call admission control for multimedia direct sequence code division multiple access (ds-cdma)...

8/7/2019 Call Admission Control for Multimedia Direct Sequence Code Division Multiple Access (DS-CDMA) Wireless Networks

1/70

Chapter 1 Introduction

Call Admission Control for Multimedia Direct

Sequence Code Division Multiple Access (DS-CDMA)

Wireless Networks

CHAPTER 1

INTRODUCTION

Engineeringpapers.blogspot.com

1


2/70


With rapid development and continuous expansion of mobile communications,

and explosive growth in demand for new wireless cellular services, it is expected that the

next generation wireless cellular networks will support a wide variety of services,

including voice, video, images, data, or combinations of these. Direct Sequence Code

Division Multiple Access (DS-CDMA) has emerged as the predominant radio access

technology to provide high speed multimedia services used in both wideband CDMA

(WCDMA) and cdma2000.

Great capacity gain and flexibility can be achieved with deployment of a DS-

CDMA radio interface. In narrowband time/frequency division multiple acce

(TDMA/FDMA) systems with fixed channel allocation, the capacity for each cell is time

invariant based on the specified frequency reuse pattern, ensuring the co-channel

interference level. In DS-CDMA systems, all cells share the overall frequency band, and

cell capacity is interference limited. The capacity in a DS-CDMA system depends

heavily on instantaneous traffic conditions in both home and neighboring cells, while the

transmission quality over the wireless channels can be measured in terms of the Signal to

Interference ratio (SIR) at the receiver side by taking into account both the incell and

outcell interferences.

Because of this unique interference limited soft capacity nature of the DS-CDMA

system, mechanisms such as adaptive antenna arrays and voice activity factors could be

applied to improve system capacity. Despite considerable effort and progress made in

DS-CDMA system design, the accommodation of multimedia services still poses a

number of challenges. To fully utilize the scarce resources and at the same time provide

necessary quality of service (QoS) guarantees for a variety of services, it is of great

importance to design effective resource management strategies. There are several issues

of resource management in mobile communications .


2


3/70


Congestion

Cell Planning

Power and rate allocation

Call Admission Control Strategies

1.1 Problem Description

Call Admission Control (CAC) is a key element of radio resource management in

wireless and mobile communication systems. The problem of CAC is concerned with the

rules for admitting a call into the system such that the quality of service (QoS) of the call

is achieved without degrading the existing connections, fairness of resource allocation

and efficient resource usage. The problem of CAC is very challenging in multimedia

wireless systems due to the different QoS requirements, traffic asymmetry between the

uplink and downlink and differential treatment between handover and new traffic for a

given traffic class.

The objectives of this project are to:

Propose a CAC scheme for multimedia DS-CDMA wireless networks making use

of the system assumptions.

Analyze the performance of the proposed CAC algorithm using mathematical

and/or simulation techniques , using memoryless system assumptions that allow

us to model the system as a multidimensional continuous time Markov chain.

1.1.1 CAC Schemes

Call admission control is one of the most important aspects of r

management, which determines whether to admit or reject a call upon its arrival. The

objective is to maximize the utilization of resources by admitting as many new calls as

possible while maintaining the fairness and QoS of ongoing services. There are many call

admission control schemes proposed , most of which can be classified into the following

three categories.


3


4/70


Number-Based CAC - assumes time invariant cell capacity and simple for

implementation and analysis.

Interference-Based CAC - considers interference limited nature of the system and

offers a better performance, suitable for FDMA and TDMA systems.

SIR-Based CAC - best characterizes transmission quality and offers QoS

guarantee with considerable difficulty in system design and measurements,

preferred for CDMA systems.

1.2 Motivation for this Project

Since the radio spectrum is a very scarce resource, resource management is one of

the most important engineering issues in wireless and mobile communications systems.

The performance of a system with a given physical resource ( for example given the

bandwidth), depends heavily on the resource management schemes.

To provide high speed multimedia services , high capacity of a system is a basic

necessity. The Direct Sequence Code Division Multiple Access (DS-CDMA) is the most

widely used for the second and third generation mobile communications systems because

of its advantage of the soft capacity and frequency planning. Here, the term call at air-

interface means not only a voice call, but also a session for any multimedia application.

The CAC for a call request is to determine whether to accept it or not. The

objective of CACis to maximize the utilization of resource (e.g., frequency spectrum in

wireless systems) as long as the required QoSs for all calls are guaranteed.

We consider two kinds of call request, new call and handoff call. In resource

sharing between the call requests, since premature termination of connected calls is

usually more undesirable than rejection of a new call request, it is widely accepted that a

system should give higher priority to handoff call requests as compared with new call

requests. The CAC scheme proposed herein guarantees the priority of handoff calls overnew calls within a service class.

Several CAC schemes have been proposed. Some proposals assumed that the cell

capacity of a system with given frequency bandwidth is time-invariant. This type of CAC

is simple and sufficient for frequency division multiple access (FDMA) or time division


4


5/70


multiple access (TDMA) systems. In practice (especially in CDMA systems), even if a

fixed frequency band is used in a cell, the capacity of the cell varies with the loading of

the home and neighboring cells mainly because the co-channel interference changes

according to the loading. Thus, some CAC proposals for CDMA systems are based on the

QoS parameter such as the signal to interference ratio (SIR).

Here we consider a CAC policy under which a call can be admitted when the SIR

requirements of both the existing calls and the new call are guaranteed. The performance

of the algorithm is analyzed using the Markovian assumptions. The performance

measures that are considered are the blocking probabilities of the handoff and new calls,

throughput of the uplink and downlink and the outage probability of a call in progress

within a cell.

1.3 Tools used

The traffic is modeled using the NS2 simulator (version 2.31) and the packet

delay and packet loss are observed using the Xgraph. The performance of the algorithm is

analyzed and the results are plotted using MATLAB (version 7.1).

1.4 Overview of the thesis

To go straight to the subject, a little introduction about CAC is presented in the

motivation of the project. Then a detailed discussion about the DS-CDMA, its features,

and the various issues in resource management is presented in chapter 2. Then a brief

introduction to UMTS networks and a detailed discussion about the various CAC

schemes is presented in chapter 3. The CAC scheme is proposed and a detailed analysis

of the performance of the algorithm is presented in chapter 4 i.e.., here we will discuss

about the proposed CAC scheme with the algorithm for the performance analysis.

Chapter 5 presents the results with a brief introduction of the simulator tool used. Chapter

6 includes the conclusion part of the project and the future scope of the work.


5


6/70


CHAPTER 2

DS-CDMA AND RESOURCE MANAGEMENT


6


7/70


2.1 DS-CDMA

With rapid development and continuous expansion of mobile communications,

and explosive growth in demand for new wireless cellular services, it is expected that

next-generation wireless cellular networks will support a wide variety of services,

including voice, video, images, data, or combinations of these. DS-CDMA(Direct

Sequence Code Division Multiple Access) has been the predominant radio access

technology to provide high-speed multimedia services used in both wideband CDMA

(WCDMA) and cdma2000.for the present generation wireless networks(3g and 4g)

because of its unique features and soft capacity nature (compared to TDMA and FDMA).

To achieve a higher system capacity one of the most efficient methods is to reduce the

multiple access interference among the users. Great capacity gain and flexibility can be

achieved with deployment of a DS-CDMA radio interface.

2.1.1 Capacity in DS-CDMA

In DS-CDMA cellular networks, all users share the same total frequency band for

transmissions. Each user is assigned one or more distinct spreading codes, and all these

codes generally bear noise-like characteristics with very small cross-correlation to each

other. The quality of communication is primarily determined by the detected SIR level atthe receiver, with the generally accepted measurement on the bit energy to-noise density

ratio ( 0bE N), expressed as

t

N

P RSIR

I W= (2.1)


7


8/70


where R is the baseband information bit rate; Wis the total radio frequency band for

transmission, so the corresponding spreading gain is G = W R ; tPis the received signal

power for the reference spreading code; and NIdenotes the total detected interference at

the receiver. In order to guarantee transmission quality, the received SIR for each service

should be maintained above a certain threshold.

