Enhance user experience based on traffic in operator network

(1)

Linköpings universitet SE–581 83 Linköping

Linköping University | Department of Computer science

Master thesis, 30 ECTS | Computer Science

2016 | LIU-IDA/LITH-EX-A--16/054--SE

Enhance

user

experience

based on traffic in operator

network

Sruthi Kodoth

and

Juan Manuel Jiménez

Supervisor : Ola Jonsson, Anders Lundström and Robert Forcheimer Examiner : Niklas Carlsson

(2)

Upphovsrätt

Detta dokument hålls tillgängligt på Internet – eller dess framtida ersättare – under 25 år från publiceringsdatum under förutsättning att inga extraordinära omständigheter uppstår. Tillgång till dokumentet innebär tillstånd för var och en att läsa, ladda ner, skriva ut enstaka kopior för enskilt bruk och att använda det oförändrat för ickekommersiell forskning och för undervisning. Överföring av upphovsrätten vid en senare tidpunkt kan inte upphäva detta tillstånd. All annan användning av dokumentet kräver upphovsmannens medgivande. För att garantera äktheten, säkerheten och tillgängligheten finns lösningar av teknisk och admin-istrativ art. Upphovsmannens ideella rätt innefattar rätt att bli nämnd som upphovsman i den omfattning som god sed kräver vid användning av dokumentet på ovan beskrivna sätt samt skydd mot att dokumentet ändras eller presenteras i sådan form eller i sådant sam-manhang som är kränkande för upphovsmannenslitterära eller konstnärliga anseende eller egenart. För ytterligare information om Linköping University Electronic Press se förlagets hemsida http://www.ep.liu.se/.

Copyright

The publishers will keep this document online on the Internet – or its possible replacement – for a period of 25 years starting from the date of publication barring exceptional circum-stances. The online availability of the document implies permanent permission for anyone to read, to download, or to print out single copies for his/hers own use and to use it unchanged for non-commercial research and educational purpose. Subsequent transfers of copyright cannot revoke this permission. All other uses of the document are conditional upon the con-sent of the copyright owner. The publisher has taken technical and administrative measures to assure authenticity, security and accessibility. According to intellectual property law the author has the right to be mentioned when his/her work is accessed as described above and to be protected against infringement. For additional information about the Linköping Uni-versity Electronic Press and its procedures for publication and for assurance of document integrity, please refer to its www home page: http://www.ep.liu.se/.

c

Sruthi Kodoth and

(3)

Abstract

The increasing usage of numerous mobile applications can cause impairments in cel-lular network performance. Such impairments result in performance degradation that can reduce the satisfaction level of subscribers. Consequently, subscribers may switch between different network operators to get good user experience. Thus the success of any network operator will primarily depend on the ability to ensure quality of experience (QoE), where QoE is a measure of the subscriber’s satisfaction level and is closely related to the per-formance of networks. Our work aims to identify the key perper-formance indicators (KPI) which in turn can comprehensively model the QoE. Since the popularity of web browsing and video streaming applications continues to increase rapidly, analyzing the KPI of such applications will help to identify the parameters which degrade network performance the most. The analyzed KPIs are tested with different user equipments and different network load. This thesis work also includes tuning the Radio Network Controller (RNC) param-eters to analyze the variation in user experience. Important performance metrics of web browsing and video streaming applications have been considered to measure the QoE. A test environment for QoE estimation was developed using real Radio Network Controller (RNC) and simulatable models of the Core Network(CN) and User Equipments (UEs). Sim-ulations with this test set up and subsequent analyses help to identify some of the RNC pa-rameters which influence the QoE. Furthermore, simulatable models of widely used UEs such as iPhone 6 and iPhone 3 were included in the test environment to assess their relative performance for web browsing and video streaming applications. Our simulation results confirm the superior performance of iPhone 6 which reinforces the reliability of our test bed. Finally, the simulations also helped to illustrate the degradation in QoE caused by the increase in RNC load.

(4)

To my brother Krishnaraj Kodoth

Sruthi Kodoth

To my aunt and godmother Lola, my grandfather Luis, to the

siblings I never had Benji and Pablo and to those who did not

believe in me.

Juan Manuel Jiménez

(5)

We would like to express our gratitude to

• Åsa Lindgren for providing us a great opportunity to pursue the thesis work at Ericsson AB, Linköping.

• Our supervisors Ola Jonsson and Anders Lundström for the fruitful discussions throughout the thesis work. Despite their busy schedules, they responded very quickly to our emails and allocated time every week in order to ensure that we never deviated from the defined scope of the work.

• Dr. Niklas Carlsson and Dr. Robert Forcheimer for their numerous useful suggestions and continued support through out the thesis work.

• Dr. Prakash Harikumar for his constant support throughout our M.Sc studies. I (Sruthi) would like to express my gratitude to

• Dr. V. Thankamani, my mother-in-law for the unstinting support, unconditional love and encouragement throughout my M.Sc studies. Resourceful and resilient, she has a nimble intellect yoked to an indomitable spirit that is anchored on immaculate probity. • Dr. Niklas Carlsson who kindled in me a genuine interest for Computer Networks through his courses. He led me along the magnificent TDTS06 avenue and provided the imposing arena of TDTS21 where I subdued my inexperience and honed my skills for research. His erudition and inspiring guidance constitute the touchstone for the consummate preceptor.

