Performance Management for Efficient QoS provision

and Resource Utilisation in Broadband Internet

Infrastructures

Ilka Miloucheva

, Dirk Hetzer

, Pedro A. Aranda Gutierrez

Salzburg Research, Jakob-Haringer Str.2 /II, 5020 Salzburg, Austria

T-Systems, Goslarer Ufer 35, 10589 Berlin, Germany

Telefonica I+D, Madrid, Spain

Abstract. An important problem of broadband Internet infrastructures is effi-

cient resource utilisation at access points, while keeping the Quality of Service

(QoS) demands of applications stable and optimal. Considering the resource

bottleneck of broadband access networks, there is a need for integrated per-

formance management of such networks providing “feedback” from monitoring

and analysis of traffic, QoS parameters, topology and anomaly effects for the

purpose of short and long term bandwidth resource planning.

This paper is aimed to discus design challenges of advanced performance

management architecture for efficient bandwidth resource planning of broad-

band access networks with monitoring “feedback”. Based on resource planning

and performance data base, the proposed architecture is designed to include

techniques and algorithms for modelling and simulation of optimal resource al-

location strategies in advance considering impact of traffic, topology selection,

and anomaly analysis as well as feedback from QoS analysis.

Application specific QoS monitoring and analysis is used for validation of re-

source allocation planning considering QoS based applications, such as VoIP,

multimedia and Grid. The architecture is derived from the experiences of

INTERMON project for inter-domain QoS analysis studying the impact of to-

pology and traffic (see [1], [2], [4]). A scenario for the integration of QoS and

topology analysis of INTERMON toolkit in the proposed performance man-

agement architecture is described.

1 Introduction

Currently, there is a challenge for flexible broadband Internet infrastructures based on

more efficient and economical usage of resources in order to provide stable QoS

provision of applications [3].

Resource planning for QoS based applications in Internet is a promising area for more

efficient resource utilisation and enhanced QoS support. Especially, for the broad-

band Internet access networks, with their constraints of bandwidth resources and

possibility for flexible Internet connectivity, there is a need of performance manage-

Miloucheva I., Hetzer D. and A. Aranda Gutierrez P. (2004).

Performance Management for Efﬁcient QoS provision and Resource Utilisation in Broadband Internet Infrastructures.

In Proceedings of the 1st International Workshop on Shaping the Broadband Society, pages 41-48

DOI: 10.5220/0001403400410048

 SciTePress

ment toolkits, which provide monitoring and modelling “feedback” for more efficient

resource allocation planning according to the QoS demands of applications.

Today, there are two research directions, which need to be merged together for more

efficient QoS provision and resource usage in broadband Internet:

- advanced performance management systems and tools able to study different

factors, such as topology and traffic, on the QoS of applications

- techniques and technologies for optimal resource allocation which, could be used

for efficient resource planning.

Performance management of application QoS in the area of inter-domain networking

was the focus of INTERMON project [1], [4]. Other works studied traffic and con-

gestion impact on the QoS of applications [5], QoS parameter behaviour dependent

on the link failures [6], effect of inter-domain routing on QoS parameters (see [7],

[8]).

The importance of resource allocation in advance for efficient QoS provision in

Internet was considered in [10], [14], [15], [25]. Resource reservation planning af-

fects the complexity of the routing and path selection process (see [11], [12]), and

requires extensions of protocol, admission control and resource management, as for

instance RSVP extensions for reservation in advance [13]. Efficient techniques and

algorithms for optimal resource allocation were developed using different methods,

for instance operation research algorithms adapted for flexible resource reservation

requests [16] and neuro-dynamic programming using reinforcement learning [17].

Examples for integration of resource optimisation techniques and algorithms in prac-

tical tools are the automated bandwidth allocation planning tool called IconoNET

[18] and Globus Architecture for Reservation and Allocation (GARA) [19], [26].

Algorithms and tools were designed to model and simulate optimal resource alloca-

tion strategies for QoS based applications, for instance real-time [20], and Grid appli-

cations [21].

Although there is a wide research on efficient resource allocation techniques consid-

ering demands of QoS based applications, integration of real performance monitoring

data in the modelling and simulation strategies for optimal resource allocation is still

a challenge [9]. This is a technology gap in today bandwidth resource planning for

QoS based applications. It is mainly based on resource modelling by using of power-

ful mathematical methods (operation research, artificial intelligence, etc), without to

consider different kinds of performance management “feedback” [10].

For bridging the gap and achieve a performance-oriented resource planning, which

could support more efficient QoS of applications based on Internet broadband infra-

structures, we propose a new performance management architecture for operational

and long term resource allocation planning of broadband access networks, aimed to

combine techniques for optimal resource allocation in advance with “feedback” from

performance analysis considering QoS, traffic, topology and anomaly effects.