To achieve higher system capacity, one of the most efficient means is to mitigate

multiple access interference (MAI) among multiple users.Two techniques that are

commonly used are:

i. Cell Sectorization, which deploys directional antenna arrays at the base station

and splits a cell into different sectors. Generally speaking, only signals from theusers within a sector are received at the corresponding antenna array. Thus, the

number of users one cell can serve could be increased by approximately the same

factor as the number of sectors.

ii. Voice Activity Monitoring, which switches off signal transmission when the

mobile terminal is not active to reduce interference.

In order to accommodate high-speed multimedia services and support variable

transmission rates, two mechanisms can be used:

a. Multi Code(MC) CDMA system.

b. Variable Spreading Factor(VSF) CDMA system.

In an MC-CDMA system, all data signals over the radio interface are transmitted

at a basic rate, and the spreading gain G over each code channel is a constant. Multiple

orthogonal spreading codes are transmitted simultaneously for a highspeed application.

In a VSF-CDMA system each user transmits over one single code channel.

Higher transmission rate can be achieved by varying the spreading factorG inversely

with the desired data rate. Thus, in VSF-CDMA systems, users are assigned variable

length codes and different power levels, based on data rates and QoS requirements. The

two mechanisms provide comparable performance in high-speed transmission.

2.1.2 SIR Model of a CDMA system


8


9/70


In general, the uplink and downlink use different bandwidth regions

transmission in a DS-CDMA cellular system. Thus, they are usually consi

separately for capacity calculation.

Referring to the figure below , in the uplink, the received signal power at the

reference base station, BS0, from a specific mobile station, jMS , is jMS s transmitted

power jPmultiplied by the pass loss factor 0,jL ; the corresponding interference is the

Figure 2.1:The SIR model for a DS-CDMA cellular system.

received power from other active mobile stations, both within reference cell BS0 (e.g.,

iMS ) and out cells (e.g., kMS in BS1), plus the background thermal noise. The capacity

in the uplink is calculated as the maximum number of users that could be accommodated,

subject to the SIR requirement for each user in the reference cell.

In the downlink, a fraction of total transmission power at the base station is

dedicated to the control channels, and all traffic channels share the remaining power. As

shown in Figure 1, reference cell BS0 allocates a fraction of its traffic channel power to

any in-cell mobile user (e.g., iMS ); the remaining power received at iMS from its home

base station BS0 and other base stations (e.g., BS1) will appear as interference. The

capacity in the downlink is defined as the maximum number of users that could be

admitted, under the constraints that the transmission quality of each user, in terms of SIR


9


10/70


threshold, is guaranteed, and the overall power needed does not exceed the maximum

power supply at the base station.

In DS-CDMA systems, all cells share the overall frequency band,and cell capacity

is interference-limited. The capacity in a DS-CDMA system depends heavily on

instantaneous traffic conditions in both home and neighboring cells, while

transmission quality over the wireless channels can be measured in terms of the signal-to-

interference ratio (SIR) at the receiver side by taking into account both the incell and out-

cell interferences. Because of this unique interference-limited soft capacity nature of the

DS-CDMA system, mechanisms such as adaptive antenna arrays and voice activity

factors could be applied to improve system capacity. Even then, the accommodation of

multimedia services poses a number of challenges.

To fully utilize the scarce resources and at the same time provide necessary

quality of service (QoS) guarantees for a variety of services, it is of great importance to

design effective resource management strategies.

2.2 Features of DS-CDMA

There are also several distinctive characteristics in a DS-CDMA system that can

be explored for better performance.

a) Universal frequency reuse

b) Power control.

c) Soft Handoff.

d) Voice Activity.

e) Propagation Model.

2.2.1 Universal Frequency Reuse

The universal frequency reuse of a DS-CDMA system allows all cells to share thesame wide frequency band. This reduces the complexity in cell planning on co-channel

bandwidth allocation and potentially leads to higher capacity gain for a cell under

asymmetric loadings and increases the dependency among neighboring cells and incurs

significant difficulty in resource management, as more frequent coordinations are


10


11/70


required. Because of universal frequency reuse with interference-limited transmission,

even if the cell occupancy information is available, outage might still occur due to user

mobility, propagation variation, voice activity, and so on. As a result, more precise

control and finer tuning are generally required in resource management.

2.2.2 Power Control

Power control in DS-CDMA systems helps reduce excessive interferen

throughout the system and prolong battery life . In general perfect power control is

assumed . In the uplink, the received power from all users in the reference cell is kept

equal to overcome the near-far problem . In the downlink, the assigned power to each

user is adjusted to achieve exactly the required SIR level. In practical systems, power

control imperfection may occur occasionally and cause some misadjustment of received

power. Consequently, the received power level of both the required signal and the

interference might change over time which in turn exerts additional difficulty on power

allocation and adjustment, and requires resource management strategies to incorporate

more accurate estimation of power control imperfections.

2.2.3 Soft Handoff

Soft handoff denotes the state where a mobile transmits to and receives from more

than one base station simultaneously. On the downlink, the mobile combines the signals

from the base stations in connection and adds the different multipaths to reinforce the

received signal, which leads to improved communication quality. On the uplink, the

neighboring base stations independently decode the signals received, and the best replica

is selected. As a result, soft handoff ensures smoother user communications, better

communication quality during handoff, and larger cell coverage in DS-CDMA systems.

The trade-off lies in higher complexity and additional network resource demands. As

longer handoff transfer delay is generally allowed, it is possible for handoff traffic to be

queued. Furthermore, as more base stations become involved in communication with one


11


12/70


mobile user, power transmission and allocation should be carefully adjusted to maximize

utilization and avoid excessive interference in the overall system.

2.2.4 Voice Activity

In DS-CDMA systems, cell capacity is mainly constrained by interference from

simultaneous transmissions. In such cases, the voice activity factor might be incorporated

into system design, and spectrum utilization can be significantly improved by shutting

down mobile stations during their silent periods. As a result, the system does not need to

provide full bandwidth to the admitted calls as long as their received SIR levels during

active periods exceed the required threshold. However, with partial bandwidth allocation,

outage might occur occasionally if most of the users are actively transmitting. The

additional capacity gain should be well justified to maintain the required QoS, and

congestion control is needed to combat the more varying system behavior.

2.2.5 Propagation Model

Generally, the propagation model might change greatly in a variety

environments or/and during different time intervals. This imposes great difficulty as well

as new challenges on system management. For example, under poor channel conditions,

greater power should be allocated to compensate for severe path loss. However, this may

cause overprovisioning of system resources and incur additional interference to others in

better channel conditions. On the other hand, such diversity gain might help improve

performance in terms of higher system throughput and lower inter and intracell

interference by precisely tracking channel fluctuations and scheduling transmissions

when the corresponding channel quality is near its peak. Therefore, a precise loss model

and accordingly better resource management strategies should be designed to properly

utilize the system resources.

2.3 Resource Management

Resource management strategies include how to efficiently allocate the available

resources to optimize channel utilization, how to adjust service rate to relieve congestion,


12


13/70


how to provide diverse QoS requirements and how to provide priority among the various

users. There are also several distinctive characteristics in a DS-CDMA system that can be

explored for better performance. For example, universal frequency usage enables more

efficient resource provisioning; proper rate and power allocation helps to reduce

interference with greater capacity gain. On the other hand, these features inevitably

increase the complexity of system design, and thus must be carefully addressed.

