Printer Friendly


Byline: Waseem Abbas Nasim Abbas and Uzma Majeed


In this paper we proposed user satisfaction architecture for packed-switched services in 3rd generation cellular networks for end-to-end Quality of Service (QoS) provisioning in Diffserv IP network environment. The paper focuses on mapping of QoS classes from Diffserv to UMTS Admission control Buffering and scheduling. The Diffserv code point was utilized in the end-to-end quality of service provisioning to differentiate various type of multimedia real time traffic. This paper proposes the

WCDMA based prioritized uplink call admission control that combines the QoS tolerance and service differentiation for data and multimedia traffic by priority. This algorithm reserves some bandwidth margin number of users and power consumption to decrease handoff failures to give preference to high priority calls such as handoff calls additionally Low latency queuing (LLQ) is implemented to improve the quality of service. LLQ is used with the key idea of mapping voice and video traffic in two different queues but at the same time using priority queuing within LLQ for both voice and video traffic for all other QoS classes. The results obtained from simulations demonstrate that proposed algorithm meet the QoS requirement.

Keywords: CAC DiffServ EURANE LLQ NS- 2 QoS Scheduling


Third generation mobile communication technologies have gone through a very rapid growth and Universal Mobile Telecommunication System (UMTS) has emerged as a leading standard for the provisioning of third generation cellular networks. Wireless 3G network is anticipated to convey multimedia traffic such as VOIP video telephony data and other applications. The requirement of QoS while transmitting multimedia traffic on same medium is one of the major issues in order to design and analyze 3G wireless networks. Service classification and efficient resource management are quite demanding tasks due to increasing number of applications and ubiquitous bandwidth limitations for multimedia applications such as voice and video streaming.

Key advantage of UMTS is its ability to provide different services with QoS guarantees [1]. IETF standardized two different mechanisms for providing QoS in IP networks Diffserv [2] and InterServ [3]. DiffServ can be implemented in UMTS with no management complexity while InterServ have scalability and complexity problems. DiffServ uses Codepoints known as Differentiated Services code points (DSCP) attached with IP header of a packet to classify traffic with different PHBs (Per Hop Behavior) at the boundaries of the network. PHB definitions do not specify any particular implementation mechanism and therefore implementation problem of PHB has gained noteworthy attention. On the other hand QoS concept and architecture for UMTS network as specified in 3GPP [1] focuses on QoS signaling from user equipment to GGSN only and it does not support the QoS mechanisms for data transport.

For end-to-end QoS provisioning in UMTS it is required to map the IP traffic classes to the UMTS network and propose a QoS mechanism for the transport of user data.

Existing work on end-to-end QoS in UMTS [4] have analyzed advanced mapping between voice and video telephony but do not take other UMTS traffic classes into account. Other work investigates the access control (AC) in both the wired and wireless network part a multi-class AC scheme on the wired network part was proposed in [5] which supports the DiffServ approach. Authors in paper [6] have combine WFQ and LLQ but the main flaw of this idea is the property that delay in video conferencing could be reduce but at the same time voice traffic got the maximum delay time. The SEACORN [7] project contributed in the development and implementation of wireless part i-e. Radio Resource management RRM algorithms for QoS provisioning in the UMTS network. One of the major contributions of the SEACORN project is a UMTS extension for the Network Simulator 2 (NS-2) [8] known as Enhanced UMTS Radio Access Network Extension (EURANE) [9]. We choose EURANE as the tool to simulate an E2E UMTS QoS scenario.

This paper proposes detailed algorithm in which concept of mapping voice and video telephony to different QoS classes idea of implementing LLQ scheduler and packet treatment strategies for the UMTS core network and analyzes them in a large set of simulation experiments. The latter focus on revealing the impact of non-real time services on real-time services in different scenarios with and without full QoS provisioning mechanisms. The mapping will be done in the GGSN and scheduling policing and multiple queuing mechanisms will be implemented in the UMTS Core Network.

This paper is structured as follows. Section II presents the proposed algorithms and mechanism Section III explains the simulation scenario and simulation results. Section IV concludes the paper and describes direction of future work.


The aim of this section is to give the brief idea of development of E2E QoS algorithm in terms of QoS mapping Call Admission Control policing buffering and scheduling.

