Академический Документы
Профессиональный Документы
Культура Документы
doi:10.2498 /cit.1001391
Mladen A. Vouk
Department of Computer Science, North Carolina State University, Raleigh, North Carolina, USA
“Cloud” computing – a relatively recent term, builds on based on Virtual Computing Laboratory (VCL)
decades of research in virtualization, distributed com- technology. VCL has been in production use at
puting, utility computing, and more recently networking,
web and software services. It implies a service oriented NC State University since 2004, and is a suitable
architecture, reduced information technology overhead vehicle for dynamic implementation of almost
for the end-user, great flexibility, reduced total cost of any current “cloud” computing solution. Sec-
ownership, on-demand services and many other things.
This paper discusses the concept of “cloud” computing, tion 4 discusses “cloud”-related research and
some of the issues it tries to address, related research engineering challenges. Section 5 summarizes
topics, and a “cloud” implementation available today. and concludes the paper.
Keywords: “cloud” computing, virtual computing lab,
virtualization, utility computing, end-to-end quality of
service 2. Cloud Computing
also increases efficiency, quality, and reliability services. It is expected that in the next 10 years,
by capturing commonalities among application service-based solutions will be a major vehicle
needs, and facilitates the efficient sharing of for delivery of information and other IT-assisted
equipment and services.”[5] functions at both individual and organizational
Today, almost any business or major activity levels, e.g., software applications, web-based
uses, or relies in some form, on IT and IT ser- services, personal and business “desktop” com-
vices. These services need to be enabling and puting, high-performance computing.
appliance-like, and there must be an economy-
of-scale for the total-cost-of-ownership to be
better than it would be without cyberinfras- 2.2.2. Components
tructure. Technology needs to improve end-
user productivity and reduce technology-driven The key to a SOA framework that supports
overhead. For example, unless IT is the primary workflows is componentization of its services,
business of an organization, less than 20% of an ability to support a range of couplings among
its efforts not directly connected to its primary workflow building blocks, fault-tolerance in its
business should have to do with IT overhead, data- and process-aware service-based delivery,
even though 80% of its business might be con- and an ability to audit processes, data and re-
ducted using electronic means. sults, i.e., collect and use provenance informa-
tion.
2.3.1. Developers
delivery of these new services (e.g., an instruc- server, and a visualization application server, or
tor in an educational institution, with “images” a high-performance cluster. Workflow image
being cloud-based in-lab virtual desktops), they is typically a process control image that also
may oversee the usage of the services, and may has a temporal component. It can launch any
collect and manage service usage information, number of the previous resources as needed and
statistics, etc. This may require some expertise then manage their use and release based on an
in the construction of images and services, but, automated workflow.
for the most part, their work will focus on in-
terfacing with end-users and on provisioning of Users of images that load onto undifferentiated
what end-users need in their workflows. resources can be given root or administrative
Their expertise may range from workflow au- access rights since those resources are “wiped
tomation through a variety of tools and lan- clean” after their use. On the other hand, re-
guages, to domain expertise needed to under- sources that provide access to only some of
stand what aggregates of services, if any, the its virtual partitions, may allow non-root cloud
end-user needs, to management of end-user ac- users only: for example, a z-Series mainframe
counting needs, and to worrying about inter-, may offer one of its LPARS as a resource. Sim-
intra- and extra-cloud service orchestration and ilarly an ESX-loaded platform may be non-root
engagement, to provenance data analysis. access, while its guest operating system images
may be of root-access type.
Some of the components that an integration and
provisioning expert may need are illustrated
in Figure 3, based on the VCL implementa-
tion [6, 44]. The need may range from “bare 2.3.4. End-users
metal” loaded images, images on virtual plat-
forms (on hypervisors), to collections of im-
age aggregates (environments, including high- End-users of services are the most important
performance computing clusters), images with users. They require appropriately reliable and
some restrictions, and workflow-based services. timely service delivery, easy-to-use interfaces,
A service management node may use resources collaborative support, information about their
that can be reloaded at will to differentiate them services, etc. The distribution of services, across
with images of choice. After they have been the network and across resources, will depend
used, these resources are returned to an undiffer- on the task complexity, desired schedules and
entiated state for re-use. In an educational con- resource constraints. Solutions should not rule
text, this could be, for example, a VMWare im- out use of any network type (wire, optical,
age of 10 lab-class desktops that may be needed wireless) or access mode (high speed and low
between 2 and 3 pm on Monday. Then after speed). However, VCL has set a lower bound on
3pm another set of images can be loaded into the end-to-end connectivity throughput, roughly
those resources. at the level of DSL and cable modem speeds. At
On the other hand, an “Environment” could be a any point in time, users’ work must be secure
collection of images loaded on one or more plat- and protected from data losses and unauthorized
forms. For example, a web server, a database access.