2.4 Issues in Resource Management

There are several issues of resource management in mobile communications.

i. Congestion

ii. Cell Planning

iii. Power and rate allocation

iv. Call Admission Control Strategies.

2.4.1 Congestion Control

Congestion occurs if the system fails to find a set of power transmission levels

that satisfy users QoS requirements. Congestion might happen due to a variety of factors

such as the deterioration of the wireless environment, mobility of users, users activity,

and power control imperfection even with perfect admission control. As a result,

transmission power and cell interference tend to increase and outage occurs. This incurs

traffic loss and delay jitter, thus deteriorating transmission quality. Generally, when

congestion occurs three main mechanisms can be applied to relieve it:

Drop some ongoing calls. The system may choose to drop call(s) in outage condition, or

drop each existing call with prespecified probability, or drop call(s) that make the largestcontribution to alleviating congestion. The last scheme offers the best performance at the

greatest complexity.

Decrease transmission rate. This could be done by either proportionally reducing the

transmission rate of each user or decreasing the rates to the same maximal fair SIR level.


13


14/70


Reduce the number of simultaneous transmissions. The transmission probability can be

dynamically adjusted according to occupancy information. To eliminate randomness and

obtain full control over simultaneous transmissions, cells may sequentially schedule

active transmitting periods for each service type.

To conclude, dropping ongoing calls is the most straightforward way to relieve

congestion, but it is undesirable and often unbearable from the users point of view. To

simultaneously reduce the data rate of all data services yields lower throughput than to

support fewer users with full rates at the same time. Thus, decreasing data rate results in

suboptimal system utilization. On the other hand, switching off some users leads to extra

transmission jitter and delay.

2.4.2 Power and rate allocation

A variety of criteria can be optimized, such as to minimize the

consumption, which prolongs battery life and causes less interference; or to maximize

transmission rate, which indicates maximized system throughput and resource utilization.

For a downlink, the two constraints are the SIR constraint for each service type

and the average transmission power limit among all base stations. The basic problem is

that if a mobile experiences poor channel conditions, the assigned transmission rate for

this mobile should be low.To cope with these two disadvantages, suboptimal algorithms

are proposed in which rate allocation depends on interference distribution, and powerallocation depends on traffic distribution among the cells.

For an uplink, the optimization is formulated to maximize the total normalized

transmission rate subject to the constraint on transmission power and max

allowable rate. The optimal solution yields not only higher throughput but also significant

power savings. However, this can potentially lead to starvation for users with poor

channel conditions, since users with better channel conditions will always transmit first.

2.4.3 Cell Planning

Efficient cell planning is of vital importance for service providers to reduce

network cost and maximize utilization of scarce resources. Generally, cell planning

consists of a number of issues, such as

i. Bandwidth Allocation.


14


15/70


ii. Base Station Planning.

iii. Pilot Power Control, and

iv. Cell Sectorization.

Bandwidth allocation in uplink and downlink

In most systems, uplink and downlink use different bandwidth regions for

transmission, and the bandwidth allocation between them is symmetric. However, to

accommodate multimedia services, greater amounts of radio resources are required on

the downlink than on the uplink. To cope with the traffic asymmetry, unbalanced

bandwidth allocation is preferred, and great system performance gain might be achieved

by the proper assignment of bandwidth between two links according to traffic demand.

Base station planning

Since the overall frequency band is shared by all active users and the capacity of

each cell depends on the interference level, proper planning of base station locations

should consider not only cell coverage but also some other factors, such as the amount of

resources available, estimated traffic distribution in the area, and radio ch

propagation models. The main purpose of base station planning is to select the sites for

base stations by taking into account the system cost, transmission quality, servicecoverage, and so on. The optimal location of new base stations to minimize a linear

combination of installation cost and total transmitted power can be done by taking into

account traffic distribution, SIR requirements, power allocation constraints, and power

control mechanisms. More complex models also exist with considerations of the

stochastic behavior of the system or/and soft handoff.

Pilot power control

To select the proper base station(s) to connect, mobile stations need to measure

and report the 0bE N level of the received pilot power to the base stations. The pilot

power determines the cell coverage area and average number of users within a cell.

Increasing the pilot signal power of a base station expands the coverage area of the cell,


15


16/70


thereby increasing the number of users in the serving cell, but resulting in higher intracell

interference. On the other hand, it may decrease the number of users in adjacent cells

with decreased intercell interference. Efficient pilot power control needs to balance the

cell load and cell coverage area among neighboring cells, with the objectives of reducing

the variation of interference, stabilizing network operation, and improving cell capacity

and communication quality, especially under nonuniform traffic loading among cells.

Cell Sectorization

Traditional sectoring approaches divide the cell into equal width sectors, which

has been shown to provide the same capacity gain under highly uniform traffic load.

However, for a system with hot spot traffic, those sectors under high density traffic load

may suffer high outage probability. Adaptive cell sectoring can be deployed to greatly

improve the performance in such a system.The optimal cell sectoring (OS) to minimize

the total transmission power can be formulated as a shortest path problem and solved by

Dijkstras algorithm. The direction and width of the sectors can also be adjusted

according to the geographic distribution of traffic. A cluster-based sectoring (CS)

algorithm can be proposed as we observe that the sector boundaries had better be across

some low density regions in order to avoid excessive oscillations,


16


17/70


Figure:2.2 User distribution and cell sectoring.

also with the objective to achieve lower computational complexity. The complexity of CS

is much lower than OS, without any significant degradation of system performance.

Specifically, under the user distribution and sectoring for OS and CS shown in

Figure:2.2, the CS algorithm greatly reduces complexity from OS with only a slight

increase in total transmitted power. Furthermore, the sector boundaries generally cross

low density regions in the CS solution; while in OS, they may pass through two users

very close to each other. It is expected that dynamic cell sectoring could be designed to

incorporate the stochastic nature of the system, such as user mobility, channel fading,

power control, and sectorization imperfection, for better adaptation to real systems.

CHAPTER 3


17


18/70


UMTS NETWORKS AND CAC STRATEGIES

3.1 Introduction to UMTS networks

Since their inception, mobile communications have become sophisticated and

ubiquitous. However, as the popularity of mobile communications surged in the 1990s,

Second Generation (2G) mobile cellular systems such as IS-95 and Global System for

Mobile (GSM) were unable to meet the growing demand for more network capacity. At

the same time, users demanded better and faster data communications, which 2G

technologies could not support.

Third Generation (3G) mobile systems have evolved and new services have been

defined: mobile Internet browsing, e-mail, high-speed data transfer, video telephony,

multimedia, video-on-demand, and audio-streaming. These data services had differentQuality of Service (QoS) requirements and traffic characteristics in terms of burstiness

and required bandwidth. Existing cellular technology urgently needed a redesign to

maximize the spectrum efficiency for the mixed traffic of both voice and data services.

Another challenge was to provide global roaming and interoperability of different mobile


18


19/70


communications across diverse mobile environments.

Toward these ends, the International Telecommunication Union (ITU), the

European Telecommunications Standards Institute (ETSI), and other standardization

organizations collaborated on the development of the Future Public Land Mobile

Telecommunication Systems (FPLMTS). The project was later renamed International

Mobile Telecommunications-2000 (IMT-2000). The new 3G mobile cellula

communication system was set to operate at a 2 GHz carrier frequency band. For the PS

domain, the supported data rates were specified for the various mobile environments:

Indoor or stationary 2 Mbps

Urban outdoor and pedestrian 384 kbps

Wide area vehicular 144 kbps

Of the various original proposals, the two that gained significant traction were based

on Code Division Multiple Access (CDMA): CDMA2000 1X and Universal Mobile

Telecommunication System (UMTS).

i. CDMA2000 1X was built as an extension to cdmaOne (IS-95),

enhancements to achieve high data speed and support various 3G services.

ii. UMTS was based on the existing GSM communication core network (CN) but

opted for a totally new radio access technology in the form of a wideband version

of CDMA (Wideband CDMA: WCDMA). The Wideband Code Division Multiple

Access (WCDMA) proposal offered two different modes of operation: Frequency

Division Duplex (FDD), where Uplink (UL) and Downlink (DL) traffic are

carried by different radio channels; and Time Division Duplex (TDD), where the

same radio channel is used for UL and DL traffic but at different times.

