A FRAMEWORK FOR THE DYNAMIC CONFIGURATION OF

ADAPTIVE TRANSPORT PROTOCOLS

Application to QoS Requirements

F. Armando

, N. Van Wambeke

, C. Chassot

1,2

, E. Exposito

1,2

and K. Drira

LAAS-CNRS,

Université de Toulouse / INSA – Toulouse – France

Keywords: Dynamically configurable transport protocol, Policy-based management, QoS provisioning, model-based

decision.

Abstract: Self-adaptation of communication protocols is a major issue in the conception of future QoS-oriented

services for the ambient Internet. Our approach is based on behavioural and architectural adaptation

properties of dynamically configurable Transport protocols. This paper proposes an architecture for the QoS

provisioning at the Transport level. To fulfil this provisioning, the decision process follows a policy-based

framework, using different external models in order to have an extensible design. We illustrate the use of

this framework in a case study for the QoS optimization of a mobile user, roaming from a wired network to

a wireless network.

1 INTRODUCTION

Recent advances in computing and wireless

networking technologies allow considering the

deployment of complex wireless, mobile and

cooperative applications within a fully pervasive,

mobile and heterogeneous Internet. For instance,

military emergency operation management systems

are a typical example of such applications. They

support mobile users cooperating applications for

crisis management in linked to variable

communication resources that change depending on

the investigated field, the role of the participants in

the operation, etc.

From a communication point of view, multiple user

and application time-varying requirements have to

be satisfied. They depend on the communication

tools used by the participants (e.g. interactive

audio/video, sms ...); they also depend on the

evolution of the activity which can make it different

the interactions (e.g. nature, priority) between the

participants, for instance when one of them

discovers a critical situation. By the way, multiple

time-varying constraints (e.g. power, bandwidth) are

also to be considered, depending on machine and

network resources which are used by the

participants.

In such a complex and dynamic context, future

communication systems are expected to have

behavioral and architectural self-adaptation

properties, aimed at tackling different kinds of

dynamic requirements, still considering dynamic

constraints associated with the network/machine

environment.

Accordingly, several solutions have already been

proposed in the literature; they differ in several

points related to the targeted goals (e.g. QoS,

security …), the addressed levels (Application,

Transport …), the adaptation actions and their

properties. Particularly, protocols whose internal

architecture may be dynamically composed appear

to be very suitable for both behavioral and

architectural self adaptation (Bridges et al., 2001,

Exposito et al., 2003, Mocito et al., 2005, Wong et

al., 2001). Our approach is based on adaptation of

such protocols at the end-to-end level (Transport

level - TCP level -and above). The goal is to match

as best as possible dynamic application QoS

requirements, while taking into account network

resource and context changes.

Dynamic configuration of adaptive protocols raises

different classes of problems that we address for the

Transport level adaptation.

Problem 1: Design and evaluation of new Transport

mechanisms and protocols, allowing optimizing a

given QoS requirements with regard to the network

resource constraints.

275

Armando F., Van Wambeke N., Chassot C., Exposito E. and Drira K. (2007).

A FRAMEWORK FOR THE DYNAMIC CONFIGURATION OF ADAPTIVE TRANSPORT PROTOCOLS - Application to QoS Requirements.

In Proceedings of the Second International Conference on Wireless Information Networks and Systems, pages 259-267

DOI: 10.5220/0002151202590267

 SciTePress

Problem 2: Elaboration of provisioning rules for the

selection and adaptation of the Transport

mechanisms to be (re)composed / (re)parameterized.

A major difficulty, without generic solution at the

current time, is to ensure coherency of the

composition / parameterization choices, both within

and between the considered adaptation levels.

Tackling these needs by means of informal models

may lead to non generic or suboptimal solutions due

to the complexity of the problem. In front of this

limit, formal model-based design constitutes a

promising approach (Farkas et al., 2006,

Landry et al.,

2004), particularly studied in our work for

architectural self adaptation at different levels of the

end-to-end communication stack (application,

middleware and transport levels) (Chassot et al.,

2006a). In (

Van Wambeke et al., 2007), we propose an

analytical model aimed at helping the decision

process at the Transport level was designed.

Problem 3: Design of an architecture for the

provisioning process, and then the enforcement of

the operational rules on the active communication

elements (e.g. adaptive Transport protocols).

