e2e qos 3g

7/28/2019 e2e qos 3g

1/6

A Functional End-to-End QoS Architecture Enabling

Radio and IP Transport Coordination

J. Olmos, R. Ferrs, O. Sallent, J. Prez-Romero, F. Casadevall

Dept. of Signal Theory and CommunicationsUniversitat Politcnica de Catalunya (UPC)

Barcelona, SpainE-mail: [olmos, ferrus, sallent, jorperez,ferranc]@tsc.upc.edu

Abstract This paper presents a novel end-to-end Quality of

Service (QoS) architecture for wireless networks. The focus is

placed in the coordination of QoS management between Common

Radio Resource Management (CRRM), for heterogeneous radio

access technologies, and the QoS enabled IP-based network

segments. A functional reference QoS framework is introduced

and the envisaged QoS management functions are described. The

innovations proposed in the presented QoS framework arise

mainly from the study of cross effects between the radio and

fixed transport resources management. The proposed QoSframework has been designed considering the current 3GPP R6

networks as well as their long-term evolution (LTE) scenario.

Some illustrative results for a coordinated admission control

algorithm are given as a proof of concept, to show the advantagesof CRRM and IP QoS management coordination.

Keywords- IP-RAN; QoS; CRRM

I. INTRODUCTION

A significant challenge for beyond third generation (B3G)wireless networks is to coordinate the many different types ofnetworks being used for wireless communications. A plethoraof different radio access networks (e.g. cellular networks asGPRS and UMTS as well as wireless LAN hotspots) will haveto co-exist. These different networks ought to be inter-connected in an optimum manner with the ultimate objective toprovide the end-user with the requested services andcorresponding QoS requirements.

On the other hand, it is widely recognized that there will bea migration to all-IP architectures in the mobile networks in themedium-term. This migration creates the challenge to supportQoS over IP networks, and at the same time, the introduction ofnew high speed mobile services claims for a highly efficientuse of network resources. Sometimes it is argued that over-provisioning is the most straightforward way to manage QoS infixed networks. This approach can only make sense in some

scenarios where the cost of adding new capacity is low. Forexample it could make sense in the core network, where fiberoptics (FO) offer high capacity links, [1]. According to [1], inthe backhaul of the access network microwave radio is themost commonly used access technology. With the deploymentof new high speed technologies like HSDPA, the currentbackhaul network, which was basically designed for voiceservices, will have to be upgraded. Operators may be hesitantto invest in additional transport capacity to support the peakdata rates during busy hours, so it becomes important to makean efficient use of the available fixed network resources.

The underlying assumption of this work is that thescenarios where over-provisioning will be applied are restrictedto deep network core, and that in the outer core and radioaccess network big savings can be obtained by implementingappropriate admission and congestion control techniques tosupport QoS. Probably even with these techniques some degreeof over-provisioning will still be needed, but with the expectedmassive introduction of new wireless data and multimediaservices (some of them real-time) it becomes important to keep

it to a minimum. Taking this into account we face a situationwhere limiting resources (bottlenecks) do not have to benecessarily in the Radio Network Layer (RNL), but can be inthe Transport Network Layer (TNL) of the access network;then QoS guarantees depend on both radio and fixed networkresources, and it is possible that transport resources can alsoput important limits to QoS, [2].

The term access network refers to the portion of themobile network comprised between the radio controller (RadioNetwork Controller in UMTS terminology or Base StationController in GSM terminology) and the mobile terminal. Theremaining portion of the network, from the radio controller tothe border gateway, is termed core network. The whole

mobile network can be subdivided into the RNL, whichincludes all the radio network specific entities, and the TNLwhich includes all the entities that depend on specific transporttechnologies, (see Fig. 1). The TNL portion of the accessnetwork is also known as backhaul network.

Throughout this paper, the QoS management functionsspecific to the TNL are designated as Transport ResourceManagement (TRM) functions. Due to the already mentionedreasons, it is very important the joint coordination/interactionof the Radio Resource Management (RRM) entities with theTRM entities; to the point that some of the most important QoSfunctions should be seen as coordinated resource managemententities. The proposed terminology to name this set of entities,

which will be introduced in Section III of this paper, isCoordinated Access Resource Management (CARM)functions.