In section 2, we discuss the design of the advanced performance management archi-

tecture for efficient resource usage and QoS provision in broadband Internet access

networks, aimed at integration of performance monitoring and analysis data into

techniques and algorithms optimising bandwidth allocation for QoS based applica-

tions.

Based on the experiences of the European IST project INTERMON [1], [2], [4], [24],

which result is an architecture for inter-domain QoS analysis integrating monitoring,

modelling and simulation of topology, traffic and QoS parameter data, we propose a

new stage of automated performance management using integrated monitoring tools

and data bases in the area of resource planning of broadband Internet infrastructures

with focus on access networks.

For enhanced flexibility and interoperability, components and data bases of

INTERMON could be integrated in the proposed performance management architec-

ture. Section 3 gives an example scenario for usage of INTERMON topology dis-

covery and QoS analysis in the new context of resource planning.

2 Design of performance management architecture for optimal

resource allocation planning in broadband Internet

The proposed performance management architecture for broadband Internet infra-

structures is focused on the provision of different kinds of performance monitoring

and modelling data as “input” and “feedback” for techniques and algorithms for

optimal resource allocation in advance. The goal is optimal resource allocation plan-

ning of broadband access networks based on learning of monitoring “feedback” of

traffic, topology, QoS and anomaly effects.

Resource allocation planning is directly impacted by the volume and multiplexing of

monitored traffic flows, discovered topology changes and detected anomalies. For

instance, change in topology and traffic has influence on resource utilisation, which

should be considered in the algorithms and techniques for optimal resource alloca-

tion. Anomalies, dependent on the source, could lead to increasing traffic load, which

impacts the resource usage. Therefore study of anomalies and prediction of their

effects (“what if” analysis) is important for the resource allocation planning.

Discovery of alternative topologies and path information (like path quality and stabil-

ity metrics [4]) between broadband access networks could be considered in tech-

niques for optimal resource allocation. A “feedback” from QoS parameter monitor-

ing and modelling could provide insight on the efficiency of the selected resource

allocation strategies and needs for their enhancements.

Algorithms and techniques for resource allocation have to support flexible interfaces

to resource reservation requests of different kinds of QoS based applications like Grid

[19], VoIP [23], real-time [20].

Improvement and adaptation of basic optimisation technologies is required to con-

sider different kinds of resource requests for flexible QoS provisioning [16] and res-

ervation interfaces for specific application classes like Grid [21].

The general design of the performance management architecture for efficient resource

planning in broadband access networks includes:

- performance management components (monitoring and modelling of QoS pa-

rameters, traffic, topology and anomalies)

- resource planning techniques and algorithms for modelling and simulation of

optimal bandwidth allocation considering resource demands of QoS based appli-

cations and performance management “feedback”

- integrated data base for related performance and resource planning data.

The interaction of resource planning and performance management components is

provided based on integrated data base including resource allocation planning infor-

mation related to performance monitoring data describing topology, traffic, QoS and

anomalies. Figure 1 describes the general concept of the proposed performance man-

agement architecture for broadband access networks:

Integrated data base of monitoring and resource

allocation planning data

Algorithms and techniques for modelling and simulation

of optimal resource allocation enhanced with learning of

erformance management data, i.e. „feedback“ from

QoS, traffic, topology and anomaly analysis

Performance management for resource planning

QoS

monitring &

Analysis of

application

classes

Topology

and routing

analysis

Monitoring &

modelling

application

traffic

(SNMP,IPFIX)

Detection

anomaly

impact

Fig. 1. Integrated performance management architecture for resource planning with monitoring

“feedback”

A critical foundation of the proposed performance management architecture is re-

source allocation planning at access networks with possibility for exchange of per-

formance management and resource planning data between different access networks.

For this purpose, distributed performance management data bases of broadband ac-

cess networks could be linked by usage of controlled access.

An example scenario of the proposed performance management architecture is aimed

to provide efficient QoS provisioning and resource allocation of broadband access

networks based on the resource allocation requirements of different kinds of applica-

tions with monitoring “feedback”. The main steps of this scenario are:

1. Monitoring and modelling of traffic of broadband access networks to obtain

forecasting models of resource utilisation considering different traffic classes

which will be considered in the resource allocation planning. For this purpose,

the architecture includes traffic monitoring and analysis tools, as for instance

tools based on IPFIX traffic flow concepts [22].

2. Detection of anomaly effects caused by topology changes, failures and intru-

sions. This is used to model the resource utilisation based on “normal” traffic in

contrast to resource usage due to the anomalies. The topology discovery and

anomaly detection are used to detect “outliers”, which has to be considered in the

resource modelling framework of the bandwidth optimisation techniques [24].