It is the major goal of the NETQoS project to design

such architectures. The proposed approach starts

from the policy-based network management

(PBNM) model, which is extended to reach several

goals:

 management of several kinds of actor’s policy:

operator, service provider, user and

application;

 dynamic adaptation, not only at the network

level but also at the Transport level.

The main contribution of this paper has been

performed within the NETQoS project, and mainly

deals with the third problem exposed hereafter. We

present a framework for a model-based provisioning

and enforcement of configuration / adaptation rules

for adaptive Transport protocols. This framework is

then illustrated with a case study.

The rest of the paper is structured as follows.

Section 2 describes related work. Section 3 describes

the elements of the framework and the provisioning

process. Section 4 provides details about the

elaborated models for Transport-level adaptation.

Section 5 provides a case study illustrating the

provisioning process within the proposed

framework. Section 6 provides conclusion and future

work.

2 RELATED WORK

There are many existing solutions for context

adaptation. A complete study and classification can

be found in one of our previous works (

Chassot et al.,

2006b). In what follows, this classification is

summarized and dynamically configurable transport

protocols are presented.

2.1 Classification of Context

Adaptation Solutions

Adaptation objectives, techniques and properties are

among the main facets of adaptability. They are

studied and classified in this section.

2.1.1 Adaptability Objectives

Adaptability targets several objectives. QoS aspects

such as connectivity or access bandwidth issues in

roaming are considered in (

Kaloxylos et al, 2006). End

to end QoS optimization in the Best Effort Internet

makes heavy use of adaptation techniques (Akan and

Akyildiz, 2004). Security in wireless networks, such

as firewalls activation and deactivation, can also

benefit from adaptability (Perez and Skarmeta,

2004). Resources optimization related to device

power, computation or storage capability are

presented in (

Marshall and Roadknight, 2001).

2.1.2 Adaptation Techniques

Adaptation techniques target all layers of the OSI

model.

Application layer – (

Wu et al., 2001) addresses

adaptation of video streaming applications for the

Best-Effort Internet. The proposed techniques are

based on two mechanisms: an applicative congestion

control (rate control, rate-adaptive video encoding)

and time aware error control with FEC.

Middleware layer - Reflexive architectures such as

OpenORB or Xmiddle (

Capra et al., 2003) are good

supports for adaptation as they allow run-time

modification of the architecture.

Transport layer - TCP’s congestion control is a

well-known adaptation example. The IETF DCCP

protocol allows users to choose the congestion

control. SCTP targets adaptation to network failures

using multi homed associations. (Akan and

Akyildiz, 2004) studies various types of mobile

applications in wireless Internet. Adaptation consists

in parameterization of congestion control

mechanisms using context information. (Bridges et

al., 2001, Exposito et al., 2003, Mocito et al., 2005,

WINSYS 2007 - International Conference on Wireless Information Networks and Systems

276

Wong et al., 2001) study the architectural adaptation

of transport protocols by dynamic composition of

protocol modules. Next section (2.2) is dedicated to

these frameworks illustrating the modular

architecture concept targeted by our work.

Network layer – (

DaSilva et al., 2004) addresses

QoS-aware routing problems within ad-hoc mobile

networks. In (Wong et al., 2001), dynamic and

secure provision of IP services for military

wired/wireless networks is considered. In a policy-

based networking management context, the need for

self-adaptation is considered in (

Samaan and

Karmouch, 2005)

, using a learning-based approach.

MAC layer - The solutions handle connection and

access QoS problems for mobile users using

different terminals and roaming protocols. (

Kaloxylos

et al., 2006) provides a solution for optimizing the

handover latency but the other QoS requirements are

not considered.

2.1.3 Adaptation Properties

The adaptation solutions suggested in the literature

are defined in various ways.

The adaptation is behavioral when the execution of

a service can be modified without modifying its

structure. TCP and specific protocols such as the

ones in (Akan and Akyildiz, 2004) provide

behavior-based adaptation. It is easy to implement

but limits the adaptability range. Indeed, the addition

of new behaviours requires the component to be

recompiled and the adaptation can no longer be

performed during run-time.

The adaptation is architectural when the structure of

adapting services can be modified. The replacement

of a component by another can be implemented

following a plug and play approach where the new

component has the same interfaces as the replaced

one.