It is important to remark that 3GPP will not specify whichQoS management functions should be implemented internallywithin the TNL; because the TNL QoS management is seenmainly as an operator dependent option. However, 3GPPclearly states that RNL and TNL interaction should not beprecluded if in the interest of improved system performance,[3]. Our target is to fill this gap by establishing and developingsuitable interaction mechanisms for the QoS management

1525-3511/07/$25.00 2007 IEEE

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the WCNC 2007 proceedings.

7/28/2019 e2e qos 3g

2/6

between RNL and TNL. The proposed E2E QoS architecture isthe one envisaged at the EU IST research project AROMA, [4].

The paper is organized as follows: Section II identifies thekey drivers in the two reference QoS frameworks that havebeen considered. Section III introduces the proposed E2E QoSmanagement framework and describes the envisaged CARMQoS management functions and finally Section IV shows theresults on coordinated admission control obtained by means ofsimulations.

II. E2EQOS APPROACH FOR EVOLUTIONARY 3GPP-BASEDNETWORK ARCHITECTURES

Currently there are ongoing efforts at 3GPP aiming at thespecification of an evolved network architecture that willensure continued competitiveness of the 3GPP technologies forthe future. These efforts are mainly coordinated by the projectsentitled Long Term Evolution (LTE), [5], and SystemArchitecture Evolution (SAE), [6]. In this section we identifythe key drivers for E2E QoS of the medium-term and long-term3GPP network architectures. By medium-term architecture we

understand a 3GPP R6 scenario where the access network hasbeen entirely migrated to IP transport (IP-RAN), and where theIub interface must still be fully supported over IP; whereas bylong-term evolution we mean a scenario aligned with [5],where the RNL consists only of two types of nodes, the AccessGateway (aGW) and the Evolved UTRAN Node-B (eNB), andwhere the Iub interface with its stringent delay constraints, isno longer needed. In both cases we assume that GERAN,WLAN and possibly other non3GPP Radio AccessTechnologies (RATs) are also present in the networkarchitecture with specific degrees of coupling to UTRA. TheIP-backhaul provides the transport for all RATs. Transportresources in the RAN backhaul network can be either assignedindependently to the different RATs or shared among them.

Since a mobile network will always show a high degree ofcapillarity to inter-connect a lot of Access Points to the corenetwork, the problem of achieving an efficient usage ofresources in the TNL will persist in the long-term evolution.

On of the main features of the medium-term networkarchitecture is that IP transport is used within the radio accessnetwork (RAN) from radio network controllers to base stations.The scope of E2E solutions over this medium-term architectureis illustrated in Fig. 1. As key drivers of such solutions weclaim:

Common Radio Resource Management, devoted tocoordinate and optimize the usage of the several

heterogeneous radio access interfaces. Radio and IP transport coordination, that is, the

deployment of QoS control mechanisms (e.g. admissioncontrol, traffic engineering, etc.) considering both radioand IP transport resource availability.

In the medium-term E2E scenario, the most critical part isthe Radio Access Network (RAN). Managing QoS in the corenetwork is viewed as a challenge, but large traffic volumes,statistical laws, DiffServ and sometimes the low marginal costof over-dimensioning make it more manageable; however,managing QoS on a narrow n*2Mbit/s link to a multi-RAT

base station site can be much more challenging. For thesereasons, the coordination between QoS management in theaccess network and in the core network in the medium-termarchitecture is not so critical and is not addressed in this paper.

RNC SGSNBSCMSCServerGANC

PSTN/ISDN

CS-MGW GGSN

IMS Services

External PDN

WLANAP

Node-B

BTS

access network IP-based transport

radio

IPtransportIP transport

IPtransport

Mobile network E2E QoS

RRM RRM RRM

CRRM

RRM

Multi-mode terminals

RadioNetworkLayer

(RNL)

TranspotNetworkLayer(TNL)

IP-based core network

RNC SGSNBSCMSCServerGANC

PSTN/ISDN

CS-MGW GGSN

IMS Services

External PDN

WLANAP

Node-B

BTS

access network IP-based transport

radio

IPtransportIP transport

IPtransport


RRMRRM RRMRRM RRMRRM

CRRMCRRM

RRMRRM

Multi-mode terminals

RadioNetworkLayer

(RNL)

TranspotNetworkLayer(TNL)

IP-based core network

Figure 1. E2E QoS approaches in the medium-term network architecture

PSTN/ISDN

MGW

IMS Services

ExternalPDN

Evolved Node-B

IP-based a ccess networ k IP-based core n etwork

RRC

Border GW

radio

IP nativeIP transport/native?