3.2 UMTS Network Topology

When deploying a WCDMA network, most operators already have an existing 2G

network. WCDMA was intended as a technology to evolve GSM network toward 3G

services.


19


20/70


3.2.1 GSM Network Architecture

Figure 3.1 illustrates a GSM reference network , showing both the nodes and the

interfaces to support operation in the CS and PS domains. In this reference network, three

sub-networks can be defined

Figure: 3.1 GSM Reference Network

Base Station Sub-System (BSS) or GSM/Edge Radio Access Network (GERAN)

This sub-system is mainly composed of the Base Transceiver Station (BTS) and

Base Station Controller (BSC), which together control the GSM radio interface , eitherfrom an individual link point of view for the BTS, or overall links, including the transfers

between links for the BSC. When data functionality was added to GSM with the

deployment of General Packet Radio Service (GPRS), an additional node was added to

the interface between the GPRS-CN and the radio interface, that is the Packet Control

Unit (PCU).

Network and Switching Sub-System (NSS)

This sub-system mainly consists of the Mobile Switching Center (MSC) that

routes calls to and from the mobile. For management purposes, additional nodes are

added to the MSC, either internally or externally. Their main purpose is to keep track of

the subscription data, along with associated rights and privileges, in the Home Location

Register (HLR), or to keep track of the subscribers mobility in the HLR and Visitor


20


21/70


Location Register (VLR). Two other nodes manage security issues: the Equipment

Identity Register (EIR) verifies the status of the mobile phone (i.e., the hardware), while

the Authentication Center (AuC) manages the security associated with the Subscriber

Identity Module (SIM). The last node is the Gateway-MSC (GMSC). The MSC and

GMSC are differentiated only by the presence of interfaces to other networks, the Public

Switched Telephone Network (PSTN) in the GMSC case. Typically, the MSC and the

GMSC are integrated.

General Packet Radio Service, Core Network (GPRS-CN)

Within the NSS, two specific nodes are introduced for the GPRS operation: the

Serving GPRS Support Node (SGSN) and the Gateway GPRS Support Node (GGSN). In

the PS domain, the SGSN is comparable to the MSC used in the CS domain. Similarly, in

the PS domain, the GGSN is comparable to the GMSC used in the CS domain. These

nodes rely on existing BSS or NSS nodes, particularly the VLR and HLR, to manage

mobility and subscriptions .

3.2.2 UMTS Overlay Release 99

UMTS is based on the GSM reference network and thus shares most nodes of the

NSS and General Packet Radio Service, Core Network (GPRS-CN) sub-systems. The

BSS or GERAN is maintained in the UMTS reference network as a complement to the

new Universal Terrestrial Radio Access Network (UTRAN), which is composed of

multiple Radio Network Systems (RNS) as illustrated in Figure 3.2. Compared to the

GSM reference network, the only difference is the introduction of the Radio Network

Controller (RNC) and Node Bs within the newly formed RNS. Essentially, these two

nodes perform tasks equivalent to the BSC and BTS, respectively, in the GSM

architecture.


21


22/70


Figure: 3.2UMTS Reference Network

3.2.3 UMTS Network Architecture beyond Release 99

The initial deployments of WCDMA networks comply with Release 99 of the

standard. This standard, or family of standards, began to evolve even before being fully

implemented, to address the limitations of the initial specifications as well as to include

technical advancements. At a higher level, migrating from Release 99 to Releases 4, 5,

and then 6 does not change the structure of the network.

3.3 Call Admission Control (CAC)

Call admission control is one of the most important aspects of r

management, which determines whether to admit or reject a call upon its arrival. The


22


23/70


objective is to maximize the utilization of resources by admitting as many new calls as

possible while maintaining the fairness and QoS of ongoing services. There are many call

admission control schemes proposed , most of which can be classified into the following

three categories.

a. Number-Based CAC.

b. SIR-Based CAC or

c. Interference-Based CAC.

3.3.1 Number-Based CAC

In number-based CAC, the QoS requirement for the upper bound on packet error

probability is mapped into the maximum number of voice/data calls that can be

simultaneously accommodated in the system, denoted v dK K. To reflect voice calls

capability to tolerate higher bit error rates, vKis set greater than dK. As shown in

Figure:3.3, the admission of voice calls depends on a threshold value v . Newly arriving

data either joins the backlog pool or is discarded if the pool is full.

To provide priority to voice calls, the admission threshold v can be set equal to

vK(i.e., admit as many voice calls as the system can support). Some fairness can be

offered to data traffic by setting v less than vK, taking into account the number of

backlogged data packets. At any time slot t, voice calls can transmit without delay and

the backlogged data packets occupy the silence period of voice calls for transmission

with probability tP . tP is calculated according to a function with the parameters for

maximum allowed simultaneous data transmissions dK, current active voice calls, and

number of backlogged data packets obtained from system feedback at the beginning of

each time slot.Based on the time-invariant cell capacity assumption, the operation of number-

based call admission strategies is very similar to those in narrowband systems, in which


23


24/70


call admission only depends on current cell loading. Such a scheme is simple for

Figure: 3.3 Number-based admission control.


24


25/70


implementation and analysis. However, in a DS-CDMA system, the capacity varies with

the interference level, and the number-based schemes completely ignore the soft nature of

DSCDMA capacity, so the control is generally inaccurate and nonadaptive.

3.3.2 Interference-Based CAC

In DS-CDMA cellular networks, all users share the same wide frequency band.

When a call ( MS_Arrival ) is admitted into a cell, it transmits to and receives information

from the corresponding base station ( BS0), as shown in Figure: 3.4. In the uplink the

transmitted power from MS_Arrival increases interference levels on other in-cell (

MS_In ) and outcell ( MS_Out ) users receivers at their base stations BS0 and BS1. In

the downlink the power allocated fromBS0 to MS_Arrival causes additional

interference for all other users (MS_In and MS_Out ) in the system.

The interference level depends heavily on overall system conditions. As a result,

the admission of a new call can gracefully degrade the performance of all users currently

in service. Accordingly, in a DS-CDMA system the number of calls that can be admitted

to a cell is not a fixed value. Thus, a more reasonable measurement should be the

interference level at the receivers.

An admission control scheme based on the received interference at base stations

was proposed for the uplink DS-CDMA system. Three interference margins are defined:

the total interference margin ( TIM ), current interference margin ( CIM ), handoff

interference margin ( HIM ).

Here, TIM is the maximum acceptable link interference such that the QoS in

terms of lower bound b 0E N is guaranteed; CIM is the estimated interference level

taking into account the assignment of the channel to the newly arrived call; HIM is the

interference estimation further considering the reserved channels for handoff calls. The

BS interprets the current interference from the measured power strength and calculates

CIM and HIM accordingly. If there is a call request and HIM < TIM (i.e., handoff

arrival can be safely reserved), the call will be admitted. Otherwise, the base station will


25


26/70


check whether it is a handoff call and CIM < TIM . If so, the base station will assign a

Figure: 3.4 MS Admission


26


27/70


new channel for the handoff call; otherwise, the call will be rejected. Compared with

number-based schemes, interference based call admission control better represents the

inherent interference limited nature of DS-CDMA systems at the expense of extra

complexity in interference detection.