Finally, adapting components can reside on a single

machine or be distributed. In the first case,

adaptation is vertical and changes are performed

only locally. In the second case, it is horizontal and

synchronization between peer adapting entities has

to be managed.

2.2 Dynamically Configurable

Protocol Architectures

Dynamically configurable protocol architectures are

based on the protocol module concept. A protocol

module is a primitive building block (

Hutchinson and

Peterson, 1991)

resulting from the decomposition of

the protocol’s complexity into various successive

elementary functions. A protocol is then viewed as

the composition of various protocol modules in

order to provide a global service.

These architectures can be refined into two different

categories depending on their internal structure: the

event based model (followed by Coyote and Cactus)

and the hierarchical model (X-Kernel (

Hutchinson and

Peterson, 1991) and APPIA). ETP follows an hybrid

approach combining both models (Exposito, 2003).

These protocol architectures appear as a good choice

for future communication protocol’s self-adaptation

as they are capable of run-time architectural

adaptation, meaning that the modules composing

them can change during the communication. This

run-time architectural adaptation raises many

problems such as: (1) synchronization of adapting

peers; and (2) the choice of the best composition

guided either by the user’s requirements or by the

modification of the context.

3 THE FRAMEWORK

COMPONENTS AND THE

PROVISIONING PROCESS

The NETQoS IST project is addressing the problem

of QoS in IP Networks from Policy based network

management architecture, to provide flexible and

adaptive end to end QoS provision, unlike number of

solutions provided by bottom-up approach. The

NETQoS system introduces the notion of predefined

policies as a promising solution to address the needs

of QoS traffic management.

In order to address the policy based QoS

management, policy architecture was defined.

Policies prescribe a set of rules based on the

users/application requirements, which are

transformed into high level network level policies.

Since the network level-policies are derived from

business objectives, users and applications

requirements described in the SLA, these policies

have to evolve and to be adapted to the changes in

these objectives and requirements in a timely

manner. An autonomous self-adaptable policy-based

management framework with inherent dynamic

capabilities to best manage, customize and extend

the underlying complex infrastructure of

communication systems resources in response to the

continuously changing business objectives and users

requirements, is the main goal of the NETQoS

project. The enforcement policies thus generated are

translated into network device dependent rules to

configure the network

A FRAMEWORK FOR THE DYNAMIC CONFIGURATION OF ADAPTIVE TRANSPORT PROTOCOLS -

Application to QoS Requirements

277

3.1 The framework for Dynamic Policy

Provisioning

The general architecture of the NETQoS system

distinguishes four main entities or systems (Fig. 1):

 the Policy Description (PD) implements a set of

ontologies used to specify the actor-level

policies, the operational policies, etc;

 the Actor Preference Manager (APM) provides

NETQoS GUI/API allowing users to define

actor-level dynamic policies to the NETQoS

system based on ontologies. These policies

(e.g. requirements, preferences, profile,

quality reporting…) may be expressed before

or during the communications;

 the Automated Policy Adaptor (APA) is the

central entity of the NETQoS system. It does

not provide QoS by itself, but provides and

dispatches operational policies (namely

Network and Transport level policies) that

allow the communication system to take into

account the actor-level dynamic policies;

 the Monitoring and Measurement (MoMe)

implements all the monitoring and

measurement tasks associated with: (1)

context evolution, e.g. actor’s policies

evolving, end systems/network resource

changing …; and (2) evaluation of the

operational policy efficiency and the reporting

of quality information to the actors.

Acto rs policy

re pository

Con text M an ag e r

MoMe

AP M

Polic y Enf o rce ment MG

Policy Decision MG

App. A PISP GUI Op. G UI User GUI

Context Repository

APA

Transport PDM

Ne t w o r k P D M

Multi level

decis ion

coordinator

Transport PEM Network PEM

MoMe Manager

P oli cy Adap tor MG

NetworkTransport

APA

Operation al Policy

Repo sito ry

Policy Descriptio n

Communication link

MoMe Tool

Figure 1: The general NETQoS architecture.