For example, the resource needs of educational sources. There are VCL pilots with a number of
end-users (Figure 4) may range from single- University of North Carolina campuses, North
seat desktops (“computer images”) that may Carolina Community College System, as well
deliver any operating system and application ap- as with a number of out-of-state universities –
propriate to the educational domain, to a group many of which are members of the IBM Virtual
of lab or classroom seats for support of syn- Computing Initiative”.
chronous or asynchronous learning or hands-
on sessions, one or more servers supporting Figure 5 illustrates NC State Cloud based on
different educational functions, groups of cou- VCL technology. Access to NC State Cloud
pled servers (or environments), e.g., an Apache reservations and management is either through
server, a database server, and a workflow man- a web portal, or through an API. Authentication,
agement server all working together to sup- resource availability, image and other informa-
port a particular class, or research clusters, and tion are kept in a database. Resources (real
high-performance computing clusters. Figure 4 and virtual) are controlled by one or more man-
shows the current basic services (resources) de- agement nodes. These nodes can be within the
livered by VCL. The duration of resource own- same cloud, or among different clouds, and they
ership by the end-users may range from a few allow extensive sharing of the resources pro-
hours, to several weeks, a semester, or an open- vided licensing and other constraints are hon-
ended period of time. ored. NC State undifferentiated resources are
currently about 1000 IBM BladeCenter blades.
About 40% to 50% of them are serving high-
3. An Implementation performance computing needs, the rest are in the
individual seat mode. Its differentiated services
“Virtual Computing Laboratory (VCL) – are teaching lab computers that are adopted into
http://vcl.ncsu.edu is an award-winning VCL when they are not in use (e.g., at night).
open source implementation of a secure produc- In addition, VCL can attach other differenti-
tion-level on-demand utility computing and ser- ated and undifferentiated resources such as Sun
vices oriented technology for wide-area access blades, Dell clusters, and similar. More detailed
to solutions based on virtualized resources, in- information about VCL user services, functions,
cluding computational, storage and software re- security and concepts can be found in [6, 44].
[15] The Grid: Blueprint for a New Computing Infras- [29] E. NAONE, “Computer in the Cloud”, Tech-
tructure, 2nd Edition, Morgan Kaufmann, 2004. nology, Review, MIT, Sept 18, 2007,
ISBN: 1-55860-933-4. http://www.technologyreview.com/printer
friendly article.aspx?id=19397
[16] D. GEORGAKOPOULOS, M. HORNICK, AND A.
SHETH, “An Overview of Workflow Management:
From Process Modeling to Workflow Automation [30] THE NETAPP KILO-CLIENT: http://partners.
Infrastructure”, Distributed and Parallel Databases, netapp.com/go/techontap/tot-march2006/
Vol. 3(2), April 1995. 0306tot kilo.html, March 2006.
[17] GLOBUS: http://www.globus.org/, accessed [31] NEWCOMER, ERIC; LOMOW, GREG, “Understanding
May 2008. SOA with Web Services”, Addison Wesley, ISBN
[18] HADOOP: http://hadoop.apache.org/core/, 0-321-18086-0, 2005.Ö
accessed May 2008.
[32] CORBA, http://www.omg.org/corba/
[19] ELIAS N. HOUSTIS, JOHN R. RICE, EFSTRATIOS GAL-
LOPOULOS, RANDALL BRAMLEY (EDITORS), En- [33] http://www.service-architecture.com/
abling Technologies for Computational Sci- web-services/articles/object management
ence Frameworks, Middleware and Environments, group omg.html
Kluwer-Academic Publishers, Hardbound, ISBN
0-7923-7809-1, 2000.