3.3.3 SIR-Based CAC

Admission control based on the uplink SIR requirements was proposed . A crucial

measurement, residual capacity, is defined as the additional number of calls that can be

accepted by a base station so that the systemwide outage probability will not exceed a

predefined SIR level. In the localized algorithm, the residual capacity is calculated solely

based on the SIR measurement at the local base station, and a call will not be admitted

unless the residual capacity is greater than zero. In the global algorithm, all adjacent

cells SIR levels and residual capacities will be calculated at call admission. A call is

accepted if and only if the minimum of all cells residual capacity is greater than zero.

This ensures that the admission of the new call will not affect the QoS in all surrounding

cells, which is particularly important in a nonuniform traffic scenario. This can be

extended to support multimedia services, where uplink and downlink were considered

separately. The average SIR level for each call class is measured periodically for both

links. Accordingly, for each class i, the system estimates the expected i,jSIRat a class j

call arrival. Distinct admission thresholds i,j are set for admission if i,jSIR i,j , for all

call classes. Priority is provided for handoff calls by setting higher SIR thresholds for

new arrivals.

To better reflect the stochastic system behavior and guarantee long-t

transmission quality, an admission control scheme for the downlink MC-CDMA system

supporting multiple classes of traffic was proposed. The lower-bound SIR requirement

for each service type is formulated, incorporating statistical factors such as voice activity

and log-normal shadowing in propagation. The long-term outage probability

calculated, and the capacity constraint is derived for admission control. Reservation

based on an iteratively estimated handoff arrival rate is performed to provide priority

support for handoff traffic.


27


28/70


In summary, number-based call admission control assumes a time-invariant cell

capacity and is simple for implementation and analysis. However, it does not consider the

inherent soft capacity nature of DS-CDMA system and could lead to inaccurate results.

On the other hand, the SIR-based schemes best characterize the transmission

quality and thus offer QoS guarantee. But the varying system environment poses

considerable difficulty in system design and measurements.

Interference-based call admission control also considers the interference-limited

nature of the system. The trade-off lies in simpler implementation for

measurements of QoS.

The CAC scheme that is proposed here is based on the measured Signal to

Interference Ratio (SIR).

3.4 System Model

CDMA systems are interference-limited systems. Thus, the capacity of a cell that

is, the total resource in the cell) varies with the loading of the home and neighboring cells

because the co-channel interference changes according to the loading. On the other hand,

for guaranteeing adequate call quality, the SIR of a call should be maintained to be higher

than a predefined value. To accomplish this objective, a call request is admitted only

when even though it is accepted, the SIR of an ongoing call is expected to be not smaller

than a threshold value. This type of CAC scheme is called the SIR-based scheme. The

proposed scheme here decides whether to admit or reject a call request based on the

measured bit-energy-to-noise density ratio 0bE Nat receiver.

3.4.1 Model of Multiple Class Calls

Assume that there are L classes of calls in the system. While a call is connected,

it alternates between active and dormant states according to the characteristics of traffic

source. Let us assume that only the active calls generate traffic. The data rate of a call

varies according to its state with this model. Let us denote the uplink and downlink data

rates of a class i call in active withu

iR andd

iR ( 0 1i L ) respectively.

Call requests are classified into handoff call and new call requests. The proposed

scheme gives higher priority to handoff calls over new calls within the same class. If


28


29/70


i j< the admission priority of a call with class i is higher than that of a class j call

which means a handoff (new) call of class i has a higher priority over a handoff (new)

call of class j . A new call request of high-priority class may have higher priority as

compared with a handoff call of low priority class.

3.5 Proposed Call Admission Control Scheme

In practical systems, a base is equipped with a finite number of receiver elements

and there can be only a finite number of spreading codes for users in a cell. When all

receiver elements have been already assigned to ongoing calls, any call connection

request is rejected. In this system, the number of calls accommodated simultaneously in a

cell is limited to the number of OVSF codes.

When a class- i call request (either a new call or a handoff call) arrives at a base,

the base decides whether to admit or reject the request. If all receiver elements have been

already assigned to ongoing calls or there is no available spreading code, the call request

is rejected. If both receiver elements and spreading codes are available, the following call

admission scheme is used.

The class- i call request can obtain an admission only when even though the

request is accepted, a bit energy-to-noise density ratio of an ongoing active call within a

home cell is expected to be not lower than a threshold.

The CAC procedure consists of three stages.

3.5.1 Stage 1

Let us first consider uplink. The base measures the uplink 0bE Nfor each active

call and calculates the average of the measured 0bE Ns for each class periodically (e.g.,

every power control step). Letu

kMbe the average of the measured uplink 0bE Ns for

class- kactive calls. Then the base estimates how much the mean uplink 0bE Nof class

kcalls decreases, due to the acceptance of a class- i call request.

Let ,u

k iEdenote the estimate of the resulting 0bE Nfor class k,when the class- i

call is accepted which is given by equation (3.1).


29


30/70


1

,

1{ . }

u uu i ik i u u

k u k

RE

M W

= + (3.1)

where uWis the uplink spreading bandwidth andu

kdenotes the required value of

uplink 0bE Nfor class- kcalls, for maintaining adequate transmission quality.

3.5.2 Stage 2

Consider downlink. Assume that each active mobile measures its downlink

0bE Nand can report it to the base periodically. Then the base calculates the average of

the reported 0bE Ns using the reported information, separately for each class which is

denoted byd

kM.Then the base determines ,

d

k iE, denoting the estimate of the resulting

0bE Nfor class kwhen the class- i call is accepted, given by equation (3.2)

1

,

1{ (1 ) . }

d dd i ik i d d

k d k

RE

M W

= + (3.2)

where dW is the downlink spreading bandwidth and denotes the average

orthogonality factor in downlink.

3.5.3 Stage 3

Let ,uk iand ,dk i , respectively, denote the uplink and downlink threshold 0bE Ns

of a class- kcall for controlling admission of a class- i new call request. Let ,u

k i and

,

d

k i ,respectively denote the uplink and downlink threshold 0bE Ns of a class- kcall for

controlling admission of a class- i handoff call request.

The proposed CAC scheme accepts a new call request with class- i if the

following condition (i.e.., equation (3.3)) for any class- kis satisfied.

, ,u uk i k iE and , ,d dk i k iE for0 k L (3.3)

If a class- i handoff call arrives , then the base decides admit the call if the

following condition (i.e.., equation (3.4)) for any class- kis satisfied.

, ,u u

k i k iE and , ,d d

k i k iE for0 k L (3.4)


30


31/70


3.6 Threshold values for CAC

Conditions for supporting CAC priorities among call classes are as follows:

Between Handoff and New Calls Within a ClassThe possibility that a connection request is admitted increases as the threshold

0bE Nbecomes lower. In the proposed CAC scheme, handoff requests have higher

priority over new call requests of the same class. This is supported by the following

inequalities (i.e.., equation (3.5))for any class- i ( 0 1i L ).

, , , ,,

u u d d

k i k i k i k i < < for0 k L (3.5)

Between Calls With Different Class

Let us consider new calls. From our assumption that if , i j< the priority of a

new (handoff) call with class- i is higher than that of a new (handoff) call with class- j ,

the higher priority of a class- i new call as compared with a class- j new call means that

the condition (3) is more easily satisfied for class- i than for class- j .For example, let us

examine uplink condition in (3). As shown in (1), the estimated 0bE N, ,u

k iE, is

dependent on the data rateu

iR and the required 0bE N,

u

i . Ifu u u u

i i j jR R < , then

, ,

u u

k i k jE E> . In this case, even though ,u

k i is equal to ,u

k j, uplink condition of (3) is

satisfied with higher probability for a class- i new call than for one with class- j . If

u u u u

i i j jR R = , then ,u

k i should be smaller than ,u

k j . In summary, CAC threshold values

should be determined so as to guarantee the priority between call classes, considering the

data rates and the required 0bE Ns of all call classes.