3.1.1 The Automated Policy Adapter

The APA is aimed at deciding, dispatching and

adapting the operational policies that allows the

communication system to take into account the

actors' level dynamic policies taking into account the

context evolving. The APA is composed of three

main components:

 the Policy Decision Manager (PDM) has in

charge deciding the set of operational policies

that implements the actor-level policies. This

mapping may change depending on context-

related information;

 the Policy Enforcement Manager (PEM) has in

charge the deployment of the policies decided

by the PDM on their Network/Transport-level

enforcement points;

 the Policy Adaptation Manager (PAM) has in

charge the adaptation of policies, individually

or by grouped, when the success criteria

associated with a policy is not reached.

Policy decision manager

. The PDM is aimed at

deciding an optimal set of policies to be settled at

the Network and/or at the Transport level to satisfy

the set of actor-level policies. This policy

provisioning may be performed using rules taking

into account the dynamicity of the actor’s policies

and changes in the context information; these rules

are elaborated using a model-based approach (see

section 3.3.2). The policy provisioning may lead to

policy conflicts that the PDM has to solve, for

instance when actor-level policies cannot be reached

as required.

Each time the PDM decides a (new) operational

policy, it provides the corresponding rules to the

PEM. If the policy is enforceable, the PDM informs

the PAM of the enforcement of a new policy.

Policy enforcement manager

. The PEM is in charge

of dispatching the operational policies decided by

the PDM to the actual policy enforcement point

(PEP). For instance, for a Transport level adaptation,

the PEM dispatches the Transport protocol

configuration rules to be applied on the end nodes.

The PEM is independent of the Network and

Transport technologies that are used to really

enforce the policies, i.e. the PEM provides the rules

to be performed in a language that is independent of

the ones used by targeted QoS-oriented

communication system. Consequently, the PEM

provides generic interfaces allowing the different

entities to communicate with the actual PEP, and

adaptors have to be implemented, for instance on the

PEP themselves, to translate the generic PEM rules

into specific technology-dependant rules.

Policy adaptation manager.

The PAM has in charge

the adaptation of a policy when its success criteria

are not reached. The PAM may decide to adapt the

policy or an associated subset of relevant policies.

When adaptation is not possible, the PDM is

informed of the current policy failure, possibly

associated with a diagnosis.

The PAM mainly acts in the following situations:

WINSYS 2007 - International Conference on Wireless Information Networks and Systems

278

 when the PAM is informed by the PDM that a

new operational policy is enforced, the PAM

stores the policy in the operational policy

repository together with some success criteria

(e.g. end-to-end packet loss must be smaller

than 5%).

 when the PAM receives alarms from the

MoMe, it tries to adapt the failed policy and

informs the PDM of the result.

3.3.2 Policy Provisioning

Two different policy enforcement levels are

considered: the transport level and the network level.

For both levels, policy provisioning deals with the

way the operational policies are elaborated, selected

and adapted.

This process is complex as it requires knowledge of

many context aspects such as access network or

actors’ preferences. The component that performs

policy provisioning is the central element of the

NETQoS framework.

Various techniques exist for the provisioning

process. The simplest one consists in having a set of

predefined rules which are hard coded into the

deciding component producing a set of predefined

responses that depend on the environment. The main

aspect (among many others) that makes this simple

technique inefficient is its lack of extensibility.

Indeed, being hard coded, the policies that manage

the decision process are not changeable during run-

time.

In order to have an extensible decision process, it is

possible to guide it using different external models

such graph-based models presented in (

Chassot et al.,

2006b)

where all aspects of communication can be

represented and the different evolutions of the

system are characterized as elements of a graph

grammar. In such models, the adaptation is thought

ahead of time providing fast response to changes.

Moreover, these models can be stored in an external

repository and may change during the

communication. By doing so, the set of responses

that the decision component may have is not

statically predefined at design-time and can be

further extended by refining the models that manage

the system evolving during run-time.

Moreover, different models can be used at different

abstraction levels. Hence, the cooperation and

interaction description can be implemented by a

labelled and directed graph model that may be

transformed following a set of graph grammar rules

into a transport connection model. At the transport

level, specialized models such as analytical models

can be used in order to get a policy response which

is best suited to the environment while taking both

actor requirements and preferences into account.

In the NETQoS system, the component in charge of

this decision process is the APA. Various sub

components take place in this provisioning process,

the PDM is responsible for performing the

provisioning. The PAM then takes the necessary re-

provisioning actions in order to perform adaptation.