3. Obtaining of strategies for optimal resource allocation planning (daily and

weekly planning) using appropriate algorithms and techniques which consider

the resource modelling based on monitoring “feedback” from traffic analysis

(step 1) and topology discovery / anomaly detection (step 2).

4. QoS monitoring and analysis for validation of optimal resource allocation strate-

gies for different application classes. The QoS monitoring and analysis could be

based on active or passive measurements. When active QoS measurement is

used, the application traffic should be emulated to consider application traffic

patterns, as for instance in the case of VoIP [23]. This allows efficient applica-

tion oriented QoS parameter monitoring [5] and resource planning evaluation.

5. Learning of optimal resource allocation strategies. The feedback from QoS moni-

toring and analysis of applications provides information on how efficiently QoS

is being provided based on the actual selected resource allocation strategies (step

3). “Learning” from QoS monitoring “feedback” allows to improve the resource

allocation planning considering the actual traffic patterns, which are obtained in

operational mode. For this purpose, the concept of neuro-dynamic programming

and reinforcement learning could be applied in the framework of QoS-aware re-

source planning [17].

3 Integration of INTERMON QoS and topology analysis

For enhanced interoperability and more flexibility, the proposed performance man-

agement architecture is based particularly on the INTERMON architecture, developed

in an European IST project [1] and used in large scale broadband Internet infrastruc-

tures in the related area of inter-domain QoS analysis [2]. INTERMON technology

for analysis of QoS behaviour in inter-domain environment is based on the automated

tool interaction and data base integration ([1], [2], [8], [23],[24]) including:

- Monitoring tools and data bases for inter-domain topology discovery, QoS pa-

rameters and traffic, considering SNMP and IPFIX traffic measurements [22]

- Pattern technology to describe QoS behaviour in structures, i.e. “patterns”, for

more efficient QoS analysis (e.g. “linear approximation” [2] for capacity plan-

ning and “outliers” [24] for fault analysis)

- Modelling and simulation environment with automated integration of measure-

ment and topology discovery data to provide “what if” analysis.

INTERMON is aimed at scenarios focussing on end-to-end QoS monitoring and

analysis in inter-domain environment considering specific of applications, and impact

of topology, traffic, network, and resources on the QoS.

The end-to-end performance management strategy of INTERMON tools allows the

integration of selected INTERMON components in the framework of the proposed

performance management architecture for broadband access networks and their fur-

ther enhancements to provide monitoring “feedback” for resource planning. Figure 2

shows an example scenario for usage of INTERMON inter-domain topology discov-

ery to obtain alternative paths and route qualities of end-to-end connection between

two access Internet Service Providers (ISP) connecting Madrid and Salzburg [2], [4].

Fig. 2. INTERMON discovery of alternative inter-domain topologies between two access ISPs

Further INTERMON component, which could be used and enhanced for QoS moni-

toring “feedback” in the proposed broadband performance management architecture,

is the analysis of end-to-end QoS parameter patterns.

Figure 3 shows an example of daily QoS parameter patterns obtained for inter-

domain connection Madrid – Salzburg, discussed in [2].

Fig. 3. Daily analysis of QoS parameter patterns

Based on possibilities to abstract QoS parameter values dependent on the require-

ments for pattern usage, the pattern based QoS parameter analysis could be success-

fully integrated in optimisation algorithms for resource allocation planning.

4 Conclusions

This paper proposed a performance management architecture for optimal resource

allocation in Internet broadband access networks considering “monitoring feedback”

of topology, traffic, QoS parameter and anomaly detection data.

It was shown, that enhanced flexibility and interoperability of the proposed architec-

ture could be achieved based on integration and extension of tools and data bases for

topology and QoS parameter analysis, developed in the European research project

INTERMON. The provision of common interfaces to performance management tools

and data bases, developed in different European research activities, will contribute to

more efficient QoS and resource analysis of Internet broadband infrastructures based

on integrated technologies considering different factors and optimisation goals.

References

1. Advanced architecture for INTER-domain quality of service MONitoring, modelling and

visualisation, INTERMON project, http://www.ist-intermon.org.

