physical or virtual include a network storage system and a distributed
servers, storage systems, datacenters, networking devices, data system.
hypervisors and middleware. The architecture of cloud Clouds can be divided into three categories: Private
services is based on a real-time scalable approach. Cloud Clouds where the infrastructure is dedicated to a particular
computing is the delivery of computing resources as a organization, Public Clouds where resources are shared
service, whereby shared resources, software, and information between multiple organizations and a combination of both
are provided to computers and other devices as a utility over called Hybrid Clouds where some resources are in-house and
a network [3]. Cloud Computing thus provides to users on- others are provided externally [5].
demand services, which can be Software as a Service (SaaS),

This work was supported in part by the I/UCRC Program of the National
Science Foundation under Grant No. IIP-0706352.

978-0-7695-4691-9/12 $26.00 © 2012 IEEE 690

DOI 10.1109/CCGrid.2012.129
 III. CLOUD MIDDLEWARE FUNCTIONALITIES Eucalyptus (Elastic Utility Computing Architecture for
Cloud middleware is an abstraction layer that hides Linking Your Programs To Useful Systems) uses existing
system complexity and enables seamless communication infrastructure to create a scalable, secure web services layer
mechanism between the cloud computing components that abstracts CPU, network and storage to provide IaaS.
[6,7,8]. Eucalyptus web services are designed for hybrid clouds. It is
The most important functionalities and characteristics of composed of several components that interact with one
Cloud middleware and related definition and benefits are another, enabling developers to modify existing modules.
presented in Table I. Eucalyptus is an open source implementation of Amazon
EC2 (Elastic Compute Cloud) and is compatible with most
business interfaces [9].
TABLE I. MIDDLEWARE FUNCTIONALITIES The OpenNebula is a toolkit to build private, public and
hybrid clouds. It has been designed to be integrated with
Middleware heterogeneous distributed data center infrastructures.
functionalities/ Definition Benefits OpenNebula orchestrates storage, network and virtualization
Characteristics technologies to deploy multi-tier services, such as groups of
Data storing the same data in - improving performance interconnected virtual machines, on distributed
replication multiple devices - increasing availability infrastructures, combining both data center resources and
remote cloud computing resources, according to allocation
- Managing the job
Scheduling Resource allocation and
queue for a cluster
policies [10].
jobs job execution Nimbus is known as a cloud computing solution
- Run-time optimization
providing IaaS. It consists on a set of software to be installed
A transparent takeover of a to one service node and a separate piece of software to be
- system recovery after
Fault tolerance failed machine by another
with continuity of service
outages installed on any number of virtual machine monitor (VMM)
nodes. Nimbus provides an implementation of Amazon's
Load balancing
Optimizing job distribution - Increased availability Elastic Compute Cloud (EC2) that allows using clients
on multiple computers. and data accesses developed for the real EC2 system against Nimbus based
clouds. It also allows users to create auto-configuring
Support interoperation - Easy access to more
Interoperability among different Cloud external services clusters. The solution provided enables users to even store
services - Easy data exchanges Virtual Machines (VMs) with no private credentials on-
board [11].
Support to security - Protection against
Security (passwords, SSH, X509 intruders V. ESTABLISHMENT OF METRICS FOR MIDDLEWARE
certificates, LDAP, etc.) - Securing data exchange
COMPARISON
Image disk Storing, transferring and
- Avoids loss of data
On the basis of the above listed functionalities (Table I),
management cloning image disk we defined some conceptual metrics to compare cloud
Enabling viewing middleware, and we assigned a weight to each of them based
Resources - Controlling the on its perceived importance from experience. The metrics are
information about the use
monitoring resource usage
of resources in real time the following:
• Virtualization management support: working in a
Creation of virtual versions
Virtualization
of the system, such as OS,
- Enabling resource scalable manner under XEN, KVME and VMwre
management sharing and reliability hypervisors.
storage device or networks.
• Scheduling: support for round robin, first-fit, using
adding, deleting or - Controlling user batch scheduler such as Haizea, PBS and SGE, and
User
modifying user’s account access to cloud services
management
and authorizations - Managing privileges
enabling integration of new algorithms.
• Security standards: support of security technologies
Control panel
User interface used to
- Easy manipulation
such as VPN tunnel, SSH authentication and ACL.
execute tasks • Cloud interfaces: use of standardized interfaces for
public cloud such as EC2 and S3.
• Control panel: managing and administering
Middleware performance criteria are not considered here resources with user-friendly and rich interfaces,
and will be treated in future work that will include including command line and web interface, support
experimental benchmarking. for users’ management and accounting.
Each metric is evaluated based on the number of
IV. CLOUD COMPUTING MIDDLEWARE technologies included. Accordingly, a metric’s score is equal
Nowadays, there are many commercial and open source to 1 if all the above listed solutions are available and
middleware. This paper focuses on the comparison between provided.
three popular open source solutions: Nimbus, Eucalyptus and
OpenNebula.