1) Provisioning. In order to perform provisioning at

the Transport and Middleware levels, policies in

place at the network level have to be known. Due to

this constraint, the natural order for the provisioning

consists in initially performing network provisioning

prior to transport provisioning. Once these two steps

are done, middleware provisioning can start.

2) Adaptation. Once the adequate communication

services have been selected and deployed, specific

adaptation actions are performed in order to

maintain the QoS required. The adapting conditions

are expressions based on the current media flows

and applications regarding the QoS goals computed

during the decision process. One possibility is to

adapt the middleware level first then the transport

level and finally the network level.

3.3.3 Policy Provisioning Process

In order to perform provisioning at both the network,

transport and middleware levels, the PDM

component has been divided in two different

components that are responsible for (N-PDM)

network level provisioning, (T-PDM) transport level

provisioning and (M-PDM) middleware level

provisioning.

The policy provisioning process can be lead using

different models at these various levels. In order to

support this adaptation, various architectures are

possible. The configuration and outsourcing models

defined by the IETF (Boyle et al., 2000) are suitable

for such task.

For instance, for an interactive and high priority

flow, a network service characterized by low delay

and high reliability could be selected in a first time

(e.g. EF service). In a second time the selection of a

basic UDP transport service without additional

middleware mechanisms would be enough to

guarantee the satisfaction of the user. In contrast, if

the EF service has not been declared as available for

this user (e.g. following the service provider

policies), a Best-Effort network service could be

selected, combined with an UDP or DCCP transport

service and completed with an error control

mechanism implemented at the middleware level.

A FRAMEWORK FOR THE DYNAMIC CONFIGURATION OF ADAPTIVE TRANSPORT PROTOCOLS -

Application to QoS Requirements

279

The design can be done as illustrated by the figure 2

below. In this architecture, the outsourcing model is

followed; the PDM component represents any of the

N-PDM, T-PDM or M-PDM. The decision is not

directly performed by the PDM but it is delegated to

external modules (represented by squares on the

diagram). In each of these modules, a specific

decision algorithm is implemented in order to

construct a potentially partial policy. The successive

invocation of various modules (TD...ND) will lead

to a valid policy being generated.

For each of the outsourced modules, the access

method can be of any kind (from simple RPC to

XML/SOAP Web Services). This design

architecture allows easy extension of the system by

simply adding new decision components. By such,

any new model that is produced by the different

actors (Operator/Service provider) can be deployed

and used without requiring modification of the PDM

itself. The PDM acts here as a system kernel

managing the process by delegating tasks to the

various external elements successively.

Figure 2: Policy provisioning process architecture.

The above figure presents an outsourcing based

model for the PDM decision process. When

necessary (e.g. when a new connection is to be

established in the NETQoS domain), the PDM

retrieves the information regarding the actor

preferences and associated policies from the

repositories. These policies contain information that

allows the PDM to contact the appropriate software

module to perform the decision. Technologies such

as RMI, Web Services or CORBA may be used for

this task. As this software module is external to the

PDM, this model is largely extensible by the simple

addition of new modules and simply describing them

in the repositories. This action also allows users to

choose and classify the adaptation actions provided

by these new modules in their preferences.

In order to support different interfaces, the

invocation takes place in two stages. Stage 1 allows

for discovering the parameters required by the

outsourced modules (i.e. inputs to the decision

model they hold inside). The PDM is then

responsible for retrieving the necessary information

from the other components in the NETQoS system

(e.g. MoMeTool or Context Manager). Once these

parameters have been retrieved, the outsourced

decision service can be invoked by the PDM

returning the decided policy. The PDM could then

continue invocation following the users' preferences

(i.e. perform adaptation both at Middleware and

Network levels simultaneously) which would lead to

further modification of the policy being decided

upon. At the end of this process, the PDM has

constructed the policy to be activated and deployed

in order to support the new communication. This

decision is then transmitted to the PAM and PEM in

order for them to take the necessary actions to

respectively, monitor and react to adaptation needs

by enforcing and deploying the policy via the PDPs.

