CLOUD MANAGEMENT ARCHITECTURE IN NGN/NGS CONTEXT

QoS-awareness, Location-awareness and Service Personalization

Rachad Nassar and No¨emie Simoni

TELECOM ParisTech - LTCI - UMR 5141 CNRS, 46, rue Barrault F 75634, Paris Cedex 13, France

Keywords:

Cloud computing, NGN/NGS, E2E QoS-awareness, Location-awareness, Service personalization, SOA.

Abstract:

Cloud computing has become one of today’s hot topics. The major contribution of this Internet-based service

delivery paradigm consists in offering computing, storage and network resources able to guarantee information

technology externalization. In parallel to this novel trend, cloud users requirements are quickly emerging due

to both network and service convergence. Therefore, beyond its externalization solution, cloud must also re-

spond to users needs within this “Next Generation Networks/Next Generation Services” (NGN/NGS) context.

Hence, it should offer service personalization for cloud users, take into consideration their mobile context,

and guarantee an end-to-end QoS. In this paper, we propose a QoS-based cloud management architecture that

overcomes the aforementioned challenges through several mechanisms. First, we surpass mobility and E2E

QoS challenges by gathering ubiquitous elements into ubiquity-based virtual communities. Second, we ensure

service personalization by proposing a seamless and dynamic service composition based on stateless services.

Finally we take into consideration user’s ambient context by using location-based virtual communities. Com-

puting models for QoS-aware and location-aware clouds are also provided.

1 INTRODUCTION

Nowadays, cloud computing is a new buzzword

around the globe. Many different deﬁnitions has

been given for this concept. In this paper, we have

chosen the one proposed by Prof. Ian Foster (Fos-

ter et al., 2008), since it combines the main key as-

pects that distinguish clouds from previous distributed

computing paradigms. According to this deﬁnition,

cloud computing is “a large-scale distributed comput-

ing paradigm that is driven by economies of scale, in

which a pool of abstracted, virtualized, dynamically-

scalable, managed computing power, storage, plat-

forms, and services are delivered on demand to exter-

nal customers over the Internet”. Based on this deﬁni-

tion, cloud computing is considered as a vast quilt on

which externalized services are running. These ser-

vices arise from heterogeneous domains (Telco, Web

and IT) and are executed separately. Hence, we be-

lieve that the success of cloud computing should not

be limited to externalization, but it should also be re-

lated to service convergence which is the heart of our

Next Generation Services (NGS) context. The lat-

ter consists in seamlessly composing global Cloud

Services (CSs) by using stateless service elements.

Cloud Service Providers (CSPs) are then able to dy-

namically create and offer users any desired CS.

Moreover, based on this same deﬁnition, we

strongly believe that clouds success also depends on

the ability of CSPs to dynamically maintain and man-

age their resources in order to best answer the users

requirements. However, with the rapid evolution of

Next Generation Networks (NGN) and its network

convergence aspect, users needs are quickly emerg-

ing. In fact, within the NGN context, all access net-

works convergeinto one IP core network. Cloud users

are then able to access different CSs that are provided

by different CSPs, while using different equipment

technologies and access points. Thus, cloud users be-

come more nomadic and want to access their services

anywhere, anytime and anyhow. Therefore, to attract

more users, CSPs should guarantee an End-to-End

(E2E) Quality of Service (QoS) by taking into con-

sideration real-time users preferences and contexts.

In essence, users can have functional and non-

functional preferences. The former are characterized

by personalized CSs that dynamically and transpar-

ently adapt themselves to users needs without session

interruption; the latter are characterized by QoS re-

quirements that appear in Service Level Agreements

(SLAs) signed between CSPs and cloud users. In fact,

this user-centric approach that is based on service per-

629

Nassar R. and Simoni N..

CLOUD MANAGEMENT ARCHITECTURE IN NGN/NGS CONTEXT - QoS-awareness, Location-awareness and Service Personalization.

DOI: 10.5220/0003390106290635

In Proceedings of the 1st International Conference on Cloud Computing and Services Science (CLOSER-2011), pages 629-635

ISBN: 978-989-8425-52-2

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

sonalization and QoS preferences is highly desirable

in future cloud environments since it increases the

Return On Investment (ROI) for CSPs on one hand

and allows to enrich users experiences on the other

hand. However, to the best of our knowledge, this re-

search area is not fully investigated within the context

of cloud computing.