2. Miloucheva I., Gutierrez P.A.A., Hetzer D., Nassri A., Beoni M.: INTERMON architec-

ture for complex QoS analysis in inter-domain environment based on discovery of topol-

ogy and traffic impact, IPS Workshop, Budapest, March (2004)

3. Tsuda T.: R&D for a Broadband and Flexible Network Infrastructure, FUJITSU Sci.

Tech. J., 39 (2)p.224-233, http://magazine.fujitsu.com/us/vol39-2/paper11.pdf, December

(2003)

4. Gutiérrez P.A.A., Miloucheva I.: Integrating Inter-domain Routing Analysis in novel

management strategies for large scale IP networks, International IEEE Conference on

Next Generation Teletraffic and Wired/Wireless Advanced Networking (NEW2AN'04),

St.Petersburg, Russia, February (2004)

5. Papagiannaki, Cruz K.R., Diod C.: Network Performance Monitoring at Small Time

Scales, IMC 2003, Oct. (2003)

6. Boutremans C., Iannaccone G., Diot C.: Impact of Link Failures on VoIP performance,

NOSSDAV (2002)

7. Markopoulou A., Tobagi F., Karam M. Assessment of VoIP quality over Internet back-

bones, in Proc. IEEE INFOCOM, New York, NY, pp. 150- 159, June (2002)

8. Gutierrez P.A.A., Miloucheva I.: Analysis of end-to-end QoS behaviour in inter-domain

environment, Inter-domain Performance and Simulation Workshop, Salzburg, February

(2003)

9. Hetzer D. : Measurement based Resource and QoS Planning in Large Scale Internet, Inter-

domain Performance and Simulation Workshop, Budapest, March (2004)

10. Heckmann O., Schmitt J., Steinmetz R.: Robust Bandwidth Allocation Strategies, 10

International Workshop on QoS, IWQoS, Miami Beach, May 15-17 (2002)

11. Gufferin R.A., Orda A.: Networks with advance reservations: The routing perspective.

IEEE INFOCOM Conference (2000)

12. Lewin-Eytan L., Naor J., Orda, A.: Routing and admission control in networks with ad-

vance reservations. 5

International Workshop on Approximation Algorithms for Combi-

natorial Optimization, APPROX'02, LNCS (2002)

13. Schill A., Kühn S., Breiter F.: Design and Evaluation of an Advance Reservation Protocol

on top of RSVP, Broadband (1998)

14. Wolf L., R. Steinmetz R.: Concepts for reservation in advance, Kluwer Journal on Multi-

media Tools and Applications, vol. 4, May (1997)

15. Berson S., Lindell R., Braden R. : An Architecture for Advance Reservations in the

Internet. Technical report, USC ISI, Technical report, USC Information Sciences Institute,

Marina del Rey, CA , July (1998)

16. Erlebach T.: Call Admission Control for Advance Reservation Requests with Alternatives,

Eidgenössische Technische Hochschule Zürich, TIK-Report Nr. 142, July (2002)

17. Brown T.X., Tong, H., Singh S.: Optimizing admission contreol while ensuring quality of

service in multimedia networks via reinforcement learning, Advances in Neural Informa-

tion Processing Systems 11, MIT Press, Cambridge, MA (1999)

18. Frei C., Faltings B. , Melissagris G., Pu P.: IconoNET: a Tool for Automated Bandwidth

Allocation Planning, IEEE/IFIP Network Operations and Management Symposium, Ha-

waii, April (2000)

19. Foster I., Roy A., Sander V.: A Quality of Service Architecture that Combines Resource

Reservation and Application Adaptation 8

International Workshop on Quality of Service,

IWQOS, pp. 181–188, June (2000)

20. Gupta A: Advance reservation in real-time communication services, in Proc.of the IEEE

Annual Conference on Local Computer Networks (1997)

21. Roy A., Sander V.: Advance Reservation API, Scheduling Working Group, Scheduling

Working Document: 9.2, Draft for First Global Grid Forum, 3. March (2001)

22. Raspall F., Tartarelli S., Molina M., Quittek J.: Implementing an IETF IPFIX meter,

Inter-domain Performance and Simulation Workshop, Salzburg, February (2003)

23. Miloucheva I., Nassri, A., Anzaloni A.:Automated analysis of network QoS parameters

for Voice over IP applications, Inter-domain Performance and Simulation Workshop, Bu-

dapest, March (2004)

24. Miloucheva I., Anzaloni A., Müller E.: A practical approach to forecast Quality of Service

parameters considering outliers, First international workshop on Inter-domain perform-

ance and simulation Workshop , Salzburg 20-21 February (2003)

25. Degermark M., Kohler T., Pink S., Schelen O.: Advance reservations for predictive ser-

vice in the internet. ACM/ Springer Verlag Journal on Multimedia Systems, 5(3), (1997)

26. Foster I., Kesselman C., Lee C., Lindell B., Nahrstedt K., Roy A. : A distributed resource

management architecture that supports advance reservation and co-allocation, Interna-

tional Workshop on Quality of Service (IWQoS) (1999)