Reduction in the Number of CAC Threshold Values


31


32/70


As seen in (3) and (4), there are 4 thresholds for each class and 24L in total.

Therefore, the work to determine CAC threshold values may be very hard. To reduce the

complexity, we set the CAC threshold values using the required 0bE Ns as follows,

given by equation (3.6).

, , , ,, , ,u u h u u n d d h d d n

k i k i k i k i k i k i k i k i = = = =

for0 1i L and 0 k L (3.6)

wheren

iandh

iare, respectively, the CAC parameters for the new call requests and

handoff call requests with class- i , and are larger than one. As seen in (6), the number of

CAC parameters, s, is reduced to 2 in total and this makes us determine CAC

thresholds more easily. The priority of handoff calls over new calls within any class- i is

guaranteed byh n

i i < .

CHAPTER 4


32


33/70


PERFORMANCE ANALYSIS AND

IMPLEMENTATION

4.0 Introduction

A cellular system consists of several cells. We assume that the overall system is

homogeneous in statistical equilibrium. For a homogeneous system , a cell is statistically

the same as any other cell. Thus, for each class, the mean handoff arrival rate to a cell

should be equal to the mean handoff departure rate from the cell. With this observation,

we can decouple a cell from the rest of the system and evaluate the system performance

by analyzing the performance of that cell.

4.1 Assumptions

The following assumptions regarding memoryless properties allow us to model

the system as multidimensional continuous time Markov chain.


33


34/70


A1) New calls of class- i , 0 1i L , arrive at the cell according to Poisson process

with rate i .

A2) The dwell time of a class- i call in a cell is exponentially distributed with mean 1 i.

A3) The service time of a class- i call is also exponentially distributed with mean 1 i.

A4) During a connection, a class- i call alternates between active state and dormant state,

and the data activities of uplink and downlink channels of a call are independent of each

other. The amount of time which the class- i call spends in each of active and dormant

states for an uplink (downlink) channel are exponentially distributed with mean 1u

i(

1 di) and 1u

i(1d

i), respectively.

According to the above assumptions A1)A3), handoff calls of class- i arrive

from adjacent cells according to Poisson process. We denote this arrival rate of class- i

handoff calls by i. From the assumption A4), we can obtain the uplink and downlink

activity factorsof a class- i call, denoted byu

iandd

i, respectively. That is,

uu ii u u

i i

=

+and

dd ii d d

i i

=

+.

4.2 Flow Balance Equations

When in ( 0 1i L ) denotes the number of class- i calls in progress within the

cell, the system state can be defined by a row vectors

s 0 1 1( , ,......, )Ldef n n n .

When a call request arrives at a base, the base performs two tests. It first checks

whether there are available receiver elements and available spreading codes for the

request. Let us define a feasible state as a possible state for the given number of

receiver elements and the spreading code assignment scheme.We denote the state space

of all feasible states by S. Then, the first test implies that the base checks whether the


34


35/70


state will be within S, after admitting the call request. If this simple test is passed, the

base secondly decides whether to admit or reject the request, using the proposed CAC

scheme. The CAC scheme influences the state transition rate

between feasible states .

Within the feasible state space S, any state transition is caused by one of the

following events:

1. Arrival of a new call.

2. Arrival of a handoff call.

3. Termination of an ongoing call and

4. Handoff of an ongoing call toward neighboring cells.

Let us denote the possible successor states from state s as

is + 0 1 1 1 1( , ,...., , 1, , ...., )i i i Ldef n n n n n n + +

is 0 1 1 1 1( , ,...., , 1, ,...., )i i i Ldef n n n n n n +

.

Consider the transition from state s to stateis + , caused by the origination of a

class- i new call. Let us denote this transition rate by ( , )nq s i . When ( , )

nA s i is the

probability which a base admits a class- i new call in state s , is expressed as

( , ) ( , )n n

iq s i A s i= .

Consider the transition from state s to stateis + , caused by the origination of a

class- i handoff call. Let us denote this transition rate by ( , )hq s i . When ( , )

hA s i is the

probability which a base admits a class- i handoff call in states , is expressed as

( , ) ( , )h h

iq s i A s i = .

Let us consider the state transition from state s to stateis , due to the

completion of a class- i call in process. This state transition rate, denoted by ( , )cq s i , is

expressed as


35


36/70


( , )c

i iq s i n = .

Let us consider the state transition from state s to stateis , due to the

departure of a class ongoing call toward adjacent cells. This state transition rate, denoted

by ( , )xq s i , is expressed as

( , )x i iq s i n= (4.5)

Let ( )s be the stationary probability of state s . The stationary probabilities

should then satisfy the following flow balance equations:

1

0

1

1

1

( ) { ( , ) ( , ) ( , ) ( , )}

. ( ){ ( ) ( , )}

. ( ){ ( ) ( , )}

i

i

Ln h c x

i

L n h

n i i i

i o

Lc x

s S i i i

i o

s q s i q s i q s i q s i

I s q s q s i

I s q s q s i

+

=

=

+ + +

=

+ + +

= +

+ +

for all s S (4.6)

with the additional normalization equation that

( ) 1s S

s

= (4.7)

In (4.7), cIis an indicator function of which value is one if the condition is true;

otherwise, the value is zero.

4.3 Call Admission Probability

Here we calculate the call admission probabilities of new calls and handoff calls.

In the proposed scheme, the call admission decision is based on measurement and

estimation of 0bE N. Since the calls in dormant state do not generate traffic, we should

know the number of active calls of each class for calculating 0bE N. Let us assume that

the system state is s . Let us introduce a state s, which explicitly describes the number

of active calls in each of two links for all classes. Let ildenote the number of class- i calls


36


37/70


whose uplink channel state is active. And let im be the number of downlink active calls

with class- i . Then, sis defined as

s 0 0 1 1( , ,....., , ,....., , )i i L Ldef l m l m l m (4.8)

Let us call sas an activity state ofs .

Let ( )Q s be the state space of all feasible activity states of

system state s . Then

( ) { : (0 )s i iQ s l n= and (0 )i im n

for0 i L } (4.9)

Let us calculate the mean0b

E Ns of uplink and downlink channels of class- k

call in state swith the background noise ignored.

4.3.1 Uplink Bit energy to noise density ratio

Consider the uplink channel of an ongoing call with class- k.Let jCbe the average

of the signal powers that the base receives from mobiles with class- j call. Then, the

average home-cell interference of class- kuplink channel in state s, denoted by

( )uk sH, is as

1

( )L

u

k s j j k

j o

H l C C

=

= (4.10)

Let jNbe the mean number of class- j calls in process within a cell. Then, the

mean number of class- j calls whose uplink channel state is active isu

j jN. Whenu

denotes the ratio of the uplink interference from other cells to that from home cell, the

mean interference of the class- kuplink channel coming from other cells in state s,

( )uk sO , is as

1

0

( )L

u u u

k s j j j

j

O N C

=

= (4.11)


37


38/70


The average of total interference for class- kuplink channel is a sum of ( )u

k sH

and ( )u

k sO . Then, the mean 0bE Nof the class- kuplink channel in state s, ( )

u

k sM,

is as

( ) .( ) ( )

u k uk s u u u

k s k s k

C WMH O R

=+

(4.12)

The mean received signal power at a base for any class call is proportional to the

uplink data rate and the required 0bE Nof the class. Therefore, ( ) ( )u u u u

j k j j k k C C R R = .