4 THE PROPOSED MODELS FOR

THE TRANSPORT-LEVEL

ADAPTATION

4.1 Adaptation at the Transport Level

Transport-level adaptation consists in dynamically

selecting and adapting parameters and/or internal

architecture of the adaptive Transport protocol used

to transfer an application data flow. The goal is to

match as best as possible dynamic and hierarchical

requirements, i.e. an actor’s policy, associated to a

given application, taking the current communication

context (e.g. access networks used by the terminals

hosting the applications):

 the processing of an actor’s policy, typically a

user’s policy for a given application when this

one is launched, may lead to Transport-level

adaptation for the connection underlying this

application only. It may also lead to

Transport-level adaptation for other

connections when hierarchical relationships

(e.g. priorities) have been defined between the

actor’s policies, for instance when a user has

defined priorities between his/her

applications;

 the communication context may evolve

depending on several factors, for instance

when the hosting terminal is moving from an

access network to another.

WINSYS 2007 - International Conference on Wireless Information Networks and Systems

280

To handle dynamic and hierarchical policies,

together with an evolving communication context,

two kinds of Transport policies, namely per

connection handling, and per group of connections

handling, may be applied:

 per connection handling means that the

adaptation rules are applied independently,

connection per connection;

 per group of connections handling means that

several connections may be coordinated by the

adaptation rules, allowing taking into account

hierarchical policies, for instance expressed by

a user on his/her applications.

Both policies are refined in operational rules that

consist in composition and parameterization rules of

the transport protocol implementing each

connection.

4.2 The Proposed Models for Policy

Provisioning

In what follows, a model for automatically selecting

the best composition of protocol modules at the

transport/middleware level is presented. This model

has been previously presented in (

Van Wambeke et al.,

2007)

and is only briefly recalled here.

We model a protocol module (A) by the following

matrix:

A =

[ A

]

Given this description, we are able to describe a

valid composition as respecting the following

constraints:

Additionally, for each module, efficiency vectors are

defined which allow for evaluating the module’s

efficiency in the various contexts that are

considered.

Based on the above, the decision process comes to

solving the following optimization problem:

The additional constraints are produced by the

refinement of the policies. For example, a user might

specify that he is using a low memory device which

would be refined into a constraint that limits the

number of unused variables production.

5 CASE STUDY: QOS

OPTIMIZATION FOR MOBILE

USERS IN NETQOS

5.1 Scenario Description

This scenario aims at illustrating Transport

adaptation by instantiating the above presented

model in the decision process.

Assumption

For this scenario, let’s assume that

 the user’s terminal and the server are connected

to high speed wired access networks,

interconnected by a Best Effort Internet

domain;

 the user's policies (applicable to the present

context) are as follow:

− “when I'm on a wired connection, I would

rather have ensured QoS of Gold quality

for all my applications”;

− “when I'm on a wireless network, I would

rather have my communications classified

by decreasing level of quality

requirements as follow: Audio call, Video

call, VoD movie, Internet Radio, Web,

Mail and similar, File transfers”.

While the user is connected to the wired network,

the service provider's policies are such that the

maximum amount of bandwidth that the user might

use is 10 Mbps, thus, resource reservations will be

performed and updated until this limit is reached.

Note that these reservations are handled by any QoS

system in place such as the one proposed in the IST

FP6 EuQoS project.

Let’s now assume that:

 at t = t0, the user starts to work at his office

desk;

 at t = t1 > t0, the user disconnects his laptop

and the wireless interface becomes the default

interface (this is handled by the OS itself).

5.2 Interaction of Policy Actors and

NETQOS Components

The NETQoS components that are involved in this

scenario are:

⎩

⎨

⎧

otherwise 0

i variablerequiresA if 1

⎩

⎨

⎧

otherwise 0

i variableproducesA if 1

A FRAMEWORK FOR THE DYNAMIC CONFIGURATION OF ADAPTIVE TRANSPORT PROTOCOLS -

Application to QoS Requirements

281

 MoMe (Monitoring and Measurement): Context

Manager (CM) and MoMeTool,

 APA (Automated Policy Adapter): PDM, PEM

and PAM (for the Transport only, i.e. adaptation

at the network level is not considered in this

scenario),

 REPO (Policy Repository).

The following illustrate the main scenario steps.

1) At the NETQOS system initialization, the APA

subscribes to the CM for a set of events: application

launch, policy violation, context modification etc.