Another main aspect that CSPs should take into

consideration is users ambient context. The latter can

be characterized by users locations or by their activ-

ities. In the scope of this paper, we treat the spatial

mobility aspect and the inﬂuence of users locations

on CSs provisioning. In fact, with the resource exter-

nalization trend into cloud systems, CSs sources are

often far away from end devices. Thus, the latency -

the time to send or receiveone byte of data - is directly

inﬂuenced, since it depends on the distance separating

a user from his CS. For instance, network latency of

few milliseconds becomes critical for highly interac-

tive services such as immersive applications, since it

can cause a noticeably degraded users experience. In

order to minimize latency, CSPs must offer a more

adapted scalable distribution of their resources. In

addition, they should connect cloud users, specially

those requiring highly immersive interactions, to the

closest CSs sources.

After having introduced the main aspects that mo-

tivate our work, we propose in the following of this

paper solutions to overcome the aforementioned chal-

lenges. The proposed solutions answer the follow-

ing questions: how can CSPs dynamically and trans-

parently guarantee personalized services for users?

how can they maintain, in real-time, the required E2E

QoS? and how can they deliver CSs while taking into

consideration users locations?

The remainder of this paper is organized as fol-

lows. In Section 2, we discuss some of the related

works. In Section 3, we propose a QoS model, an

ubiquity-based virtual community concept and a dy-

namic service composition, in order to ensure QoS-

aware clouds and personalized CSs. In Section 4, we

allow CSPs to provide location-based CSs by propos-

ing a location-based virtual community concept. The

conceptual QoS and location-aware cloud architec-

ture is presented in Section 5. Finally, conclusion and

future perspectives are presented in Section 6.

2 RELATED WORK

Service personalization, E2E QoS management and

location-awareness are highly important research top-

ics for future cloud environments. In this section, we

discuss some of the investigations in these aspects.

In the context of Personalized Services, the au-

thors in (Guo et al., 2009) focus on how to pro-

vide personalized services for users by using a client

side program that records user activities and computes

models on personal devices. Afterwards, selected

models are uploaded to service providers so they can

have better knowledge of that user and consequently

provide him personalized services. The drawback of

this approach lies in the fact that pushing all mod-

els and computing programs to clients side contradicts

the cloud computing externalization concept. More-

over, this previously mentioned paper investigates the

commonly known description of service personaliza-

tion as the ability of service providers to offer, or

rather impose adequate services to users, while be-

ing based on their analytical records. In opposite to

the previous description, our service personalization

approach is based on a dynamic service composition

where users can impose their own preferences.

Some of the papers found in the literature reveal

a growing interest in the topic of QoS-awareness in

cloud environments. In (Spillner and Schill, 2009),

the authors present, according to the service run-time

behavior, a technique to dynamically adjust SLA con-

straints after adjusting the values of service descrip-

tion’s nun-functional properties. In addition, QoS-

awareness is also discussed in (Korn et al., 2009)

where SLA monitoring is delegated to a third in-

dependent party, namely Service Level Management

Authority (SLMA). In opposite to these approaches,

our proposition overcomes QoS violation without

modifying any SLA parameter. Moreover, we pro-

pose auto-managed QoS-aware CSs.

Finally, to the best of our knowledge, the topic of

location-aware clouds has not yet received much at-

tention. Moreover, research communities who treat

this subject, propose externalized services that pro-

vide location-based information. For instance, the

authors in (Wang and Yang, 2009) propose a mo-

bile information retrieve system based on GPS and

Web2.0 applications. Through this platform, users

can obtain location-based information and personal-

ized recommendations. In our proposition, we guar-

antee location-based information for users, but we

also apply the location-awareness concept on CSs. In

this context, based on their users locations, CSPs of-

fer distributed CSs. Hence, we overcome the latency

problem by allowing users to access the closest CS.

CLOSER 2011 - International Conference on Cloud Computing and Services Science

630

3 QoS-AWARE CLOUD

MANAGEMENT

In this section, we introduce in the ﬁrst subsection our

QoS model according to which we gather CSs into

Ubiquity-based Virtual Communities. The latter con-

cept is introduced in the second subsection. In addi-

tion, a QoS agent is integrated into each CS in order

to ensure an autonomous management of these com-

munities. All these concepts are merged into a QoS-

aware computing model that is presented along with a

QoS management architecture in the last subsection.

