RESEARCH ON THE CLASSIFICATION CONSULTATIONS

AND SELF-ADAPTATION OF QOS IN COGNITIVE NETWORKS

Xiujun Gu, Runtong Zhang, Zhongtian Li and Xu Han

Institute of Information Systems, Beijing Jiaotong University, Beijing 100044, China

Keywords: Cognitive Networks, Quality of Service, Scalability QoS Index, Self-adaptive Mechanism.

Abstract: Classification of scalable QoS systems requires more reasonable index, it needs comprehensive and deep

analysis and evaluation of business requirement to realize the expansibility of the index and their mapping

algorithm. In this paper, it proposes a set of extensible QoS index system on the existing basis that these

mechanisms divide grade too simple, and the kinds of index are so few. Making good use of the advantages

of cognitive networks on the dynamical re-configuring and communicating capabilities, meeting amounts of

various requirements of QoS, designing interface of expandable QoS index, providing more precise and

concrete QoS service and improving the efficiency of the networks resources are the main works of this

paper.

1 INTRODUCTION

In the current information network, the network

elements, such as nodes, protocol layers, strategies

and behaviors, are restricted greatly in the aspects of

status, scope of application, and response

mechanism, therefore they cannot satisfy the QoS

demands of clients with diversification and

individuation. Some scholars have already proposed

adaptive mechanism (Zhang L et al., 2007, Wang F

et al., 2003, Partha P et al, 2000, Stroud R et al.,

2004) in order to ensure the service qualities of sorts

of business. These adaptive mechanisms which have

been already existed, however, are all reactive

modes that can only make adjustment after

occurrences of problems. Also, the optimization

objects are usually aiming at a layer of protocol

stack, some key nodes of network, or a certain part

of transmission link. Hence the end-to-end QoS

performance of a network cannot be performed

optimized. The information networks of new

generation are with more complication and

summarization. Therefore proactive adaptive

algorithm needs more research to provide more

This work is partially supported by National Nature Science

Foundation Of China (No.60773033) and the National High

Technology Research and Development Program Of China (863

Program) (No.2009AA01Z211)

suitable guarantee of QoS on business. Cognitive

network (CN) (Thomas R W, 2005) can exactly

meet this demand because of its unique ability of

self-learning and re-configured. As the more

intelligent, cognitive networks can help to provide

users with richer types of business, more reliable

information transmission and more optimized

end-to-end QoS performances.

The principle of cognitive behavioral model

(Feizi-Khankandi S et al., 2007) is: first, it detects

the current state of the network, and then executes

adjustment, adjudication, and implementation based

on the observed network conditions and parameters.

Cognitive technology makes communication entities

possess a cognitive ability of the surrounding

environment, and can make dynamical change which

is intelligent, independent, and adaptive according to

the change of surrounding environment.

The research of the cognitive networks on the

goal of end-to-end QoS performance optimization in

China and in abroad are both at an embryonic stage,

with only a rough conceptual model and a small

amount of cognitive network routing algorithm

(Sahoo A, 2002). The implement of QoS

mechanisms in Cognitive network includes:

classification and definition of QoS metrics,

admission control and consultation, resource

reservation, resource scheduling and management

issues which are still lack of practicable solutions

nowadays.

218

Zhang R., Han X., Li Z. and Gu X.

RESEARCH ON THE CLASSIFICATION CONSULTATIONS AND SELF-ADAPTATION OF QOS IN COGNITIVE NETWORKS.

DOI: 10.5220/0003266702180225

In Proceedings of the Twelfth International Conference on Informatics and Semiotics in Organisations (ICISO 2010), page

ISBN: 978-989-8425-26-3

Based on cognitive network which has the advantage

of dynamical reconfiguration and information

interaction capability as the background, this paper

is to carry out extensible QoS classification

mechanism which possess sorts of index options, it

executes the reasonable QoS classification according

to the needs of the business. The mechanism takes

all factors into consideration such as the character of

different users and the change of network state. It

performs real-time end-to-end QoS classified

negotiation and revision before the start and also

during process of business, in order that it could

provide more suitable service for users, and also

raise the utilization rate of resources. This paper

proposes many reasonable indexes and mapping

algorithm, designs the expansion interface, based on

the actual demand of the business to dynamically

increase the new QoS index according to the sorts of

business. To design the QoS negotiation and

adaptive adjusting strategy for progressing the end to

end QoS negotiation and re-negotiation, it fully

considers the network state information of the nodes

of the cognitive network, then, it could adjust the

QoS classification.