2) A t = t0, the launch of various applications is

detected by the CM; for each one:

 the CM informs the PDM that an application is

launched;

 the PDM retrieves from the REPO the policies

related to the identified user and application;

these policies contain information that allows

the PDM to contact the appropriate external

software module that decides the

Middleware/Transport protocol configuration

to be selected with regard to the policy

defined for the ftp application;

 the PDM contacts the outsourced software

module, and provides it with the necessary

information to decide the rules to be applied;

 the appropriate rules for Transport protocol

configuration are returned to the PDM by the

outsourced software module following the

model presented above;

 those rules are then transmitted to the PAM and

PEM in order for them to take the necessary

actions to respectively, monitor and react to

adaptation needs and enforce and deploy the

policy via the PEP at the Transport layer;

 if the MoMeTool, via the CM informs the PAM

that success criteria associated to the selected

policy are no longer met (via alarms); the

PAM may try to adapt the policy and may also

inform the PDM if necessary. For instance, the

adaptation action can consist in changing

parameters of the micro-protocols.

3) A t = t1, the change of network context is detected

by the CM

 the CM informs the PDM that the user is

moving to a different network;

 the PDM retrieves from the REPO the policies

related to the identified user and application

for the new context; these policies contain

information that allows the PDM to contact

the appropriate external software module that

implements the model presented in the

previous section in order to decide on the

Middleware/Transport protocol configuration

to be selected with regard to the policy

defined for the currently running applications.

 At this step, two possibilities may be

considered.

possibility

− the outsourced software module provides

the PDM with the set of rules that can be

applied for the application;

− taking into account its current execution

context (here, various active applications

for which the QoS requirements have

changed), the PDM/PAM selects the

Transport protocol configuration /

adaptation to be applied for each

connection.

possibility

− the appropriate rules for Transport

protocol configuration/adaptation for the

new connection and possibly existing

connections are returned by the

outsourced software module;

− those rules are then transmitted to the

PAM and PEM in order for them to take

the necessary actions to respectively,

monitor and react to adaptation needs and

enforce and deploy the policy via the PEP

at the Transport layer;

 if the MoMeTool, via the CM informs the PAM

that success criteria associated to the selected

policy are no longer met (via alarms); the

PAM may try to adapt the policy and may also

inform the PDM if necessary.

When an application is stopped, the PAM and PDM

may also apply reconfiguration actions.

6 CONCLUSION

In this article, we presented a framework for the

dynamic configuration of adaptive Transport

protocols in order to support policy-based QoS

provisioning for heterogeneous, mobile and

cooperative activities in a pervasive Internet. The

main components of the general policy architecture

have been described, especially the APA

component, responsible for the policy decision

process of adaptation actions, in particular at the

Transport level. A case study aimed at displaying the

framework’s usage and internal interactions in the

case of mobile, multi-network, continuous, seamless

communications has been presented. The models

which govern the policy-based decision and

WINSYS 2007 - International Conference on Wireless Information Networks and Systems

282

provisioning process have been illustrated in this

context.

Future works on the topic include, but are not

limited to, the implementation of the presented

framework. Moreover, evaluation of the efficiency

of the different mechanisms that compose the

current architecture has to be carried on.

ACKNOWLEDGEMENTS

Part of this work presented has been carried on as

part of the European IST FP6 NetQoS project. In

conjunction with this, several aspects of this works

are funded by DGA grant.

REFERENCES

Bridges, P. G., Chen, W.-K., Hiltunen, M. A., Schlichting,

R. D., 2001. Supporting coordinated adaptation in

networked systems, In 8th Workshop on Hot Topics in

Operating Systems (HotOS-VIII), Elmau, Germany.

Exposito, E., Diaz, M., Senac, P., 2003. Fptp: the xqos

aware and fully programmable transport protocol, In

11th IEEE International Conference on Networks

(ICON’2003), Sydney (Australie) - pp.249-254.

Mocito, J., Rosa, L., Almeida, N., Miranda, H., Rodrigues,

L., Lopes, A., 2005. Context adaptation of the

communication stack, In Proceedings of the 25

IEEE

International Conference on Distributed Computing

Systems Workshops (ICDCSW’05), Colombus,

Colombus, OH, USA.

Wong, G. T., Hiltunen, M. A., Schlichting, R. D., 2001. A

configurable and extensible transport protocol, In

Proceedings of INFOCOM, Anchorage, Alaska, USA,

pp. 319–328.