3.1 QoS Model

Nowadays, industrial providers are offering different

types of Infrastructures as a Service (IaaS), Platforms

as a Service (PaaS) and Software as a Service (SaaS).

Therefore, heterogeneous CSs have been provided

for cloud users. In parallel, the heterogeneity aspect

also appears on the users access level, since they can

choose among different types of terminals and access

points. Consequently, if CSPs want to provide CSs

with a respected E2E QoS, they must ﬁrst overcome

the heterogeneity challenge. For this purpose, we pro-

pose to provide an homogenous vision for all existing

cloud elements by introducing a new QoS model. The

latter is based on four criteria that are deﬁned as fol-

low:

• Availability. It represents the ability of a CS to be

accessed at a certain time. It indicates the ratio of

accessibility for a CS.

• Reliability. It represents the ability of a CS to

be used without deteriorating the information and

while respecting contract conditions. It indicates

the percentage of unintentional modiﬁcations in

the information caused by the CS.

• Delay. It indicates the duration of a request treat-

ment by a CS.

• Capacity. It reﬂects the ability of a CS to treat

normally a request while using all the possible

treatment means. It indicates the charge rate of

the CS.

We note that there is no need to combine these

four criteria into one formulation, since each variation

in one of the CS QoS characteristics will be solved by

passing the request into a ubiquitous element as we

will see in the next subsection.

3.2 Ubiquity-based Virtual

Communities: VXCU

Due to the NGN evolution, cloud users are able to

access different CSs while using different terminals,

access points and networks. In addition, they be-

come more nomadic and aim to conserve their contin-

uous session despite their situations. To reach these

goals and to guarantee an E2E QoS for their users,

CSPs should overcome the mobility challenge con-

sisting of four mobility types (Guo et al., 2008): ter-

minal mobility, user mobility, network mobility and

service mobility. In this subsection, we surpass the

aforementioned problem by gathering ubiquitous CSs

into Ubiquity-based Virtual Communities (VXCU);

X={S,C,E} which represents different types of CSs:

applicative Services (S), abstracted Connectivity net-

works (C), and Equipment (E) such as servers and

platforms. We note that in our context, ubiquitous

CSs correspond to those having the same functional-

ity and an equivalent QoS. Therefore, throughout end-

users situations, we manage the continuity of their

sessions by provisioning VXCUs that respond to their

functional and nonfunctional (QoS) preferences.

In many cases, specially in mobile context, a CS

that is used in a user session might not continue to

function normally or to fulﬁll the user QoS require-

ments. For this reason, we solve the problem by us-

ing the corresponding provisioned VXCU. In fact, we

dynamically replace the current CS by an ubiquitous

counterpart that belongs to the same VXCU. There-

fore, we seamlessly adapt the user’s session against

any degradation and we guarantee the continuity of

this session while maintaining the E2E required QoS.

In the following, we divide the VXCU management

process into two phases: the creation and the exploita-

tion phases.

In the creation phase, each CS joins an existing

VXCU or creates a new one. The creation process is

based on three basic services:

• Ubiquity Inquiry Service: determines the func-

tionality and QoS criteria of a CS while having

its ID as input.

• Discovery Service: launches a search in order to

discover CSs that verify the criteria indicated at

the input. In the ubiquity case, a functionality and

QoS criteria are considered as inputs.

• Presence Service: ﬁlters an obtained list accord-

ing to CSs states. It selects among a list of CSs

IDs the ones that are “Available” (Accessible by

the end-user), “Activable” (Activated by the CSP

but not yet used) and “Activated” (Activable and

used).

CLOUD MANAGEMENT ARCHITECTURE IN NGN/NGS CONTEXT - QoS-awareness, Location-awareness and

Service Personalization

631

The creation process is launched during each CS

deployment. First, the Ubiquity Inquiry Service is in-

voked in order to get the functionality and the QoS

criteria of this CS. Second, the DiscoveryService uses

the former output as searching criteria. Thus, we ob-

tain a list of CSs having the same functionality and

QoS as the CS in question. Finally, the Presence Ser-

vice applies its ﬁltering process on the obtained list.

In consequence, the VXCU verifying the functional-

ity and QoS of this CS is determined. Thus, the latter

joins this community. We note that if no existing VX-

CUs verify the required QoS and functionality crite-

ria, the CS in question creates a new VXCU having

its own characteristics.