Then

1 1

0 0

( )u

u u kk s L L

u u u u u u u

j j j k k u j j j jj j

WM

l R R N R

= =

=

+ (4.13)

4.3.2 DownlinkBit energy to noise density ratio

Consider the downlink channel of a class- kcall in state s. Let jp be mean

transmission power at a base for a class- j user, xPbe total transmission power of a base

in cell x , and let zdenote the portion of overhead channel (e.g., pilot channel) power (

ohP) for the maximum transmission power ( maxP ). That is, maxohz P P= and is assumed to

be a fixed value. When the home cell is numbered as cell 0, 0( )ohP P z and

1

0

0

L

oh j j

j

P P m p

=

= . So,1

0

0

(1 )L

j j

j

z P m p

=

. As the load increases and 0Papproaches to

maxP , 0Pmay be regarded as1

0

( ) (1 )L

j j

j

m p z

=

. Since admission control becomes full of

interest mainly at heavy load, we approximate the transmission power of the home cell

base as follows:

1

0

0

1.

1

L

j j

j

P m pz

=

=

(4.14)


38


39/70


We know that in the proposed CAC , each active mobile reports the measured

0bE Nfor its downlink channel periodically and the base calculates the avera

downlink using the reported 0bE Nvalues. Let us calculate the average downlink 0bE N

for class- kcalls. We assume that the average of the measured downlink 0bE Ns for

class- k calls is equal to the measured 0bE Nof a class- kmobile at an average

location (the tagged mobile). This is known as the concept of average location. We

assume that the transmission power required for the tagged mobile is same as mean

transmission power, kp , for class- kcalls. We consider the simplest model for radio

channel, which is a propagation loss inversely proportional to the distance between

transmitter and receiver. Let 0dbe the distance from the base of home cell (cell 0) to the

average location. When the average orthogonality factor in downlink is represented by ,

home cell interference of the tagged mobile at average location in state sis as

0 0( ) ( )(1 )

d

k s kH cd P p = (4.15)

where is the path-loss exponent, and and c are constants.

Let us evaluate other cell interference of the tagged mobile. Assume that only

cells in two tiers are considered as interference source. There are six cells in first tier and

12 cells in second tier, which are numbered from 1 to 18. When xddenotes the distance

from the base of cell x to the tagged mobile at average location, the interference at the

tagged mobile from the base of cell x is x xcd P

.Then, other cell interference at the

tagged mobile is18

1

x x

x

cd P

= . Since the mean number of class- j calls in downlink active

state isd

j jN

1

0

1.

1

Ld

x j j j

j

P N pz

=

=

(4.16)

Therefore, the downlink 0bE Nof the tagged mobile in state s, denoted by

( )dk s

M, is given by


39


40/70


0

1 1 18

0 0 1

( ) .1 1

(1 )( ) ( )( )1 1

d d kk s L Ld

dkj j k j j j x x

j j x

W cd pM

Rm p p N p cd P

z z

= = =

=

+

(4.17)

Let

18

0

1( )d x

xd d

== .

drepresents the average ratio of the downlink

interference from other cells to that from home cell and since the mean transmission

power at a base for a call is proportional to the data date and the required 0bE Nof the

call, ( ) ( )d d d d

j k j j k k p p R R = . Therefore, ( )

d

k sMis

1 1

0 0

(1 )( )

(1 ) (1 )(1 )

dd d kk s L L

d d d d d d d

j j j k k d j j j j

j j

z WM

m R z R N R

= =

=

+ (4.18)

4.3.3 Call Admission Probability

Let us calculate the probability that the activity state is swhen the system is in

state s . This probability is denoted by ( )sP s| . Since the activity of a call is

independent of any other call and the uplink activity of a call is also independent of its

downlink activity , ( )sP s| is given by

( ) ( )

1

0

! !( ) { ( ) (1 ) }{ ( ) (1 ) }

! ! ! !

j j j j j j

Ll n l m n mj ju u d d

s j j j j

j j j j j j j

n nP s

l n l m n m

=

| =

(4.19)

In the proposed CAC scheme, a new call request and a handoff call request with

class- i are accepted if the conditions on the estimates of uplink and downlink 0bE Ns

for any class- k, (3.3) and (3.4) are respectively satisfied. When a class- i call request

arrives, the base calculates the estimates using (3.1) and (3.2), respectively. That is, if a

class- i call requests a connection in state s, the base calculates , ( )u

k i sE and , ( )

d

k i sE for

any k, as follows:


40


41/70


1

,

1( ) { . }

( )

u uu i ik i s u u

k s u k

RE

M W

= + (4.20)

1

,

1( ) { (1 ) . }

( )

d dd i ik i s d d

k s d k

RE

M W

= + (4.21)

Let us introduce a function, , ( , )u

k i sg x , whose value is zero if the estimate of

average uplink 0bE Nfor class- kwith at least one active call in state sis smaller than

x ; otherwise the value is one.

, ( , )u

k i sg x = {0, if0kl> and , ( )u

k i sE x< ; 1, otherwise}

Similarly, , ( , )d

k i sg x denotes a function which has a value of zero when the

estimate of average downlink 0bE Nfor class- kin state sis smaller than x ; otherwise

the value is one. That is

, ( , )d

k i sg x = {0, if 0km > and , ( )d

k i sE x< ; 1, otherwise}

We know that ,u

k iand ,d

k i, respectively, denote the uplink and downlink threshold

0bE Ns of a class- kcall for controlling a class- i new call request. Then, the probability

that a class- i new call obtains an admission in state s , ( , )nA s i , is

1

, , , ,

( ) 0

( , ) { ( ) ( , ). ( , )}i

s

Ln u u d d

s S s k i s k i k i s k i

Q s k

A s i I P s g g

+

=

= | (4.24)

Similarly, we know that ,u

k i and ,d

k i , respectively, denote the uplink and

downlink threshold 0bE Ns of a class- kcall for controlling a class- i handoff call

request. Then, the probability that a class- i handoff call

obtains an admission in state s , ( , )hA s i , is

1

, , , ,

( ) 0

( , ) { ( ) ( , ). ( , )}i

s

Lh u u d d

s S s k i s k i k i s k i

Q s k

A s i I P s g g

+

=

= | (4.25)

whereis S

I+ checks whether there are available receiver elements and spreading codes.


41


42/70


4.4 Performance Measures

The various performance measures that are considered are

1. Blocking probability of class- i new calls.

2. Failure probability of class- i handoff calls.3. Outage probabilities of class-i call in progress for both the uplink and the

downlink.

4. Throughputs of the uplink and the downlink.

4.4.1 Blocking probabilities of handoff and new calls

Let ibe the blocking probability of class- i new calls. The connection request of

a class- i new call in state s is rejected with the probability {1 ( , )nA s i }. Thus, iis

given by

{1 ( , )}. ( )n

i

s S

A s i s

= for0 1i L (4.26)

Similarly, let ibe the failure probability of class- i handoff arrivals. The class- i

handoff call in state s is forced to terminate with the probability {1 ( , )hA s i }. Thus,

iis given by

{1 ( , )}. ( )h

i

s S

A s i s

= for0 1i L (4.27)

4.4.2 Outage Probability of a call in progress

Let us calculate the outage probability of a call in progress. The outage

probability of a call is the probability that the measured 0bE Nof the call is smaller than

the required 0bE Nfor maintaining adequate transmission quality. Letu

iandd

ibe the

uplink and downlink outage probabilities of class- i calls. We know that ,u

k iand ,

d

k i,

respectively, denote the required 0bE Ns for uplink and downlink traffic of class- i calls.

Then


42


43/70


0 { ( ) }( )

0

( )

( )

( )( )

u ui i s i

s

i

s

s l MQ su

is S s l

Q s

P s I I

sP s I

>

|

=|

(4.28)

0 { ( ) }( )

0

( )

( )

( )( )

d di i s i

s

i

s

s m MQ sd

is S s m

Q s

P s I I

sP s I

>

|

=|

(4.29)

4.4.3 Throughput of uplink and downlink

Let uand dbe the throughputs for uplink and downlink, respectively. Here, the

throughput is a data rate under the condition that the measured 0bE N

is larger than or

equal to the required 0bE N. Therefore

1

{ ( ) }( ) 0

( ) ( ) u ui s i

s

Lu

u s i iMs S Q s i

s P s I l R

=

= | (4.30)

1

{ ( ) }( ) 0

( ) ( ) d di s i

s

Ld

d s i iMs S Q s i

s P s I m R

=

= | (4.31)

Then, the total system throughput is a sum of uand d.