The structure of the paper is as follows: The

second section analyzes the existing QoS

mechanisms, the third section presents scalable QoS

indicators system, the forth section designs and

presents an adaptive QoS mechanism, the fifth

section simulates the mechanism proposed, and the

last section makes a conclusion.

2 LITERATURE REVIEW

At present, the research on QoS is mainly based on

MPLS (Dekeris B and Narbutaite L, 2004), DiffServ

(Lee G, 2008) and other models for the mechanisms

optimization, as well as mapping algorithm of QoS

mechanisms. While relatively less research done on

refined and extensible QoS mechanism for

diversification of the business. Most of them are at

steps of tracking, introduction, digestion and

absorption.

In July 2006, ITU-T adopted a recommendation

called ITU-T recommendation Y.2111 (Yamada H

et al., 2007). The proposal addressed a key area in

NGN, provided end to end QoS capability. The

recommendations involved resource and admission

control functions (RACF) (Kamatani O et al., 2008)

which will help operators to ensure the quality of

end-to-end multimedia service in NGN, such as

VoIP and IPTV. The key of the approach is the

ability that operators designated criteria for specific

types of traffic capacity, for better organization of

the network resources. RACF meets the demand of

more intelligent control based on packet network

infrastructure. The recommendation defined the

relevant requirements and functional architecture,

including resource reservation, admission control

and access control, Network Address Port

Translation (NAPT) and firewall control, and

Network Address Translation (NAT traversal).

Currently the more research architecture of

end-to-end scalable QoS is a feasible framework

presented by a scholar of Australia University of

Technology academics (Hoang D B and Phan H T,

2007). This architecture took into account the

distinction between the business service capabilities

and operational QoS support capabilities for the IP

network orientation during the design of network

services. This architecture has the capabilities to

support the DiffServ extensions and explores a new

path congestion control method across multiple data

streams of resources between the rational

distribution and aggregation. Also, it can make QoS

consultations among several domains by using QoS

extended functions of BGP. This architecture uses a

simple router control panel, and the sending panel

does not require other complex PHB scheduling.

For the IP network, it is essential that its network

architecture can provide differentiated services to

applications, to ensure quality of service, and can be

called service-oriented architecture. In this literature

(Hoang D B and Phan H T, 2007), a new initiative

QoS framework called End-Diff has many features

required for this paper. This architecture is more

scalable than DiffServ. Through simulation, this

End-Diff QoS architecture has achieved good results

on scalability, system overhead, routing strategy,

fairness, packet delay and jitter conditions.

There was a scalable IP network QoS

mathematical model (Fgee E B et al., 2008)

proposed, taking into account the requirements of

the network forwarding service assurance by VOIP,

E-Commerce and other new services. In this model,

a new QoS mathematical model was designed based

on network calculus theory, which means the

maximum rate of a new access transport stream is

limited by a correlative arrival curve. The

restrictions of transmission rate and time extension

are realized by the service-by-hop delay time on the

transmission path.

Some scholars proposed an internet transmission

strategy (Chan H. et al., 2005) which can support

scalable QoS security capability. For the

disadvantage that the traditional port-based

classification methods cannot obtain a valid result of

RESEARCH ON THE CLASSIFICATION CONSULTATIONS AND SELF-ADAPTATION OF QOS IN COGNITIVE

NETWORKS

219

the business classification, such a mechanism

achieves scalable QoS protection by distinguishing

Internet transmission level by the application of

business type. The strategy is divided into two steps,

first, select the characteristics of the business which

mainly based on information provided by the ISP;

second, balance the system benefits and system

complexity through the features of degradation, and

then use REPTree bagging mechanism (Park J et al.,

2006) to conduct service classification. From the

above analysis we can see that most of the above

research is QoS mechanism optimization based on

service demands of single business, under the

balance between the interests of the various network

elements, optimizes the QoS guarantee mechanism,

but not for the diverse needs of the business,

particularly in the characteristics of cognitive

network, to carry out the research of QoS

mechanisms scalability and its implementation

strategy.