RRM

aGW

MMEUPE

InterASanchor

PSTN/ISDN

MGW

IMS Services

ExternalPDN

Evolved Node-B

IP-based a ccess networ k IP-based core n etwork

RRC

Border GW

radio

IP nativeIP transport/native?


RRMRRM

aGW

MMEUPE

InterASanchor

Figure 2. E2E QoS approaches in the long-term network architecture

The main aspects of the long-term architectural frameworkare summarized in Fig. 2. Although Fig. 2 only reproducesnetwork architecture for E-UTRAN, we assume that the IP-backhaul will also provide the transport for UTRAN, GERAN,WLAN and other non3GPP RATs. In this case, networkarchitecture is more aligned to 3GPP Evolved-UTRAN, [5], sothat new solutions for QoS and mobility management as wellas a new 4G radio interface can be introduced. In such long-term vision context, E2E QoS solutions should rely on thefollowing key drivers:

Multi-cell RRM. Current design trends towards flat

network architectures make most RRM functionality to bemoved towards the edge of the network (i.e. evolvedNode-B). However, coordination of those evolved Node-Bs in terms of overall network QoS performance remains acrucial aspect (e.g. admission control decisions not justrelying on the status of a single node-B because of thepotential movement of the terminal to another NodeB,support of fast handover mechanisms between NodeBs,etc.).

Introduction of QoS and mobility solutions based on IETFprotocols.


7/28/2019 e2e qos 3g

3/6

Radio and IP transport coordination, that is, thedeployment of QoS control mechanisms (e.g. admissioncontrol, traffic engineering, etc.) considering both radioand IP transport resource availability.

III. A FUNCTIONAL REFERENCE QOS FRAMEWORK

Our goal in this section is to define a functional framework

for a coordinated management of radio and transport resourcesthat leads to efficient resource utilization and to better supportof QoS guarantees.

Focusing on the IP-RAN and in the medium-termarchitecture, Fig. 3 highlights the different pools of resources tobe managed by QoS supporting functionalities. Also in Fig. 3the acronyms of the QoS functionalities identified for bothRRM and TRM are shown. In principle these functions couldbe conceived separately for both QoS management domains.With respect to RRM, the proposed functions take intoaccount, as a guideline, the functions proposed in [5], [7], [8],and were already studied and validated in the EVERESTproject, [9], [10]. With respect to TRM, we have identified the

proposed functions by taking into account recent researchefforts towards the introduction of QoS management in thecontext of DiffServ IP networks, [11], [12], [13] and byrealizing that the desirable QoS resource managementobjectives in the TNL claim for a set of functions that mirror,to certain extent, those already familiar at the RNL. Theimplementation of some of the proposed QoS functions can beRAT specific; for example the Radio Link Control functionmay encompass a variable range of RAT specific capabilities,typically (but not necessarily) including power control and linkadaptation mechanisms. The proposed QoS functions are: RATSelection (RS), Admission Control (AC), Congestion Control(CC), Bearer Selection (BS), Mobility Control (MC), RadioLink Control (LC), Radio Packet Scheduling (RPS), TNL

Route Control (RC) and TNL Packet Scheduling (PS).

RRM RAT n

MC

RPSLC

AC CC

BS

Radio Controller

(RNC, BSC, GANC)

Base Station

(NodeB, BTS, AP)

SGSN GGSN

Mobile

Terminal

Radio Resources IP Transport Resources IP Transport Resources IP Transport Resources

TRM

R5/R6

CC

PSRC

BS

AC

RRM RAT2

MC

RPSLC

AC CC

BS

RRM RAT1

MC

RPSLC

AC CC

BS

CRRM RS

RRM RAT n

MC

RPSLC

AC CC

BS

RRM RAT n

MC

RPSLC

AC CC

BS

Radio Controller

(RNC, BSC, GANC)

Base Station

(NodeB, BTS, AP)

SGSN GGSN

Mobile

Terminal

Radio Resources IP Transport Resources IP Transport Resources IP Transport Resources