691
 The comparison of the three open source cloud gives a final score, based on the scores found in the previous
computing middleware, based on the five metrics is evaluation “Fig. 2”, and the metrics’ weights proposed in
represented in “Fig. 2”. Table II.
“Fig. 3” shows a comparison between the three open
source cloud computing middleware based on the formula
(1).

Figure 2. Middleware comparison

From “Fig. 2”, it can figure out that the three cloud
middleware are different but there are some common
Figure 3. Representation of middleware’s final score
between them. For example all provide the same Control
panel technologies. However, there are also some According to “Fig. 3”, OpenNebula obtained the highest
differences, for instance in the way each middleware score, followed by Nimbus and Eucalyptus. This means that
manages the virtualization and security. OpenNebula provides to users the most important and
According to the cloud architecture and above relevant cloud technologies related to Virtualization,
functionalities, we assigned weights to metrics as shown in Scheduling, Security and Cloud interfaces and panel,
Table II. comparing to the two other middleware.
TABLE II. METRICS’ WEIGHTS This classification was made according to the conceptual
metrics.
Metric weight VI. CONCLUSION AND FUTURE WORK
In order to evaluate different middleware solutions, a set
Virtualization management support 0.22
of functionalities, metrics and related weights were
presented. Three open source middleware were compared.
Scheduling algorithms support 0.28
In spite of its limited scheduling options, OpenNebula seems
to consistently compare favorably with the other paradigms.
Security standards 0.28
All of the three paradigms provide acceptable basic
functionalities, however.
Cloud interfaces support 0.05
This conceptual study will be completed by a future
experimental work with extensive benchmarking and
Control panel 0.17
whereby metrics will capture among other variables, average
turn-around, response-time, throughput, and CPU and
Among the five metrics, we consider Scheduling network utilization [12].
algorithms support and Security standards the most relevant
ones, followed by Virtualization management support. The
Control interfaces support is ranked last, and is preceded by ACKNOWLEDGMENT
Control panel. The total of weights is equal to 1. The authors are grateful to the Fulbright Program for
The following formula: enabling this study to be undertaken by Professors and
graduate students from The George Washington University
in the United States and Abdelmalek Essaadi University in
Morocco.

692
 REFERENCES
[1] F. Feldhaus, S. Freitag and C. El Amrani, “State-of-the-Art
Technologies for Large-Scale Computing”, Ch. 1, pp. 1-17, in
Werner Dubitzky, Krzysztof Kurowski and Bernhard Schott, Large-
Scale Computing Techniques for Complex System Simulations,
Wiley-IEEE Computer Society Pr, 2011.
[2] L. Wang, M. Kunze, J. Tao, G. von Laszewski, “Towards building a
cloud for scientific applications”, Advances in Engineering Software,
Vol 42, pp. 714–722, 2011.
[3] G. Reese, “Cloud Application Architectures: Building Applications
and Infrastructure in the Cloud”, O'Reilly Media, 2009.
[4] T. Velte, A. Velte, R. Elsenpeter, “Cloud Computing, A Practical
Approach”, McGraw-Hill Osborne Media, 2009.
[5] J. W. Rittinghouse and J. F. Ransome, “Cloud Computing
Implementation, Management, and Security”, CRC Press, 2009.
[6] A. Nathani, S. Chaudharya, G. Somani, “Policy based resource
allocation in IaaS cloud”, Future Generation Computer Systems, Vol
28, pp 94–103, 2012.
[7] J. Gallard, A. Lebre, C. Morina, T. Naughtonb, S. L. Scott, G. Vallee,
“Architecture for the next generation system management tools”,
Future Generation Computer Systems, Vol 28, pp 136–146, 2012.
[8] T. Mather, S. Kumaraswamy, S. Latif, “Cloud Security and Privacy:
An Enterprise Perspective on Risks and Compliance (Theory in
Practice)”, O'Reilly Media, 2009.
[9] http://www.eucalyptus.com/
[10] http://opennebula.org/
[11] http://www.nimbusproject.org/
[12] T. El-Ghazawi1, K. Gaj, N. Alexandridis, F. Vroman, N. Nguyen, J.
R. Radzikowski, P. Samipagdi, S. A. Suboh, “A performance study of
job management systems”, Concurrency and Computation: Practice
and Experience,Vol 16, Issue 13, pp. 1229–1246, November 2004.

693