After explaining the creation process, we present

the exploitation phase, where the CS is already a

member of a VXCU. We mention that the VXCU

concept is based on a Peer-to-Peer (P2P) self commu-

nity management process in which each CS acts as an

auto-managed peer. For this purpose, a QoS-Agent is

integrated into each CS. During this dynamic VXCU

auto-management process, each CS compares its cur-

rent QoS and functionality with the ones of its current

VXCU. If the result is positive, then the QoS-Agent

sends a notiﬁcation called “IN Contract” in order to

inform the other VXCU members that the community

contract is still respected. If not, the QoS-Agent sends

a notiﬁcation called “OUT Contract”. In this case, the

CS leaves the VXCU and ﬁnds another one.

3.3 Computing Model for QoS-aware

Clouds

In this subsection, we explain how the aforemen-

tioned QoS management model and concepts are able

to dynamically and seamlessly adapt users sessions

against any variation in users preferences. For this

purpose, we propose a solution that allows cloud users

to create their own personalized sessions by combin-

ing different CSs provided by different CSPs.

Our personalization vision is based on a dynamic

seamless composition that is provisioned and tailored

according to users preferences. To reach this goal,

a new model is introduced, namely the Virtual Pri-

vate Network (VPXN) (Guo et al., 2008); X={S,C,E}

since for each cloud user we provision VPXNs on the

applicative Services layer (VPSN), Connectivity net-

works layer (VPCN) and Equipment layer (VPEN). In

fact, these personalized VPXNs gather CSs that best

answer the cloud user’s preferences. Consequently,

the user’s session is dynamically created by combin-

ing different CSs from these VPXNs. To support this

dynamic composition, management services are in-

troduced in (Nassar and Simoni, 2010):

Old VPXN

VXCU

QoS-aware

CLOUD

CS11

QoS agent

CS23

QoS agent

CS31

QoS agent

CS32

QoS agent

CS33

QoS agent

VPXN

Reconfiguration

Invocation of VXCU

Exploitation Service

VXCU

members

Adapted CS

• CS Profile

• VXCUs Profiles

• User Preferences

• CS delivery process

Adapting

CSs

Adapting

Community

Monitoring

Events

Figure 1: Computing Model for QoS-aware Clouds.

• Service Logic: represents the service composition

workﬂow. By analogy, there are Network Logic

and Equipment Logic.

• Semantic Routing: routes user’s request to the

next SE in the Service Logic. To support this

function-based routing, a QoS Routing Table is

used. By analogy, on the network and equipment

layers, there are the well-known routing mecha-

nisms (e.g. MPLS, OSPF, etc.) and handover

techniques.

• Service Selection: selects the chosen SE. By anal-

ogy to this service, we introduce the Network and

Equipment Selection services.

• VPXN Conﬁguration: modiﬁes and conﬁgures

cloud users VPXNs.

In addition to these management services, we in-

troduce in this paper a QoS Policy Agent that veriﬁes

the integrity of the independent modiﬁcations that are

caused, on each layer, by each of the aforementioned

services. It is a decision table that monitors E2E QoS

and is considered as a part of the AmbientGrid (Si-

moni et al., 2008). The latter is an information infer-

ence driven by the proﬁles deﬁned in the knowledge

base, namely “Infoware” (Simoni et al., 2008). It dy-

namically manages ambient resources in a personal-

ized way.

In our proposition, we favor an event-based ap-

proach. For this reason, this QoS management archi-

tecture is associated with a computing model that an-

alyzes the QoS-based event and consequently adapts

the VPXN conﬁguration. As shown in Figure 1, a

provisioned CS in the user’s Old-VPXN (e.g. CS31)

could not maintain the required QoS. Its QoS-Agent

detects this QoS degradation and notiﬁes the cloud

management system. The latter ﬁrst monitors the

event. Since it is a QoS degradation, the VXCU Ex-

ploitation service is invoked, and the “Adapting Com-

CLOSER 2011 - International Conference on Cloud Computing and Services Science

632

Geographic

Address 1

SE2

SE1

EE2

EE1

EE3

SE3

SE4

Geographic

Address 2

Old VXCL

Location-aware

CLOUD

NE1

NE2

NE3

Monitoring

Events

Adapting

Community

Ambient

Context

Reconfiguration

Invocation of VXCL

Exploitation Service

VXCLs

members

• User Profile • VXCLs Profiles

Figure 2: Computing Model for Location-aware Clouds.

munity” phase is launched. Based on the CS’s func-

tionality and QoS, this phase provides all possible

ubiquitous members. Thus, according to the afore-

mentioned CS delivery process (service composition

according to user preferences: CS logic, semantic

routing, etc.), the “Adapting CSs” phase chooses the

adequate counterpart (e.g. CS32) to replace the cur-

rent unwanted CS (CS31). Finally, the “VPXN Re-

conﬁguration” phase is launched and a New-VPXN is

then created. In consequence, this computing model

demonstrates how the VXCU concept and the per-

sonalized service composition process are used to

dynamically and seamlessly manage users sessions

while conserving their E2E QoS and taking their pref-

erences into consideration.