TRM

R5/R6

CC

PSRC

BS

AC

RRM RAT2

MC

RPSLC

AC CC

BS

RRM RAT2

MC

RPSLC

AC CC

BS

RRM RAT1

MC

RPSLC

AC CC

BS

CRRM RS Figure 3. Separate QoS functionalities for RRM and TRM

When trying to jointly optimize the use of resources, astrong interaction can be predicted among the AdmissionControl, Congestion Control and Bearer Selection of bothpools of resources; for this reason the proposed trend is to viewand design that functions as CARM functions. With respect tothe Mobility Control, it is foreseen that the transport network

will also have to be informed and checked for availableresources before making the decision of a radio layer handover,and so this function is also included among the CARMfunctions. RAT Selection, which originated as a CRRMfunction, should also be included as a CARM function due tothe possible dependence of RAT selection decisions on theTNL link loads. Link Control, on the RRM side, as well asRoute Control on the TRM side are considered specialized

functions that only make sense within their own scope (or poolof resources). The same can be said about Packet Scheduling inboth RRM and TRM. It is in the packet scheduling functionwhere some important QoS decisions (like priority) areseparately enforced in the relevant nodes from RNL or TNL.

Fig. 4 summarizes the name and scope of all the proposedQoS management functions, while Fig. 5 shows the role of thecoordinated QoS functions in the medium-term and long-termarchitectures.

Bearer

Selection

Radio Packet

Scheduling

CongestionControl

Link Control

RRM functions TRM functions

TNL Packet

Scheduling

Route

Control

Admission

Control

Mobility

Control

CARM functions

RAT Selection

Bearer

Selection

Radio Packet

Scheduling

CongestionControl

Link Control

RRM functions TRM functions

TNL Packet

Scheduling

Route

Control

Admission

Control

Mobility

Control

CARM functions

RAT Selection

Figure 4. Proposed QoS management functions

Radio Controller (RNC,BSC, GANC) or aGW (LTE)

Base Station (NodeB,

BTS, AP) or enB (LTE)MobileTerminal

Radio Resources IP Transport Resources

TRM

PSRC

RRM RAT n

RPSLC

RRM RAT2

RPSLC

RRM RAT 1

RPSLC

CRRM

CARM

MC

CC

AC

BS

RS

Radio Controller (RNC,BSC, GANC) or aGW (LTE)

Base Station (NodeB,

BTS, AP) or enB (LTE)MobileTerminal

Radio Resources IP Transport Resources

TRM

PSRC

RRM RAT n

RPSLC

RRM RAT n

RPSLC

RRM RAT2

RPSLC

RRM RAT2

RPSLC

RRM RAT 1

RPSLC

RRM RAT 1

RPSLC

CRRM

CARM

MC

CC

AC

BS

RS

Figure 5. Role of CARM QoS functions in the 3GPP architecture

In the next paragraphs the objectives of each of theenvisaged functions in the QoS architecture are described.

The CARM functions are:


7/28/2019 e2e qos 3g

4/6

1) RAT selection: This function is in charge of selectingthe most appropriate RAT, either at call establishment orduring the session life-time through the so-called verticalhandover procedure, given the requested service and QoSprofile. The RAT selection decision could be influenced bymany factors, including non-technical issues like theOperators policies or business model. It is considered aCARM function due to the need to take into account the link

load at the TNL before making a RAT selection. For example,the operators initial choice for a voice service request could bethe GERAN RAT, but an overloaded GERAN IP-backhaulcould redirect the RAT selection towards UTRAN.

2) Bearer Selection: Bearer selection is in charge ofselecting the required resources to support the requested QoSprofile at the radio and transport bearer services. This impliesthe configuration of new radio and transport bearers given therequested QoS profile and selected RAT. It also includesdynamic mapping of requested QoS parameters to the transportQoS parameters.

3) Admission Control: maintains information ofavailable/allocated resources in both the radio and the IPtransport network and performs resource reservation/allocationin response to new service requests, at call establishment orduring vertical/horizontal handover, with a given QoS profile.From the radio point of view it takes into account, for example,the interference level and the availability of codes (in aWCDMA radio interface), and from the transport network itcan take into account, for example, the current occupation ofthe bottleneck link. Depending on the implementation, it ispossible to envisage a variable degree of coupling betweenradio and IP transport admission control. For example the typeof response, at a new QoS bearer request from the radio part tothe TNL, may range from a simple success/reject message toa more sophisticated response in the form of a prioritized list of

potential cells where the new QoS bearer request could beadmitted. Prioritization would be done based on the currentload in the fixed links connecting to every managed NodeB (orAP in general).