General Assumptions and optimization target

In our proposed algorithm which is depicted in Figure 1 for provisioning of E2E QoS algorithms. There are three main components of UMTS System architecture User Equipment UE or Terminal Equipment TE UMTS Terrestrial Radio Access Network (UTRAN) and Core Network. Core network is connected to external IP domain network in which there are edge router and application servers which can send only one type of application data. GGSN and SGSN are the entities of UMTS core network and UTRAN contains Radio Network controllers RNCs and Node-Bs which are responsible for all functionality related to radio access. In our scenario both external network and UMTS network supports DiffServ. Application Servers send downlink data packets to the UEs and these packets will be controlled by Edge router GGSN and SGSN before it is segmented to RLC PDUs in RNC and forwarded to Node B and in our proposed algorithm only PDUs transmitted via Dedicated Transport Channel to UEs are considered.

The basic assumptions on the traffic model are given in Table 1 and sessions of the individual types of traffic arrive according to Poisson Processes.

Table 1: Traffic Model assumptions

Traffic Type###Application###Application###Holding

###Level###Time Model



Conversational###Voice###EXP On/Off###Exponential

###Streaming###Video###EXP On/Off###Exponential





In our scenario the bottleneck of the downlink transmissions is presumed to be the outgoing link from the GGSN to SGSN hence the design target is to enhance the bottleneck link consumption i-e minimizing session blocking rate and higher link throughput while maintaining the E2E QoS requirements of each UMTS class. The boundary conditions of each service type are shown in Table 2.

Table 2: UMTS QoS Requirements of each service class

Service Conversational Streaming###Interactive Back-



Loss###less than 10-2###less than 10-1###less than 10-3###less than 10-3



End###less than 100 ms###less than 250ms###N/A###N/A


DiffServ to UMTS QoS Mapping

The 3GPP standard defines a layered architecture for the provisioning of end-to-end QoS. To understand QoS for a certain network a Bearer Service (BS) with appropriately given functionalities has to be implemented from the source to the destination of a service and contains all characteristics to facilitate provisioning of a constricted QoS. UMTS bearer service attributes from a QoS profile and defines the level of service provided to the UMTS BS user.

UMTS specification [10] defines four QoS classes in

UMTS packet domain: Conversational Streaming

Interactive and background. These classes are categorized on the basis of Delay factor The streaming and conversational classes are the most delay sensitive classes and defined for real time applications while interactive and background classes are meant for delay-insensitive

classes and treat the traffic without restricted transfer delay requirements e-g Transmission control Protocol.

Table 3: UMTS vs. DiffServ Mapping

Application###UMTS Service###DiffServ###DSCP



Streaming###AF 12###12

Web HTTP###Interactive###AF21###18



QoS mapping between IP DiffServ and UMTS services as described in Table 3 is very essential for keeping appropriate end-to-end delay for real time traffic.

The Expedited Forwarding Per Hop Behavior (EF-PHB)

[11] is known for low packet loss low jitter and less delay services. In case of EF traffic is treated at minimum service rate for both short and long intervals. The EF packets that go beyond the particular arrival rate will be dropped in advance. This DiffServ class is very suitable with the traffic of conversational class as traffic of conversational class i-e VOIP is very delay sensitive but comparatively insensitive of packet loss. IETF specifies Assured Forwarding (AF) PHB group in [12].

AF-PHB gives guarantee of delivery provided that traffic will not surpass subscribe rate. In case of congestion traffic that goes beyond the particular rate has high probability of being dropped.

Four different independent classes with three different dropping precedence levels each were implemented in AF- PHB for IP packet delivery. The corresponding PHB is called AFij in which i represent AF class and j shows dropped precedence. Every class is implemented with different shares of bandwidth with different configurations of buffering and dropping. The AF-PHB class supports streaming class as it is delay sensitive but less than conversational class. The Interactive class is mapped to AF31 with no explicit delay boundary and do not have extra requirements apart from reliability and in last Background class of UMTS network is mapped with default PHB class i-e Best Effort

2.3 Call Admission Control

The main purpose of CAC is to restrict the interference by constraining the capacity of new calls admitted in the network as traffic load offered by uplink and downlink transmissions is different so CAC need to perform separately for each other. The preconditions of uplink and

downlink CAC must be attained by each new user while entering into the system. Due to mobility in wireless networks CAC turn out to be more complicated.

The call dropping happens in the network in most of the case during handoff in which the UE is moving away from the covered area of one cell and entering in the region covered by another cell. During the process of call handoff there is a possibility of not getting enough resources in the next cell due to limited resources in wireless networks [13]. Thus there should be a priority base mechanism for treating new calls and handoff calls in terms of resource allocation. Normally higher priority is given to handoff calls than new calls [14]. Handoff priority-based CAC schemes can be classified into two broad categories NCAC (Number based) and ICAC (Interference based).