There was a literature (Junghun P and

Hsiao-Rong T, 2006) proposed a new approach that

makes classification for network business based on

scalable QoS application type. The results of using

traditional port-based classification method were not

satisfactory because the same port number may be

shared by multiple applications. The

non-equilibrium route, such as the PCF and the Net

Flow as well as errors caused by modern

measurement tools are reasons for this problem. In

order to identify those problems, the paper described

the classification process consists of two steps:

feature selection and classification. First check those

optional features easily learned by the ISP to balance

the service performance and complexity.

Currently there is a new mathematical model for

QoS (Ducatelle F. et al., 2005) in Australia which is

based on the principle of network calculus. In this

model, the inbound service flow is regulated by the

arrival curve, and upper limit of arrival curve is

maximum flow of the inbound service flow. On each

node, the lower bound and delay time is determined

by service waiting time before the business flow

arrives the destination node. Compared with other

models, the model has an acceptable end-to-end

delay.

Among the domestic research results, there is a

paper (Yunbo W, 2006) that analyzed the business

features and characteristics of network needs of the

Internet Large-scale multimedia applications. And it

focused on requirements of the Large-Scale

Application on network transmission reliability and

scalability. Combined comparative analysis of the

existing IP QoS plans and meets the QoS

requirements of Large-Scale application, it proposes

QoS control thought based on PCF and builds the

corresponding support for Internet Large-Scale

Multimedia Applications E2E QoS framework for

the implementation of the system ON - QoS. Finally

it do research about the aggregate flow of the

ON-QoS scheduling strategy.

Another domestic study results (Suogang L et al.,

2007) proposed a scalable multicast program for

unordered QoS. The mechanism uses the order

feature of the QoS level and chooses an appropriate

multicast tree for multicast group by heuristic

algorithm. So that the multicast receivers who are

belong to different multicast groups and request the

same QoS level can share the multicast tree. The

simulation result showed that this program can

effectively improve the scalability of multicast state,

and in certain experimental settings the ratio of

multicast trees and multicast groups may be less

than 1 / 8; and can satisfy the different QoS receiver

demand at the same time.

In the ChinaCom 2007 meeting, the Chinese

scholars proposed a solution (Meng S and Bertrand

M, 2008) that can make overlay service management

aware of QoS information, so that QoS reminder

service (QSON) with overlay network can be

provided. In the QSON, the service components are

organized into different subnet, and the QoS of

service in the same subnet are similar. With the help

of QSON, QoS has been unitopia integrated into

such a similar information retrieval, so that the

service query and choice can be easily obtained.

Through a P2P-based services positioning system,

service routings become with QoS conscious and

scalable capability, and service routings are able to

achieve good service performance.

3 SCALABLE QOS INDICATOR

SYSTEM

As an important approach to improve resource

utilization and overall performance, QoS security

system almost penetrates any network with limited

capacity in the design process and evaluation criteria.

The most typically ones include distributed

computing systems of the Internet, packet-based

systems of telecommunication networks and

computer networks. The QoS parameters of former

ones include packet loss rate, delay, jitter, order

delivery and error string, while the latter ones also

include the availability, throughput and so on. At

present, a lot of cutting-edge real-time media

ICISO 2010 - International Conference on Informatics and Semiotics in Organisations

220

streaming technologies, such as Voice over Internet

Protocol (VoIP) and Internet TV technology (IP-TV),

are very sensitive to the delay and require a

relatively fixed bit-transfer rate. How to optimize

and integrate the coding method , the priority

distribution of grades (Prioritization) and real-time

scheduling mechanism, under certain bandwidth

constraints, to meet the stability and real-time

network communication requirements is the core

issue considered with priority.

Based on the feature that cognitive networks can

be dynamically reconfigured, according to the

principles of making and distribution QoS parameter

index, in this section we designed a scalable QoS

mechanism. This mechanism can increase the QoS

business indicators dynamically based on business

application requirements to provide accurate specific

QoS classification for business with different

characteristics to enable businesses to obtain more

suitable services.