4 LOCATION-AWARE CLOUD

MANAGEMENT

In order to reduce latency inﬂuence on cloud users

experiences, CSPs are supposed to distribute their

resources according to their users speciﬁc locations

(home, work, hotels abroad, etc.). In this section,

we propose a management mechanism that allows

the provisioning of location-based CSs. Therefore,

throughout cloud users movements, we anticipate

their demands by providing the nearest resources.

Our location-awareness proposition considers the

perviously discussed virtual community concept. In

fact, we sense user’s location and use it to create and

manage Location-based Virtual Communities (VX-

CLs, with X={S,C,E}). The latter gather all CSs that

have the same location as the user. Consequently,

for each users location, three VXCLs are provisioned

(VSCL, VCCL and VECL). In the following, we di-

vide the VXCL management process into two phases:

the creation and the exploitation phases.

In order to create VXCLs, we use the following

basic services:

• Geo-Location Service: determines an element’s

geographic location while having its Element ID

as input.

• Discovery Service: previously deﬁned in subsec-

tion 3.2. However, in this case, the discovery is

based on the user’s location.

• Presence Service: previously deﬁned in subsec-

tion 3.2.

• Sorting By Type Service: sorts by type (applica-

tive Services, Connectivity networks, Equipment)

a list of Element IDs received as an input.

The creation process is executed in advance for

user’s speciﬁc locations. Hence, it allows to anticipate

user’s movements. First, the Geo-Location Service

determines user’s speciﬁc location. Second, the Dis-

covery Service uses the obtained result as its search-

ing criterion. Thus, we get a list of CSs that have this

same location. Third, the Presence Service applies

its ﬁltering process on this list. The obtained VXCLs

verify user’s location. At the end of this process, the

Sorting By Type Service ﬁlters the obtained VXCLs

according to CSs types. In summary, the aforemen-

tioned steps provision the VSCL, VCCL and VECL

corresponding to user’s speciﬁc location.

After explainingthe creation phase, we present the

exploitation phase during which the user is moving.

Instantly, we offer the user the three provisioned CSs

lists that correspond to his current location.

According to our event-based approach, we intro-

duce a computing model that analyzes the location-

based event and consequently adapts the user’s am-

bient context conﬁguration. As shown in Figure 2,

a user moves towards a new location (Geographic

Address2). Consequently, a notiﬁcation is sent to

the cloud management system. The latter ﬁrst mon-

itors the event. Since, it is a modiﬁcation in user’s

location, the VXCL Exploitation service is invoked

and the “Adapting Community” phase is launched.

This phase provides all the location-based commu-

nities (VSCL, VCCL, VECL) that are provisioned

to this user’s speciﬁc location. Finally, the “Ambi-

ent Context Reconﬁguration” phase is launched and

a new ambient context (New-VSCL, New-VCCL,

New-VECL) is then created. Hence, this comput-

ing model shows how the VXCL concept can be used

to dynamically and instantly enrich user’s experience

and context-awareness.

CLOUD MANAGEMENT ARCHITECTURE IN NGN/NGS CONTEXT - QoS-awareness, Location-awareness and

Service Personalization

633

VPX Configuration

Infoware

Service

Profile

CS5

VXCU&L

QoS & Location

aware CLOUD

CS4

CS2

CS3

CS13

CS12

CS5

CS15

CS14

Equipment

Logic

Network

Logic

Service

Logic

Equipment

Selection

Network

Selection

Service

Selection

E2E QoS Management

VXCU&L Management

VXCL Exploitation

Ubiquity

Inquiry Service

Discovery

Presence

QoS

Policy

Agent

Equipment

Routing

Network

Routing

Semantic

Routing

Profile

etwork

Profile

Equipment

Profile

User

Profile

VXC

Profile

VPX

Profile

Session

Profile

Real-time

Profile

VXCL Creation

VXCU Creation

VXCU Exploitation

Geo-Location

Figure 3: Cloud Management Architecture.