NCAC is the number based call admission control that considers limited number of channels and the QoS parameter bit-error-to-interface ratio for NCAC is mapped to the number of users contained in the network while ICAC algorithm does not limit the number of channels and

it uses SIR parameter to accept or reject the newly originated call. ICAC have been further classified in three schemes [15].

Wideband Power Based CAC Throughput Based CAC and Signal to Noise Based CAC.


Figure 2 shows flow chart for handoff prioritization algorithm which is basic logic.

i. Estimate load factor threshold thresh (in term of No. of user BW Power consumption) [16].

ii. Calculate the load increase of incoming call i along with the load factor of current cell prior to accept

incoming call [17].

iii. Calculate the up to date load of target cell and compare it with threshold of load factor for newly originated call. The new call seeking admission will only be admitted in target cell when sum of load increase and current cell load factor is equal to or less than threshold of desired load factor for incoming call or else the incoming call will be rejected or queued. Soft handoff calls which are in queue can be admitted if enough bandwidth is available [21] otherwise rejected because of timeout.

Policing buffering and scheduling.

Policing buffering and scheduling algorithms deals with packets these mechanisms are executed each time a packet is sent or received. Policing and buffering algorithms decide whether or not an incoming packet is accepted. A full implementation structure of these algorithms is depicted in figure 3.

DiffServ Code Point DSCP marking is implemented by a policer at the edge router and also two code points (red and green) are assigned to each packet it means that IP packets with different color can be handled later with different physical/virtual queues and different treatment.

Multiple queues have been designed to provide different treatment to different UMTS classes according to their DSCP field. Since Random Early Detection (RED) [18] queues only impact TCP based traffic streams and cannot help in congestion control of UDP based traffic two drop- tail queues are designed for UDP based real time services Conversational (C) and Streaming (S) while RED queues are for TCP based non-real time services Interactive (I) and Background (B). The queue size was optimized according to their different delay and packet loss requirements simultaneously. Preliminary simulations show that in the RED queue design multiple virtual queues did not improve the QoS for non real time traffic i.e. the "green" packets of Web/Http traffic actually experienced higher packet dropping than the "red" packets. Hence the coloring was not use in our QoS concept.

Scheduling improves the capacity of system by sharing the common recourses dynamically among different classes. The main responsibility of packet scheduling algorithms is to assign available wireless network resources to group of mobile users who are allowed to receive data. When a link is congested there is a need of scheduler and queuing mechanisms are busy though Scheduler has to decide for the next queue to treat first. If there is no congestion on link then packets are sent as they arrive.

It is obvious that each UMTS service class should be assigned a priority levels from 1 to 4 due to the different delay sensitivity levels in which 1 is the highest. We choose Low Latency Queuing LLQ because the relative importance of each class can be modified easily according to change of traffic mix hence it is more flexible. LLQ scheduling is an enhancement of Class Based Weighted Fair Queuing CBWFQ in which one or more strict priority queues are integrated to solve latency issues in multimedia applications that makes it ideal for jitter and delay sensitive applications. LLQ have properties of both PQ and CBWFQ. LLQ queues delay sensitive traffic and allow it to be dequeued and if there are no more packets in that queue then WFQ executes the rest of the traffic and the normal scheduler logic applies to other non latency queues giving them their guaranteed bandwidth.

The main advantage of LLQ over the classic PQ is that the LLQ strict priority queue is policed which minimize the chances of starvation of other queues.


In this paper we simulated our proposed scheduling algorithm and CAC algorithm in UMTS cellular networks for different transmission scenarios. We have used NS-2 as our simulation tool which is a discrete event-driven tool

[7] and independent developed module for user defined networks. The proposed CAC algorithm is investigated based on three parameters: probability of blocking of calls requesting for new admission probability of termination of calls before completion forcibly and overall system carried traffic. Overall Grade of Service to calculate performance of the network is defined as [19]: GoSj = aPhj + Pnj (1)

Where Ph is probability of the blocking of handoff calls Pn

is probability of the blocking of new calls and priority level

of handoff over new calls are shown in terms of a. The system gives good performance when GoS is small. The system capacity is investigated by evaluating total carried traffic which is defined as [20]:

CAC Simulation Results

In our proposed algorithm higher priority is assigned to handoff calls over newly arrived calls. Following three scenarios are investigated in our simulations: (1) all calls should be equally treated when there is alike load threshold and in this scenario queuing techniques are not implemented (2) handoff calls are queued till the availability of resource or time out reach (3) over and above second case higher load threshold for handoff than newly arrived calls (proposed). This mechanism is evaluated with disparate channel holding times. In our simulation general service time is 180 sec and each service has different arriving rate. Different service times (180

120 90) are used in case of third scenario. Figure 4 shows GoS of voice calls vs. arrival rates Figure 5 shows GoS of Data calls vs. arrival rates.