4) Congestion Control: It is in charge of taking the actionsrequired to handle overload events in the radio or transportnetwork side. This function will implement the Operatorspolicy for congestion situations, for example, give priority toreal-time/premium/business users over non-real-time/consumerusers, etc, and take the necessary actions to reduce the durationof the congestion event. The methods used to handlecongestion include a range of options, for the radio and for thetransport part, which are operator/implementation dependant.

Congestion Control needs coordinated actions from the radioand transport resource management. As an example, thepossible actions range from changing the Transport FormatCombination Set (TFCS, UTRAN specific) to some users inthe RRM part, to setup alternative routes or enforce link-sharing strategies for packet scheduling in the transport part.Congestion detection could be based entirely on RNLmechanisms as is currently assumed by 3GPP (reactiveprocedure), or the TNL could set some flag or even givedetailed information about the congested links (proactiveprocedure).

5) Mobility Control: This function is basically in charge ofhandover. As stated in [5], handover decisions take intoaccount measurements from the UE and the node-B and maytake other inputs, such as neighbor cell load, trafficdistribution, transport and hardware resources and Operatordefined policies into account. We envisage that the transportresource availability could be checked before making ahandover, since it is possible that a given cell with free radio

resources can not accept a handover call due to a congestedlink in the transport network. It is also possible that transportresource availability can influence the decision about which isthe optimum cell to direct the handover to.

The objectives of the RRM and TRM specific functions aredescribed in the next paragraphs. On the RRM side we haveRadio Link Control and Radio Packet Scheduling.

6) Radio Link Control: It is in charge of dynamicallyadjusting the radio link parameters of the mobile terminals inorder to preserve the QoS for established sessions. Thisfunction will typically include power control and linkadaptation mechanisms. Power control aims at dynamicallyadjusting the power transmitted by all the terminals in a givencell. The required power for each user depends on severalfactors like radio-link propagation losses, amount ofinterference in the cell and type of service and mobility of theuser. Link adaptation functions dynamically adjust modulationand coding to maximize the throughput given the radio channelconditions.

7) Radio Packet Scheduling: This function is in charge ofmaximizing resource occupation by scheduling packets forestablished sessions taking into account several factors, like theQoS of the session, the interference level of the cell and thechannel quality for the particular user. Radio PacketScheduling is a short term strategy that tries to use freeresources that could otherwise remain underutilized.

On the TRM side we have TNL Route Control and TNLPacket Scheduling:

8) TNL Route Control: This QoS management function isin charge of selecting the optimum routes in the transportnetwork to guarantee the efficient use of TNL resources andthe QoS requested by the TNL IP bearers. This function will beapplied when setting up new QoS IP bearers and it is alsoenvisaged that this function should continuously monitor thelink utilization and buffer occupancy of the transport networknodes in order to prevent congestion and maintain efficient useof network resources. The implementation of this function willrelay on appropriate load balancing techniques, path

establishment with QoS constraints (like Constraint Routing-Label Distribution Protocol, CR-LDP, [14]), as well as networkresiliency mechanisms in case of link/node failures.

9) TNL Packet Scheduling: This function is in charge ofimplementing, at the IP transport network nodes, theappropriate QoS queuing decisions so the different flows (oraggregates of flows) receive the right QoS treatment at everynode. For that purpose the packets are marked at the ingressnode with a mark that identifies them as belonging to a givenQoS Behavior Aggregate (assuming a DiffServ scheme, forexample). The implementation of TNL packet scheduling could


7/28/2019 e2e qos 3g

5/6

range from simple priority queuing to the more sophisticatedlink-sharing techniques.

In the long-term scenario, where the functions of the RNChave moved to an evolved node-B, the core IP transportnetwork and the access IP transport network are expected tomerge, giving raise to a single IP transport network linking thegateway (i.e. AGW) and the node-B (i.e. eNB). At thatmoment, the QoS entities managing resources at the corenetwork and at the access network could also merge together.For that reason the proposed QoS functions should be able tofit both the medium-term scenario and the long-term scenario.

A. Realisation of the CARM model within current 3GPPnetwork structure

Actually the protocol structure of 3GPP radio accessnetworks consists of two main layers: RNL and TNL. AllUTRAN related issues are visible only in the RNL, and theTNL uses standard transport technology that is selected toprovide the transport services for UTRAN, but without anyUTRAN specific requirements.