[34] PARTIMAGE, http://www.partimage.org/Main
[20] M. HSU (ED.), “Special Issue on Workflow and Page, accessed Dec 2008.
Extended Transaction Systems”, IEEE Data Engi-
neering, Vol. 16(2), June 1993. [35] J. REIMER, “Dreaming in the “Cloud” with the
[21] IBM, “IBM Launches New System x Servers and XIOS web operating system”, April, 8, 2007,
Software Targeting Large Scale x86 Virtualization, http://arstechnica.com/news.ars/post/
http://www-03.ibm.com/press/us/en/ 20070408-dreaming-in-the-cloud-with-the
pressrelease/19545.wss, 21 Apr 2006. -xios-web-operating-system.html
[22] IBM, “Google and IBM Announced University Ini- [36] M. P. SINGH AND M. A. VOUK, “Network Comput-
tiative to Address Internet-Scale Computing Chal- ing”, in John G. Webster (editor), Encyclopedia of
lenges”, October 8, 2007, http://www-03.ibm. Electrical and Electronics Engineering, John Wiley
com/press/us/en/pressrelease/22414.wss & Sons, New York, Vol. 14, pp. 114–132, 1999.
[23] IBM, “IBM Introduces Ready-to-Use Cloud Com-
puting”, http://www-03.ibm.com/press/us/ [37] Secure Open Systems Initiative, http://www.
en/pressrelease/22613.wss, November 15, sosi.ncsu.edu/, accessed Dec 2008.
2007.
[38] THOMAS ERL, “Service-oriented Architecture: Con-
[24] IBM, “North Carolina State University and IBM cepts, Technology, and Design”, Upper Saddle
help bridge digital divide in North Carolina and River: Prentice Hall PTR. ISBN 0-13-185858-0,
beyond”, May 7, 2007, http://www-03.ibm. 2005.
com/industries/education/doc/content/
news/pressrelease/2494970110.html
[39] MARK VANDERWIELE, “The IBM Research Cloud
[25] IBM, “Service-oriented Architecture – SOA”, Computing Initiative”, Keynote talk at ICVCI 2008,
http://www-306.ibm.com/software/ RTP, NC, USA, 15–16 May 2008.
solutions/soa/, accessed May 2008.
[40] VIRTUAL COMPUTING LABORATORY, VCL, http:
[26] IBM, “Mirage: Virtual Machine Images as Data”, //vcl.ncsu.edu, on-line since Summer 2004.
http://domino.research.ibm.com/comm/
research projects.nsf/pages/mirage.
index.html, 6-Feb-08. [41] VMWARE, “The Open Virtual Machine Format
– Whitepaper for OVF Specification”, v0.9, 2007,
[27] S. LOHR, “Google and I.B.M. Join in http://www.vmware.com/pdf/ovf whitepaper
‘Cloud Computing’ Research”, October 8, specification.pdf
2007, http://www.nytimes.com/2007/10/08/
technology/08cloud.html? r=1&ei=5088&en= [42] M. A. VOUK AND M. P. SINGH, “Quality of Service
92a8c77c354521ba&ex=1349582400&oref= and Scientific Workflows”, in The Quality of Nu-
slogin&partner=rssnyt&emc=rss& merical Software: Assessment and Enhancements,
pagewanted=print editor: R. Boisvert, Chapman & Hall, pp.77–89,
[28] B. LUDSCHER, I. ALTINTAS, C. BERKLEY, D. HIG- 1997.
GINS, E. JAEGER-FRANK, M. JONES, E. LEE, J. TAO,
Y. ZHAO, “Scientific Workflow Management and [43] M. VOUK, R. L. KLEVANS AND D. L. BITZER, “Work-
the Kepler System”, Concurrency and Computa- flow and End-user Quality of Service Issues in
tion: Practice & Experience, Special Issue on Web-based Education”, IEEE Trans. On Know-
Workflow in Grid Systems, Volume 18, Issue 10 ledge Engineering, to Vol 11(4), July/August 1999,
(August 2006), pp. 1039–1065, 2006. pp. 673–687.
246 Cloud Computing – Issues, Research and Implementations
Contact address:
Mladen A. Vouk
Department of Computer Science
North Carolina State University
Raleigh, NC 27695, USA
e-mail: vouk@ncsu.edu