4 SELF-ADAPTIVE MECHANISM

Self-adaptive mechanism is an application-level QoS

assurance mechanism, which mainly deal with the

strategy of the parameter adjustment according to

the user’s demands. Self-adaptive mechanism is the

scope of QoS guarantee which can be defined as

constraint by the user (application) for the QoS

request strategies, through testing the state of the

system and network resources and environment to

adjust application strategy. It implements the

suitable behavior to guarantee the user acceptable

QoS request. It mainly reflects the characteristics are

as following:

First, It is designed on the base of the specific

type of application-centric. Therefore, different

application types have different mechanisms of

implementation. There is no single realization

mechanism which is suitable for all application

types.

Second, to solve the user's QoS Strategy level

demand for parameter adjustment strategy, this kind

of strategy demand proposed is mainly to adapt the

network resource change.

Third, It shows primary role of the scope of

end-node in the network, end system and above

network level. Forth, It is one end-to-end realization

mechanism, not only realize on an end system.

Fifth, It is one much more fine-grained

network-level QoS to ensure the application of

personalized expression.

4.1 Mechanism Constitution

The mechanism is mainly composed of three key

components: detection, memory base, index

mapping. As shown in Figure 1:

Figure 1: Composition of self-adaptive mechanism.

Before the service request arriving at the processor,

first it wait in the waiting stack, and then to detect

the network environment. Including the service type,

the data type, as well as the network type of the goal

terminal located. After obtaining the comprehensive

record of network environment, then put the record

into the memory base for comparing, examine that

the historical grade decision-making which the

computer adopts under the same network

environment.

Based on the multi-dimensional analysis of the

memory data base according to the different

environmental information, if it has the same

historic record, then calls the historic record in the

memory base to carry on the index mapping as well

as the grade decision-making directly. If the records

are not the same, then calls the environmental

information to the index collection to compare

directly, takes the corresponding classification

measure, and stores the top level result and the

environmental information into the memory base to

prepare for future use.

4.2 Detection

The testing process is divided into three categories,

categories of detecting content shown in Figure 2:

(1) Service type detection: to detect the terminal

of the service applicants for those who are using

application softwares. It identifies the service

request is issued by which applications, and thus

obtains the information of the service type. For

example, the text asynchronous transmission service

RESEARCH ON THE CLASSIFICATION CONSULTATIONS AND SELF-ADAPTATION OF QOS IN COGNITIVE

NETWORKS

221

request sent out by the e-mail system, the audio

document synchronized transmission request sent by

the skype software and so on.

(2) Data type detection: it distinguishes between

the service type, while detecting the data type of the

files in the waiting stack. The file block size, block

number, and the format of the waiting files, suffixes

are mainly information for carrying on the type

division.

(3) Terminal type detection: it uses cognitive

radio technology to detect the signal types of the

terminal service, such as measuring radio frequency

band, spectrum and so on. So that it obtains the type

of terminal is the wireless network or wired network,

thus gets the corresponding transport strategy.

Figure 2: The detecting content of QoS.

Through treble network environment's detection

above, we obtain comprehensive result of the

network environment. It calls in the environmental

information to the memory base to compare, so that

it finds the classification decision-making under

similar circumstance in the historical information to

achieve the adaptive purpose.

4.3 QoS Index Map

QoS is a comprehensive indicator to measure the

satisfaction of a service using. Different multimedia

applications require different service requirements,

so these services must be parameterized. At the same

time, different objects have different descriptions.

For example, the QoS requests proposed by clients

are only some brief descriptions, such as poor, in

general, better, the best and so on. It is essential to

map the user QoS to the application QoS in order to

control the QoS consultation and the permission

through the RSVP protocol. Application QoS

re-mapped to the system QoS including network

QoS for consultations. QoS description is that the

user demands of the service quality, because users

may not be computer professionals. Therefore, the

audio QoS description is a brief, unspecific

description of several service levels. But it is

insufficient to describe the complex QoS structure

which depends upon these brief level descriptions

obviously. Therefore, the mapping from user QoS to

application QoS, then to the system QoS, last to the

network QoS must be completed before the QoS

consultation.

4.4 Memory Base

The storage form of the classification historic record

is the database relation table. It should call the

memory data for comparing during the process of

each new task pushed on, so that the classification

step could be simplified. Simultaneously the historic

records which are similar to the network

environment make the logical partition.