5 CLOUD MANAGEMENT

ARCHITECTURE

After introducing the novel QoS-aware and location-

aware cloud management concepts, we structure in

this section all the aforementioned mechanisms and

basic services into a cloud management architecture.

As previously stated in the proposed computing mod-

els, we adopt an event-based approach. In addition,

for an efﬁcient and ﬂexible usage of clouds, we adopt

a Service Oriented Architecture (SOA) approach. The

latter organizes reusable, autonomous and interopera-

ble services into the service layer. However, accord-

ing to our NGN/NGS context, we do not limit our

proposition to the aforementioned SOA characteris-

tics. In contrary, the added-values of our approach

consists in adding the following architectural aspects:

• Mutualization: It represents the ability of a ser-

vice to be not only reusable, but also shareable.

Consequently, depending on its QoS criteria, the

mutualized service is shared among several re-

quests. The untreated requests are stocked in a

queue that is associated to the service. If the

Timeout of a request is less than its waiting delay

in the queue, our service delivery system reroutes

the request towards a ubiquitous CS.

• Auto-Management: It represents the ability of a

service to be not only autonomous, but also auto-

managed. For this reason, we have introduced a

QoS-Agent as described in subsection 3.2.

• Interconnection: It represents the ability of a ser-

vice to be not only interoperable, but also in-

terconnected with others in order to dynamically

and seamlessly compose personalized services.

For this purpose, we have already presented the

VPXN concept in subsection 3.3. The latter in-

troduces “Virtualization” as another characteristic

of our architecture. It allows to transparently of-

fer services for the users and to let them use these

services as if they were dedicated to them.

The cloud management architecture is shown in

Figure 3. It manages the cloud while taking into

consideration all the NGN and NGS challenges.

For this purpose, this global structure gathers all

the previously proposed management mechanisms.

The “VXCL Creation” and “VXCL Exploitation”

parts support the user-centric approach by enriching

users context-awareness. The “VXCU Creation” and

“VXCU Exploitation” parts allow to overcome the

NGN mobility challenge by gathering ubiquitous CSs

into auto-managed communities. The VXCU con-

cept is combined with the “QoS Policy Agent” and

the “E2E QoS Management” parts in order to guaran-

tee an E2E QoS for cloud users. The “VPXN Con-

ﬁguration” part supports the NGS context by ensur-

ing personalized, dynamic and seamless CSs compo-

sitions. Moreover, a knowledge base is necessary in

order to provide all needed information. For this pur-

pose, we use the Infoware (Simoni et al., 2008) which

contains several XML proﬁles (for CSs, cloud users,

real-time, sessions, communities, etc.). It efﬁciently

and dynamically manages the decisional and reactive

information. It is not a simple data base like the Home

Subscriber Server (HSS) in the IP Multimedia Sub-

system (IMS) architecture. On the contrary, it is a

well structured knowledge base acting as a real-time

informational inference. It is worth noting that within

the project in which our work is situated, security as-

pects are also investigated by another research group.

Furthermore, as previously discussed, E2E QoS is

guaranteed by creating VXCUs that gather ubiquitous

CSs which are deployed all over the world. How-

ever, creating VXCUs with huge number of ubiqui-

tous CSs is not efﬁcient, since it makes the stored in-

formation more bulky and takes longer time to decide

the adequate substitute for the degraded CS in user’s

session. In order to minimize the time and informa-

tion clarity costs, we propose to ﬁlter ubiquitous CSs

by using speciﬁc locations. Consequently, we cre-

ate virtual communities that are simultaneously based

on ubiquity and location. In comparison with normal

VXCUs, the former communities are able to manage

clouds in more efﬁcient way. These communities are

named Ubiquity and Location based Virtual Commu-

nities (VXCU&L).