4.4.4 Determination of Handoff Call Arrival Rates and Carried Load

The flow balance equations are derived using the arrival rate of class- i handoff

calls, i( 0 1i L ). Since the overall system is assumed to be homogeneous in

statistical equilibrium, the arrival rate of handoff calls with class- i should be equal to the

departure rate of class- i calls toward neighboring cells. That is

( )i i is S

s n

= for 0 1i L (4.32)

As seen in (4.32), the handoff arrival (departure) rates are dependent on state

probabilities, while the state probabilities are derived using the handoff arrival rates. On

the other hand, as seen in (4.13) and (4.18), the mean number of ongoing calls for each


43


44/70


class, jN( 0 1j L ), is used to calculate the average uplink and downlink 0bE Ns

and these are in turn used to derive the state probabilities. Conversely, the mean number

of ongoing calls is obtained from the state probabilities. That is, when calculating the

mean number of ongoing calls for each class, we meet the same problem as whendetermining handoff arrival rates. To solve these problems, we use the following iterative

algorithm that begins with the initial guess for handoff arrival rates and the mean number

of ongoing calls.

4.5 Algorithm

Figure 4.1 shows the flowchart for the algorithm to be implemented for the

performance analysis of the proposed CAC scheme.

Step 1: Set an initial value for i( 0 1i L ) which is calculated using the blocking

probabilities of handoff arrivals and new call arrivals with class-i by iand i,

respectively as follows:

, { (1 ) (1i h i i i iP = + (4.33)

where ,h iPis the probability that the connection is released by handoff departure and

, ( )h i i i iP = + . Assuming that 1i

= and 1i

= ,

,

,(1 )

h i ii i i

h i i

P

P

=

; (4.34)

We set the initial value ofito i i i .

Step 2: Estimate an initial value of jN( 0 1j L ). Let us assume that all cells are in

the same state. Then, (19) and (24)are replaced by the following equations:

1

0

( )(1 )

uu u k

k s Lu u u u

u j j j k k

j

W

Ml R R

=

=+

(4.35)

1

0

(1 )( )

(1 ) (1 )(1 )

dd d kk s L

d d d d

d j j j k k j

z WM

m R z R

=

=

+ (4.36)


44


45/70


We obtain the initial values for iteration from (4.34)(4.36).

Figure 4.1 Flowchart showing the performance analysis.

Step 3: Compute the stationary probabilities ( )s s using (4.2)(4.25).


45


46/70


Step 4: Compute the mean rate of handoff arrivals, ,i new (0 1i L ), using (4.32)and

compute the mean number of ongoing calls, ,i newN ( 0 1i L )

, ( )i new is S

N n s

= (4.37)

Step 5: Let ( 0) > be a predefined small value. We introduce a function iforisuch

that

i= {1, if,

1i new

i


47/70


CHAPTER 5

RESULTS


47


48/70


5.0 Introduction

In this chapter, first we will discuss about the modeling of the traffic with

different bit rates and their priorities set at different levels. Then the corresponding

performance is also evaluated in terms of throughput, packet loss and packet delay. This

is done using the Network Simulator NS2 (version 2.31). Then secondly, the performance

of the proposed CAC algorithm is analyzed and the corresponding plots are plotted using

MATLAB ( version 7.1).

5.1 NS2 Simulation Results

Ns-2 is an open source discrete event simulator used by the research community

for research in networking .It has support for both wired and wireless networks and can

simulate several network protocols such as TCP, UDP, multicast routing, etc. More

recently, support has been added for simulation of large satellite and ad hoc wireless

networks. The ns-2 simulation software was developed at the University of Berkeley. The

standard ns-2 distribution runs on Linux. However, a package for running ns-2 on

Cygwin (Linux Emulation for Windows) is available.

5.1.1 Simulation scenarioThe simulation scenario evaluates the performance of the traffic model in terms of

the throughput, packet delay and packet losses. The simulation topology of this scenario

is simple. It consists of 8 mobile nodes : 4 source nodes and 4 destination nodes. Each

node is transmitting with a different priority. With respect to the source nodes, Node 0 is

given a higher priority than Node 2, which is given also a higher priority than Node 4,

which, in its turn, is given a higher priority than Node 6 .

Each source is a Constant Bit Rate (CBR) source over UDP (User Datagram

Protocol). The size of a transmitted packet is 512 bytes. Transmission rate of the nodes is

set at different values. Node0 transmits at 400 Kbps, Node2 at 500 Kbps, Node4 at

600Kbps and Node6 at 800 Kbps. We assumed that the nodes are in the transmission

range at a constant distance of 195 m. The simulation time lasted for 80 sec. Table 5.1

shows the specifications of the traffic model used.


48


49/70


Channel Wireless Channel

Radio Propagation Model Two ray ground propagation

Network Interface Wireless Physical

MAC Type 802.11 MAC

Antenna Model Omni antenna

Maximum Queue size 50Queue type Drop tail/Priority queue

Number of mobile nodes 8

Routing Protocol DSDV

802.11 MAC RTS threshold 3000

802.11 MAC Basic rate 1 Mbps

802.11 MAC Data rate 2 Mbps

CBR Packet size 512 Bytes

Node 0 CBR rate 400 Kbps



Node 6 CBR rate 800 KbpsSimulation time 80 sec

Table 5.1 Traffic Model Specifications

Figure 5.1 shows the NAM layout of the simulation topology

Figure 5.1: NAM layout of the topology.


49


50/70


Node 0 starts transmitting at time T=1.4 seconds as shown in figure 5.2.

Figure 5.2 Node0 starts transmitting.

Node 2 starts transmitting at time T=10 seconds as shown in figure 5.3

Figure 5.3 Node 2 starts transmitting.


50


51/70


Node 4 starts transmitting at time T=20 seconds as shown in figure 5.4

Figure 5.4 Node 4 starts transmitting.

Node 6 starts transmitting at time T= 40 seconds as shown in figure 5.5.

Figure 5.5 Node 6 starts transmitting


51


52/70


In the next section we clearly analyze the performance of the above traffic model

in terms of the packet throughput, average packets end to end delay and the packet drop

rate.

5.1.2 Performance Evaluation

5.1.2.1 Throughput

Figure 5.6 Throughput plot using Xgraph

Figure 5.6 shows the throughput plot for the above designed traffic model. Node 0

starts transmitting at time T =1.4 sec while Node 2 starts transmitting at time T=10 sec.

During the period of time [1.4 sec, 10 sec] Node 0 is the only transmitting node using the

entire available bandwidth. This justifies the high performance of Node 0 during the

specified interval of time. At time T=10 sec, Node 2 starts transmission hence sharing


52


53/70


channel resources with Node 0. This explains the heavy reduction of bit rate. In addition,

the bit rate plot experiences heavier oscillations and reduction as the number of

transmitting nodes increases. Oscillations are reflected in heavy disorders in network

performance.

5.1.2.2 Average packets end to end delay

Figure 5.7 Average Packets end to end delay plot using Xgraph

Figure 5.7 shows the average packets end to end delay plot for the above designed

traffic model. When the number of nodes that are sharing the network resources

increases, then the delay significantly increases and readjusting CW of each node takes a


53

8/7/2019 Call Admission Control

call admission control for multimedia direct sequence code division multiple access (ds-cdma)...

Documents