First the simple two-dimensional relational table

memorizes each historic record of network service,

whose recording information includes: the network

environment of the terminal services such as service

type, data type, other side terminal service type,

during the time service occurred, the record name

(auto-coding). And it stores with the record name

and the environment name as the primary keys.

Service time, service type, and other fields are as

foreign keys. They connect with the

two-dimensional relational table of other types

named in order to facilitate the logical partition.

Table 1: Memory with the relationship table.

Field Data type

Name of Environment

Text（primary key）

Time of reservoir inflow Time (foreign key)

Type of service

Text（primary key）

Type of data

Text（primary key）

(Each indicator

corresponds to the data)

……

Grading result Numeric

In order to simplify the management of a database

large table, we divide the mechanism into two types

according to the modules: service type, data type, to

carry on logical partition management. Along with

the accumulation of system operation time, the

memory base has a gradual increase in total memory.

If everyone thinks that searching key words will cut

the working efficiency, the partition separate the

ICISO 2010 - International Conference on Informatics and Semiotics in Organisations

222

table into several different table spaces. It uses

“divide and rule” method to support an unlimited

expansion big table, which makes the large table

controllable at the physical level. Dividing the large

tables into smaller partitions, include collection,

excavation and analysis of the multi-dimensional

data such as the service type, data type, run-time, so

the table can be improved in the performance of

maintenance, backup, restore, transaction and query.

The specific reason of using logical partitions as

following: Enhancing usability which means if one

partition of the table can not be used due to system

failure, the remaining good partition of the table can

still be used; reducing the closing time which means

if the system failure affects only a part of the table

partition, then only this part of the district need to be

repaired, it could take less time to repair than the

large table; easier maintaining which means if you

need to rebuild the table, the independent

management of each partition is much easier than

manage a single large table; Balancing I/O which

means we can balance the I/O by assigning different

partitions of the table into different disk; Improving

performance which means query, add, modify and

other operations of the large table can be broken

down into different partitions of the table to execute

much faster. The partition is transparent to the users,

users will not feel the partition existent.

4.5 Features of Self-adaptive

Mechanism

4.5.1 Intelligent Learning Mechanism

The key point of cognitive networks making

improvements of the performance is that it has

specific learning mechanism. The behavior model of

their network elements is: According to different

network conditions, the results of adjusting and

saving will be classified for storage. When meeting

similar problems again, taking into account of

historic record, CN makes appraisals to each kind of

question solution's fit and unfit quality and selects a

relatively optimal strategy. Therefore, based on the

network behavioral model, CN may carry on the

revision and the expansion to the existing QoS

classification mechanism, and satisfy the various

service type and the personalized user demands.

Also CN can predict network behavior by network

status sensors to make reasonable allocation of

traffic flow; can use the smart probe packets to

establish redundant transmission path to support

real-time streaming media business better and better;

can use the resource reservation of network neatly to

set aside resources to achieve optimal use of

distribution according to need and goals.

4.5.2 QoS Needs for the Diverse Business

Before the business start, CN carries on the

end-to-end QoS classification consultation, based on

the actual situation of the network, to make

reasonable adjustments of the QoS business level.

When the service is on operation, we could obtain

the variation situation of end-to-end path status by

using the capacity of information interaction of CN

to carry out QoS re-negotiation. And also it makes

adaptive adjustment of QoS service classification to

ensure the classification of QoS service is reasonable

and reliable.

5 SIMULATION

This article uses the OPNET (Kubera E et al., 2004)

to simulate, compares and analyzes performances of

network delay, scalability, loading between the

existing mechanisms and self-adaptive mechanism

proposed in this paper.

5.1 Delay

The picture below are describing two sets of

simulation results comparing, thick line did not call

the memory base delay, thin line represent for the

use of adaptive mechanism, which means the delay

of calling the memory base. The simulation results

are shown in Figure 3.

0.5

1.5

2.5

0102030

call memory base uncall memory base

Figure 3: The simulation result of delay.

Queuing delay is determined by the service rate and

packet size under the use of self-adaptive

mechanism, that is called delay compared before and

after calling the memory base as shown in Figure 3,

Figure 3 indicated the average time of two nodes

data packets waiting in the service side. By the chart,

RESEARCH ON THE CLASSIFICATION CONSULTATIONS AND SELF-ADAPTATION OF QOS IN COGNITIVE

NETWORKS

223

delay time is about 2.40sec before calling in the

memory base; and the delay time is about 0.25ms

after calling the memory base. For servers, the

shorter delay time, the higher efficiency is in

response to user requests, which will help to

improve the quality of the service. Therefore, the

self-adaptive mechanism plays a significant role in

improving service quality.