The conception of a TNL separated from the RNL is basedon the principle that the layers, and the user and control planestherein, are logically independent of each other and, if needed,protocol layers, or the whole protocol stack in a plane may bechanged in the future. Over such a basis, for the IP transportoption, the basic interaction of the RNL and TNL layerscovered so far by 3GPP is limited to the specification of thestandards needed for user data transport protocols (e.g. UDP/IPstack) over the different interfaces (e.g. Iub, Iur and Iu forUTRAN). However, as already argued in the introduction ofthis paper, this layered model does not preclude thedeployment of control mechanisms between TNL and RNLaimed at achieving a better coordination of the resourcemanagement functionalities allocated in each layer. Hence, the

support of the CARM model may be realized through anextension of the current interaction model between both layers.

In this sense, different interaction models can be envisagedbetween the resource management entities within RNL andTNL layers to realize such coordinated QoS functions. Theseinteraction models would rely on functionalities such as: (1)exchange of reporting information (e.g. transport loadmeasurements of Iub links are available at the RNC to be usedin a coordinated admission process); (2) decision supportmechanisms (e.g. TRM entities in the TNL can provide, ondemand, a prioritization of the candidate cells based ontransport criteria during a handover process); and (3) dynamicconfiguration (e.g. some RRM/CRRM parameters in the RNL

or some TRM parameters in the TNL are dynamically adjustedso that a common strategy can be accomplished on the basis oflocal decisions on each different type of resource managemententity). The development and assessment of which of thoseinteractions may prove to be useful in terms of enhancedsystem performance is out of the scope of this paper andconstitutes one of the research activities currently addressedwithin the AROMA project, [4].

IV. COORDINATED ADMISSION CONTROL

One of the key functionalities in the CARM framework isthe coordinated admission control. In an heterogeneousnetwork environment offering flexible services, theCoordinated Admission Control could consider, not only if aparticular QoS bearer request can be admitted or not, but alsounder which conditions it can be admitted (by allowing onlycertain transport formats, for example, in the case of UTRAN)and in which radio access point it could be admitted (assumingthat there are several candidates). Among others, this decisioncan be influenced by the next criteria:

Traffic load at the candidate radio access points

Link occupations in the path between the candidateradio-access points are their corresponding controllers

In the present simulation results we concentrate in thesituations where decisions are determined by the linkoccupation in the path between the radio nodes and theircontrollers. The simulated scenario assumes an UTRAN IP-based access network where there is a single link (which acts asa bottleneck) in between a NodeB and its controller. We alsoassume that there are not alternative paths, so the questionabout which radio-access point should accept the request is outof scope. In this context we simulate three different TNLadmission strategies; among them a Coordinated AdmissionControl (CAC) that takes into account the average bottlenecklink utilization and based on that, decides which profiles ofQoS bearer request can be admitted. The link utilization isdefined as the percentage of time that the link is busy, whereasthe normalized goodput is defined as the ratio of the aggregateduser traffic generated by the tested NodeB over its linktransport capacity. Notice that the link utilization is alwayshigher than the normalized goodput due to TNL protocoloverheads and RNL retransmissions. During the simulation, the

bottleneck link utilization is sampled once every 0.1 s andaveraged by an exponentially decaying low-pass filter with atime constant of 1 s. The three different TNL admissionstrategies considered are:

1) There is not any admission control in the TNL (allrequests are granted). The graphs for this situation are labeledNo_AC. As a result of this strategy we have an impact onservice degradation in the form of a higher voice frame errorratio, Fig. 7.

2) There is admission control in the TNL, but it onlysupports the admit/reject mode. The graphs for this situationare labeled Hard_AC(=90%), where 90% is the threshold,in the average bottleneck link utilization, at which the hardadmission control rejects all new TNL admission requests. Inthis case we reduce the voice frame error ratio at the expenseof a restrictive admission scheme, Fig. 6 and Fig. 7.

3) There is admission control in the TNL and it supportsthe coordinated mode: the TNL provides the current linkutilization and this indicator helps in admission decisions. Thegraphs for this situation are labeled Coord_AC. In this caseall new voice and data connection requests can be admitted(Fig. 6) as long as, when link utilization is high, all the newadmitted data calls use a reduced transport format (i.e. areduced bit rate at the radio interface). Notice also, in Fig. 7,


7/28/2019 e2e qos 3g

6/6

that the voice frame error ratio is kept similar to the case ofHard_AC(=90%). As a trade-off we obtain an increaseddelay for data services, Fig. 8.