Farkas, K., Wellnitz, O., Dick, M., Gu, X., Busse, M.,

Effelsberg, W., Rebahi, Y., Sisalem, D., Grigoras, D.,

Stefanidis, K., Serpanos, D., 2006. Realtime service

provisioning for mobile and wireless networks, In

Elsevier Computer Communication Journal Vol. 29,

No. 5, 540–550.

Landry, R., Grace, K., Saidi, A., 2004. On the design and

management of heterogeneous networks : A

predictability-based perspective, In IEEE

Communications Magazine Volume 42, Issue 11

Page(s):80 – 87.

Chassot, C., Lozes, A., Racaru, F., Auriol, G., Diaz, M.,

2006a. A user-based approach for the choice of the ip

services in the multi domains diffserv internet, In 1st

International IEEE Workshop on Service Oriented

Architectures in Converging Networked Environments

(SOCNET’06).

Van Wambeke, N., Armando, F., Chassot, C., Exposito,

E., 2007. A model-based approach for self-adaptive

Transport protocols, to appear in Elsevier Computer

Communications, Special issue on End-to-end Support

over Heterogeneous Wired-Wireless Networks.

Chassot, C., Guennoun, K., Drira, K., Armando, F.,

Exposito, E., Lozes, A., 2006b. Towards autonomous

management of qos through model driven adaptability

in communication-centric systems, In International

Transactions on Systems Science and Applications 2

(3) 255–264.

Kaloxylos, A., Lampropoulos, G., Passas, N., Merakos,

L., 2006. A flexible handover mechanism for seamless

service continuity in heterogeneous environments, In

Elsevier Computer Communications Volume 29, Issue

6, 717–729.

Akan, O. B., Akyildiz., I. F., 2004. ATL: An adaptive

transport layer suite for next-generation wireless

internet, In IEEE Journal on Selected Areas in

Communications Vol. 22 (No. 5) 802–817.

Perez, G., Skarmeta, A. G., 2004. Policy-based dynamic

provision of ip services in a secure vpn coalition

scenario, In IEEE Communications Magazine Volume

42, Issue 11: 118 – 124.

Marshall, I., Roadknight, C., 2001. Provision of quality of

service for active services, In Elsevier Computer

Networks Volume 36, Issue 1 75–85.

Wu, D., Hou, Y. T., Zhu, W., 2001. Streaming video over

the internet: Approaches and directions, In IEEE

Transactions on Circuits and Systems for Video

Technology vol. 11, no. 1.

Capra, L., Emmerich, W., Mascolo, C., 2003. Carisma:

Context-aware reflective middleware system for

mobile applications, In IEEE Transactions on

Software Engineering 29 (10) 929–945.

DaSilva, L. A., Midkiff, S. F., Park, J. S., Hadjichristofi,

G. C., Davis, N. J., Phanse, K. S., Lin, T., 2004.

Network mobility and protocol interoperability in ad

hoc networks, In IEEE Communications Magazine

Volume 42, Issue 11 88–96.

Samaan, N., Karmouch, A., 2005. An automated policy-

based management framework for differentiated

communication systems, In IEEE Journal on Selected

Areas in Communications vol. 23, no. 12.

Hutchinson, N. C., Peterson, L. L., 1991. The x-kernel: An

architecture for implementing network protocols, In

IEEE Transactions on Software Engineering 17 (1)

64–76.

Exposito, E. J., 2003. Design and implementation of

quality of service oriented transport protocol for

multimedia applications, Ph.D. thesis, National

Polytecnic Institute of Toulouse, Network and

Telecommunications.

Van Wambeke, N., Armando, F., Chassot, C., Exposito,

E., 2007. Architecture and Models for self-adaptability

of Transport protocols, to appear in Proceeedings of

the 2nd International IEEE Workshop on Service

Oriented Architectures in Converging Networked

Environments (SOCNE’07).

Boyle, J., Cohen, R., Durham, D., Herzog, S., Rajan, R.,

Sastry, A., 2000. The COPS (Common Open Policy

Service) Protocol, RFC 2748.

A FRAMEWORK FOR THE DYNAMIC CONFIGURATION OF ADAPTIVE TRANSPORT PROTOCOLS -

Application to QoS Requirements

283