5.2 Network Scalability

It named the original network under the use of the

adaptive mechanism Scene 1, and then expand the

number of network clients to twice of the original,

and named it the Scene 2. It shows network delay

curves after running simulations as Figure 4 below.

Figure 4: The simulation result of network scalability.

In the figure 4, the thin curve represents for the

delay curve after using adaptive mechanism, the

thick curve represent for the delay curve after

expanding the number of network terminals into

twice of the thin one which also uses self-adaptive

mechanism. As shown in Figure 4, the two curves

almost overlapped. Although the number of the

network users increased doubled to the original one,

the network delay has not increased significantly.

The good network scalability is a reflect of ensuring

high quality of service during meeting a large

quantity of service in the network, in other words, it

is a necessary condition of satisfying the diverse

service to assure the service delay basically

invariable in dealing with large quantity of

diversified service. Therefore, the use of adaptive

mechanism can supply grading and coordinating

services quickly to ensure smooth network, and it is

a good guarantee of network expansion

5.3 Network Load Capacity

This section simulated two network environments, in

which an adaptive mechanism is not invoked, and

the other is called adaptive mechanism. The figure 5

shows the network load capacity curve after running

simulation.

Figure 5: The simulation results of network load capacity.

As shown in the figure, the thin curve represents for

the original network load curve, the thick curve

represents for adapting the mechanism. the thick line

is higher than the thin line level equally, thick line

index is 2000 - 5000bit/sec compared with thin line

index at each time point, which means capacity of

load improved. A good load capacity of the network

means network can complete the large number of

services and data quantity at the same time point,

that is, shorten the delay time, while ensuring the

efficient data transmission. Therefore, the use of

adaptive mechanism can improve load capacity.

6 CONCLUSIONS

This paper proposes a set of scalable QoS indicators

and designs and makes simulation of self-adaptive

mechanism based on the current situation in the CN.

The mechanism will detect the actual status of the

network and make end to end QoS classification

consultations and adjustments before business start.

We can obtain the changing situation of end-to-end

path status on time by using the information

interaction capability of CN. Then we make QoS

re-negotiation to adjust QoS classification of

business adaptively when the user business is on

operating. At last, the results are classified to import

into memory base due to the characteristic of service.

It will greatly reduce the negotiation time and

improve the efficiency of the network service when

meeting the same situation we could directly call the

classification results of the memory base.

This mechanism focuses on solving the problem

of personalization and diversification users for

protection of QoS in the cognitive network. It fully

enhances the utilization of network resources and

ICISO 2010 - International Conference on Informatics and Semiotics in Organisations

224

meets the complexity, heterogeneity, and reliability

requirements of the new generation information

network, and ultimately it achieves performance

optimization of end-to-end.

In this paper, the design of adaptive mechanism

still need improved, such as increasing memory base

integrated, the fuzzy inquiry function, storage

capacity and so on. Besides, though the simulation

of this self-adaptive mechanism is only a simulation

for three indicators, we could make simulation and

comparison more comprehensive to much more

indicators.

REFERENCES

Zhang, L., Yang, W., Wang, W., Guo, L., Yang, Y., 2007.

A Method to Enhance Service Survivability Based on

Automous Configuration. DOI10. 1109 / IMSCCS. 44.

Wang, F., Jou, F., Gong, F., Goseva-Popstojanova, K.,

Trivedi, K., 2003. SITAR: A Scalable

Intrusion-Tolerant Architecture for Distributed

Services. Proceedings of the Foundations of Intrusion

Tolerant Systems (OASIS’03).

Partha P., Webber, F., Richard E., Intrusion Tolerant

Systems [online]. Available from:

http://itua.bbn.com/isw2000.pdf. [Accessed on 1st Feb,

2010].

Stroud, R., Welch, I., Warne, J., 2004. A qualitative

analysis of the intrusion-tolerance capabilities of the

MAFTIA architecture. In: International Conference on

Dependable Systems and Networks, 2004. 453-61.