CLOSER 2011 - International Conference on Cloud Computing and Services Science

634

As shown in Figure 3, one possible technique

to create VXCU&Ls is to ﬁlter the VXCUs by

using users speciﬁc locations. For this purpose,

we combine the VXCU that contains ubiquitous

CSs with the VXCLs that are provisioned for

user’s speciﬁc location. In this way, we create

the VXCU&Ls that gather ubiquitous CSs found

around the user’s geographic location. An exam-

ple is given in Figure 3 where the VXCU contains

{CS11,CS12,CS13,CS14,CS15} and the VXCL con-

tains {CS11,CS12,CS13,CS2,CS3,CS4,CS5}. Con-

sequently, the VXCU&L is dynamically created and

contains {CS11,CS12,CS13}. Moreover,we consider

that CS11 is part of user’s session and it suffers from

a QoS degradation. Hence, we neglect CS14 and

CS15 and choose the substitute between CS12 and

CS13. This technique allows to react faster against

QoS degradation events and to manage more efﬁ-

ciently the E2E QoS.

6 CONCLUSIONS AND

PERSPECTIVES

Through its externalization aspect, cloud computing

provides a combination of cost and agility savings.

In fact, organizations prefer saving time and energy

on managing services that can rather be moved to

the cloud. However, externalizing services induces a

growing need for ways to manage and enhance cloud

environments, in order to best answer the cloud users

requirements. The latter, within an NGN/NGS con-

text, have become more mobile and more greedy.

They aim to access any type of CS anytime, any-

where and anyhow. Thus, an efﬁcient cloud environ-

ment that guarantees users preferences and needs is

required. For this reason and in order to overcome the

previous challenges, we have proposed in this paper

a cloud management architecture composed of inno-

vative mechanisms. In this ameliorated SOA archi-

tecture, we have favored service personalization over

monolithic applications. For this purpose, we have

introduced a seamless and dynamic service compo-

sition that combines stateless and mutualized service

elements. In fact, this CS composition takes into con-

sideration users preferences, such as the required E2E

QoS. In consequence, we have proposed the ubiquity-

based virtual communities concept which is based on

a QoS model. By gathering ubiquitous elements hav-

ing the same functionality and an equivalent QoS, we

have conserved the continuity of cloud users sessions

while guaranteeing their required E2E QoS. More-

over, in order to reduce the latency, we have provi-

sioned location-based virtual communities gathering

CSs close to users speciﬁc locations. Both ubiquity

and location based community concepts are supported

by event-basedcomputing models. Finally, for a more

efﬁcient management, we have proposed to ﬁlter the

VXCUs provisioning according to speciﬁc locations.

For this reason, we have created several VXCUs in

different locations in order to best manage the users

required E2E QoS.

However, for different users locations, different

VXCUs would be used to guarantee the users session

continuity. In our future work, we must investigate

the transition of the managementprocess between two

VXCUs when the user is moving. By analogy to the

handover on the access networks level, this future dis-

cussion subject is named “Semantic Handover”.

ACKNOWLEDGEMENTS

The authors would like to thank all the participants

in the UBIS project, ﬁnanced by the French ANR

VERSO 2008 in which is situated our work.

REFERENCES

Foster, I., Zhao, Y., Raicu, I., and Lu, S. (2008). Cloud

computing and grid computing 360-degree compared.

In GCE’08, Grid Computing Environments Workshop.

Guo, H., Chen, J., Wu, W., and Wang, W. (2009). Per-

sonalization as a service: the architecture and a case

study. In CloudDB’09, International Workshop on

Cloud Data Management.

Guo, W., Simoni, N., and Yin, C. (2008). Automated

management of user centric session in ngn. In GLO-

COMW’08, IEEE GLOBECOM Workshops.

Korn, A., Peltz, C., and Mowbray, M. (2009). A service

level management authority in the cloud. In HP Lab-

oratories Technical Report.

Nassar, R. and Simoni, N. (2010). Ngn/ngs components

for service personalization in a mobile and heteroge-

neous context. In IWUSP’10, International Workshop

on Ubiquitous Service Platforms.

Simoni, N., Yin, C., and Chene, G. D. (2008). An intelligent

user centric middleware for ngn: Infosphere and am-

bientgrid. In COMSWARE’08, International Confer-

ence on COMmunication System softWAre and Mid-

dlewaRE.

Spillner, J. and Schill, A. (2009). Dynamic sla template

adjustments based on service property monitoring. In

CLOUD’09, IEEE International Conference on Cloud

Computing.

Wang, Z. and Yang, F. (2009). A multiple-mode mo-

bile location-based information retieve system. In

ICWMC’09, International Conference on Wireless

and Mobile Communications.

CLOUD MANAGEMENT ARCHITECTURE IN NGN/NGS CONTEXT - QoS-awareness, Location-awareness and

Service Personalization

635