AdmissionAcceptanceRatio(%)

0.7

0.75

0.8

0.85

0.9

0.95

1

1.05

0.30 0.40 0.50 0.60 0.70

Normalized goodput

No_AC

Hard_AC(=90%)

Coord_AC

Figure 6. Admission acceptance ratio versus normalized goodput

VoiceFrameErrorRatio(%)

1

1.5

2

2.5

3

3.5

4

0.30 0.35 0.40 0.45 0.50 0.55 0.60

Normalized goodput

No_AC

Hard_AC(=90%)

Coord_AC

Figure 7. Voice frame loss ratio versus normalized goodput

MeanP

acketDelayfor

Interac

tiveTraffic(s)

0

0.1

0.2

0.3

0.4

0.5

0.6

0.30 0.35 0.40 0.45 0.50 0.55 0.60

Normalized goodput

No_AC

Hard_AC(=90%)

Coord_AC

Figure 8. Mean packet delay for interactive traffic versus normalized

goodput

V. CONCLUSIONS

In this paper we have developed a novel E2E QoS networkarchitecture for heterogeneous mobile networks. By enablingcoordination between the RNL and TNL QoS resourcemanagement functions, the proposed architecture aims atachieving highly efficient resource utilization in the IP-basedbackhaul network in order to support high speed data serviceswith minimum over-provisioning. A set of CARM functionshas been identified and their roles have been described. This setof functions is seen as the basis for RNL and TNL QoScoordination. The proposed QoS architecture has beendesigned to be consistent with the 3GPP network architectureevolution. Finally, some simulation results have been presentedto show the gains that can be obtained by introducing acoordinated admission control.

ACKNOWLEDGMENTS

This work is part of the EU IST-AROMA project, [4],partially funded by the European Commission under the ISTframework. The Authors would like to acknowledge thecontributions of their colleagues from the AROMA project.The work is also partially funded by the Spanish ResearchCouncil under COSMOS (Ref. TEC2004-00518) grant.

REFERENCES

[1] A.Bolle, H.Hertbertsson, Backhaul must make room for HSDPA,Wireless Europe, Feb/Mar 2006

[2] M.Sgfors, V.Virkki, Overload control of best-effort traffic in theUTRAN transport network, IEEE Vehicular Technology Conference

2006-Spring, pp.456-460, vol.1

[3] 3GPP TR 25.913, Requirements for Evolved UTRA (E-UTRA) andEvolved UTRAN (E-UTRAN) (Release 7)

[4] IST-AROMA project web page http://www.aroma ist.upc.edu

[5] 3GPP TR 25.912, Feasibility study for evolved Universal Terrestrial

Radio Access (UTRA) and Universal Terrestrial Radio Access Network(UTRAN)

[6] 3GPP TR 23.882, System Architecture Evolution: Report on TechnicalOptions and Conclusions (Release 7)

[7] 3GPP TS 23.107, Quality of Service (QoS) concept and architecture(Release 6)

[8] 3GPP TS 23.207, End-to-end Quality of Service (QoS) concept andarchitecture (Release 6)

[9] IST-EVEREST project web page http://www.everest ist.upc.es

[10] J. Perez-Romero, O. Sallent, R. Agusti, M.A. Diaz-Guerra, Radioresource management strategies in UMTS, Ed. John Wiley & Sons,

2005

[11] J.Lakkakorpi, Simple measurement-based admission control forDiffServ access networks, in Proc. ITCom 2002, July 2002, Boston,

SPIE.

[12] E. Mykoniati, C. Charalampous, P. Georgatsos, T. Damilatis, D.Goderis, P. Trimintzios, G. Pavlou, D. Griffin, Admission control for

providing QoS in DiffServ IP networks: the TEQUILA approach. IEEECommunications Magazine, Jan. 2003

[13] P. Trimintzios, I. Andrikopoulos, G. Pavlou, P. Flegkas, D. Griffin, P.Georgatsos, D. Goderis, Y. TJoens, L. Georgiadis, C. Jacquenet, R.Egan, A management and control architecture for providing IP

differentiated services in MPLS-based networks. IEEECommunications Magazine, May. 2001

[14] Y. Guo, Z. Antoniou, S. Dixit, IP transport in 3G radio accessnetworks: an MPLS-based approach, IEEE Wireless Communicationsand Networking Conference 2002, pp.11-17, vol.1


e2e qos 3g

Documents

Transcript of e2e qos 3g