Thomas, R., DaSilva L., MacKenzie, A., 2005. New

frontiers in dynamic spectrum access networks. 2005

First IEEE International Symposium: 352 – 360.

Feizi-Khankandi, S., Ashtiani, F., Hassani, S.H., 2007.

Tootoonchi, P. Modeling the behaviour of contending

opportunistic cognitive radios. In: 2007 IEEE

International Conference on Signal Processing and

Communications (ICSPC), Dubai, 24-27 November

2007, 720 – 723.

Sahoo, A., 2002. An OSPF based load sensitive QoS

routing algorithm using alternate paths. In: Computer

Communications and Networks, 2002. Proceedings.

Eleventh International Conference on, 14-16 Oct. 2002,

236 – 241.

Dekeris, B., Narbutaite, L., 2004. Traffic control

mechanism within MPLS networks. In: Information

Technology Interfaces, 2004. 26th International

Conference on, 10-10 June 2004, Vol. 1. 603 – 608.

Lee, G., 2008. Signaling Technogies to Support QoS in

IPv6-based NGN. In: IETF, RFC2711, 43(26): 13 -

19.

Yamada, H., Fukumoto, N., Isomura, M., Uemura, S.,

Hayashi, M., 2007. A QoE based service control

scheme for RACF in IP-based FMC networks. In:

E-Commerce Technology and the 4th IEEE

International Conference on Enterprise Computing,

E-Commerce, and E-Services, 2007. CEC/EEE 2007.

The 9th IEEE International Conference on. Tokyo,

23-26 July 2007, 611 – 618.

Kamatani, O., Kawakami, K., Yamamoto, S., 2008.

Standardization Activities for Resource and

Admission Control Functions in Next Generation.

NTT Technical Review.

Hoang, D.B., Phan, H.T., 2007. A scalable end-to-end

QoS architecture, Communications and Information

Technologies. Proc. ISCIT 07, 17-19 Oct. 2007. 677-

682.

Fgee, E.B., Phillips, W.J., Robertson, W., Elhounie, A.,

Smeda, A., 2008. A Scalable Mathematical QoS

Model for IP Networks, Information and

Communication Technologies. From Theory to

Applications. Proc. ICTTA, 2008. 7-11 April 2008,

1-5.

Chan, H., Fan P.Y., Cao, Z.G, 2005. A Utility-Based

Network Selection Scheme for Multiple Services in

Heterogeneous Networks. Proc. 2005 Interl. Conf.

Wireless Networks, Communications and Mobile

Computing, 2005, 1175 – 1180.

Park, J., Tyan, H., Kuo, C., 2006. Internet Traffic

Classification for Scalable QOS Provision.

Multimedia and Expo, 2006. IEEE International

Conference on. Toronto, 9-12 July 2006. 1221 – 1224.

Junghun P., Hsiao-Rong T., 2006. Internet Traffic

Classfication for Scalable QoS Provision, In:

Multimedia and Expo. 2006 IEEE International

Conference on.

Ducatelle F., Caro G., Ganabardella L. M., 2005. Ant

agents for hybrid multipart routing in mobile Ad Hoc

networks, In: Proceedings of the 2nd Annual

Conference on Wireless On-demand Network Systems

and Services, Moritz, Switzerland, 34(32): 44-53.

Yunbo, W., 2006. The research of Large-Scale multiple

using QoS mechanism based on Overlay. Computer

application technology. 2006.

Suogang L, Jianping W, Ke X, Ying L, 2007. Scalable

multicasting mechanism for unordered QoS. J

Jsinghua Univ (Sci & Tech), 2007, Vol. 47, No.1.

Meng S., Bertrand M., 2007. QoS-aware service overlay

network. In: Proceedings of the Second International

Conference on Communications and Networking,

China, 2008, 739-746.

Kubera, E., Sliwa, J., Zubel, K., Mroczko, A., 2004. Usage

of OPNET simulation tool for validation of routing in

ATM backbone networks of tactical communication

systems. In: Military Communications Conference,

MILCOM 2004. IEEE, 31 Oct.-3 Nov. 2004. Vol. 2.

586 – 592.

RESEARCH ON THE CLASSIFICATION CONSULTATIONS AND SELF-ADAPTATION OF QOS IN COGNITIVE

NETWORKS

225