It can be clearly seen from these graphs that performance is improved by using queue and soft guard channel.

In case of increase in mobility there is decrease in channel holding time overall system performance will be improved and in case of decrease in service time the call in queue

waiting for getting channel have more chance to get admitted in the system before time out reach. Figure 6 depicts the comparison of system carried traffic and total

offered traffic which clearly shows that in third case overall system capacity increases and it will be enhanced more when channel holding time decreases.

LLQ Scheduling Simulation Results

In this section overall end-to-end QoS investigated scenarios and experiments are discussed in which all the QoS mechanisms which are designed in last sections are applied. The IP data Packets are sent to edge router which is simplified as both the Ingress and Egress router to the external DiffServ domain here IP data packets are marked with DSCP according to their application type and forward to GGSN. The GGSN outlink differentiates each IP data flow according its DSCP (the GGSN does not change the DSCP assigned from the external network) and transmits them with the queuing and scheduling schemes described in the previous chapter. The SGSN receives these packets and forwards them to the RNC at which the IP packets are converted into RLC SDUs.

The radio link settings are shown in Table4. The DCH works in Acknowledge Mode and the maximum RLC layer retransmission time is unlimited. This setting is due to the EURANE [9] limitations. Hence all the erroneous RLC PDUs will be recovered and the E2E SDU loss will only be caused by IP packet dropping in the bottleneck link queue which is Early Drop for real-time traffic or queue overflow for all other traffic. On the other hand the unlimited RLC PDU retransmission will result in much more uncertainty for the E2E delay and make the delay control more difficult.

Table 4: Network Parameters

###Wired Part###Link Bandwidth Link Delay (ms)


Server-Edge router###10###20

Edge router-GGSN###10###2



RNC-Node B###10###5


Simulated cell###1


Active UE number###20

DCH bandwidth###384Kbps

Fast Power Control###Ideal

Radio link RLC PDU###Uniform mean=0.01


Mobility model###No

Theses parameters are investigated using different simulation experiments in which different scheduling algorithms WRR PQ and LLQ are implemented and evaluated for different scenarios of link congestion. In start link is overloaded with 10% traffic and gradually increases to 150 %. Traffic parameters are given in Table5. Conversational class produces 20 % of overall generated traffic streaming class produce 70 % interactive class 3% and background class produce 7 % of overall generated traffic. In case of PQ scheduler voice has given highest priority while background class has the least priority. In case. In other schedulers weight represent output port bandwidth percentage i-e Weight 20 weight 70 weight 3 weight 7 for conversational streaming interactive and background respectively.

Table 5: Average throughput of NRT Traffic




layer Protocol

###Size of IP###120###160###240###480


Traffic Source###Exp###Exp###Pareto###Pareto


Holding time###Exp###Exp###Log-###Pareto





It is clear from the Figure 8 that if PQ and LLQ are implemented average end-to-end delay for conversational class remains below 100ms even in case of high congestion. The average end-to-end delay of video streaming is shown in Figure 9. All schedulers have almost same performance and provide satisfactory level of QoS. Figure 10 depicts average jitter and in all experiments for conversational class jitter stays within 50ms and same results are observed in case of video streaming as depicted in Figure 11. Figure 12 depicts packet loss rate for conversational class it is clear that link congestion has great effect on WRR scheduler performance while packet loss rate stay within boundary conditions in case of PQ and LLQ. Similarly packet loss in streaming class stays below

2x10-3 for PQ and LLQ but exceed in case of WRR.

Average throughput for interactive and background class is given in Table 6. In case of PQ in high link congestion both classes have least throughput. Whereas LLQ provides fair level of bandwidth and gives the best results for interactive and background classes even in case of high link congestion.

Table 6: Average throughput of NRT Traffic

###Traffic Class###Throughput (Kbps)


Interactive Class###8.57###4.31###8.24

Background Class###16.21###10.04###17.23


This paper focuses on investigating provisioning of QoS for real-time UMTS traffic as well as to improve the system performance. The enhanced algorithm combines Diffserv-UMTS QoS mapping WCDMA based prioritized uplink call admission control multiple queuing and LLQ scheduling with optimized parameters. The performance of this CAC with multiple queuing and LLQ scheduling is investigated with various scenarios. On the basis of the better bottleneck link bandwidth utilization while keeping the end-to-end delay and IP data Packet Loss shows adequate performance with respect to end-to-end delay packet loss ratio and achieves good bottleneck link utilization with in all QoS limitations and at the same time gives the fair service even in case of highly congested links to all traffic classes and enhance the overall system capacity.

Future research work will focus on enhancing the network capacity with convolution code with QoS for real time application users by selecting suitable value of BER.


[1] TS 23.107 Release 11 End-to-end Quality of Service concept and architecture 3GPP Nov 2012.

[2] RFC 2475 An Architecture for Differentiated

services" IETF 1998.

[3] RFC 1633 Integrated Services in the Internet

Architecture: an Overview" IETF 1994

[4] R.B. Ali S. Pierre and Y. Lemieux UMTS-to-IP QoS mapping for voice and video telephony services" IEEE Network vol. 19 no. 2 PP. 26-32 2005

[5] Lei Xiao Tung Chong Wong and Yong Huat Chew End -To-End QoS of Multi-Class traffic in WCDMA and DiffServ Network" PIMRC 2004.

[6] B. Dekeris T. Adomkus and A. Budnikas Analysis of QoS assurance using weighted fair queuing (WFQ) scheduling discipline with low latency queue (LLQ)"

28th Int. Conference. Information Technology

Interfaces ITI Cavtat Croatia June 2006.

[7] SEACORN Project

[8] Network Simulator April 2012

[9] EURANE website SEACORN project 2004.

[10] TS 23.107 Release 7 Quality of Service concept and architecture" 3GPP September 2007.

[11] V. Jacobson et al. An Expedited Forwarding PHB"

Technical Report RFC 2598 IETF June 1999.

[12] J. Heinanen et al. Assured Forwarding PHB Group"

Technical Report RFC 2597 IETF June 1999

[13] V.S Kolate G.I. Patil A.S. Bhide Call Admission Control Schemes and Handoff Prioritization in 3G Wireless Mobile Network" International Journal of Engineering and Innovative technology (IJEIT) Volume 1 Issue 3 PP.92-97 March 2012.

[14] Mohamed Hossam Ahmed Calll Admission Control in Wireless Networks: A Comprehensive Survey" IEEE Communications Surveys Volume 7 No. 1 PP. 50-69 First Quarter 2005.

[15] R. Shankar P. Dhananjayan An adaptive call admission control scheme with load balancing for QoS enhancement in RoF based converged UMTS/WLAN" International Journal Of Computers And Communications Issue 4 Volume 5 PP. 271-

278 2011.

[16] N. Mohan T. Ravichandran An Efficient

Multiclass Call Admission Control and Adaptive

Scheduling for WCDMA Wireless Network" European Journal of Scientific Research Vol.33 No.4 PP.718-727 2009.

[17] Salman AlQahtani Ashraf S. Mhmoud King A

QoS-Aware Call Admission Control Algorithm for

3G Cellular Wireless Networks"GCC conference

IEEE March 2006.

[18] Floyd S. and V. Jacobson. Random Early Detection Gateways for Congestion Avoidance" IEEE/ACM Transactions on Networking August 1993.

[19] Sampath et al. Erlang Capacity of a Power Controlled Integrated Voice and Data CDMA System" IEEE VTS Proceedings of Vehicular Technology Conference PP. 1557-1561 May 1997.

[20] Jamie S. Evans and David EverintOn the Tele traffic Capacity of CDMA Cellular Networks" IEEE Transactions on Vehicular Technology vol. 48 no. 1 Jan 19991.

[21] Majid Ghaderi Raouf Boutaba Call Admission Control in Mobile Cellular Networks A Comprehensive Survey" PP. 1-45.
COPYRIGHT 2014 Asianet-Pakistan
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2014 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Publication:Science International
Article Type:Report
Date:Jun 30, 2014

Terms of use | Privacy policy | Copyright © 2021 Farlex, Inc. | Feedback | For webmasters |