• My dearest parents Sailaja Kodoth and M. Madusoodanan for their eternal love and prayers which help me to overcome all impediments.

• My most loving father-in-law M. Harikumar for the boundless affection and blessings. He is a polymath whose colossal knowledge is further embellished by his ingenuous-ness.

• My husband Dr. Prakash Harikumar for helping to improve my technical writing skills. His effortless use of complex English words always astounds me.

• The heavenly place called Linköping. The idyllic and tranquil settings of Linköping constituted the perfect ambience for translating my potential into performance. I (Juanma) would like to express my gratitude to

• My aunt and godmother Lola, because my mind is linked to you when I hear the word mum.

• My grandfather Luis. Your knowledge in maths and cards has exposed my path until here. 13 years without the wisest man.

• My grandmother Manola and my grandfather Emilio. Your discipline and fight were a mirror to look. They are interwoven into my heart, and also you.

• Benji and Pablo. Real friendship belongs forever and still is going on harder and harder, thanks for your persistent help.

• Sweden for the remarkable treatment and for the unique and unprecedented opportu-nity.

• Dr. Unmesh D. Bordoloi and Arian Maghazeh who gave me a chance to come back to this fantastic place called Linköping.

(6)

• Google and Stackoverflow, for improving my troubleshooting skills during my educa-tion.

• Manolo, who saw some light on me and bet blindly on me. You introduce me to the fantastic world of Magic: The Gathering and Poker. I am also grateful to all of the friends there. 84 PAM.

• My cousins, aunts, uncles and myriad friends in Spain.

• The innumerable friends in Sweden, legendary memories with all of you. Special men-tion for Sruthi, Erind, Bego and Oskar.

• My cousins Miguel, Raúl and friend Ismael.

• Ared, because you showed to me the meaning of sharing. Thanks for your support and comprehension during the first year of the Master.

• The unforgettable flatmates David “Mi chiki”, Paula, Alejandra “Don’t call my name” and Angela. Crazy 80’s and laughing moments, Domino Dancing and surviving the good and the bad days, but always together.

• Emma J, to show me the real courage. First time I saw you I knew I had in front of the strongest. Your unknown impact on my motivation reached top levels.

• Elena and her family. During 5 years, Paco you were like a father, Bea like a real sister and Elena, you guided me energetically to find my current status. Wish you the best wherever you are.

(7)

List of Figures

2.1 WCDMA architecture . . . 7

2.2 WRAN connected to CS and PS Core Networks . . . 8

2.3 UE states . . . 11 2.4 Connections . . . 12 2.5 Transport network . . . 13 2.6 WCDMA channels . . . 14 2.7 Protocol stack . . . 16 3.1 QoE vs QoS . . . 18 5.1 Simulator tool . . . 26 5.2 Data flow . . . 27

5.3 Post processing phase . . . 28

5.4 Wireshark set up . . . 28

6.1 Web browsing scenario . . . 31

7.1 Web browsing KPI . . . 34

7.2 Analysis of SLI value for multiple UEs . . . 36

7.3 Analysis of KPI while tuning the RNC parameters . . . 38

7.4 Total signaling load . . . 39

7.5 Channel switching intensities . . . 39

7.6 Analysis of QoE while tuning the RNC parameters . . . 40

7.7 Number of active UEs . . . 41

7.8 Payload . . . 41

8.1 Overview of video streaming system . . . 44

9.1 Video streaming KPI’s . . . 51

9.2 SLI Vs Processor load . . . 53

9.3 KPI variation with processor load . . . 53

(10)

List of Tables

7.1 SLI generation . . . 33

7.2 iPhone capability . . . 35

7.3 iPhone 3 Vs iPhone 6 . . . 35

9.1 SLI generation for video streaming . . . 51

(11)

List of Tables

Terminology

QoE Quality of Experience

SLI Service Level Index

KPI Key Performance Indicator

EUL Enhance UpLink

HSDPA High Speed Downlink Packet

WCDMA Wide band Code Division Multiple Access

UE User Equipment

UMTS Universal Mobile Telecommunication Systems

RNC Radio Network Controller

RBS Radio Base Station

RRC Radio Resource Control

RAN Radio Access Network

SLA Service Level Agreement

FACH Forward Access Channel

TCP Transmission Control Protocol

UDP User Datagram Protocol

RA Routing Area

SGSN Serving GPRS Support Nodes

GGSN Gateway GPRS Support Nodes

ARQ Automatic Repeat reQuest

PDP Packet Data Protocol

SDN Software Defined Networking

CDN Content Delivery Networks

HS High Speed

UL Up Link

DL Down Link

UTRAN UMTS Terrestrial Radio Access Network

(12)

1 Introduction

Mobile communication has become a ubiquitous facet of modern-day life. This trend is be-ing spurred by the explosive growth in the popularity of smartphones and tablets as well as the deployment of high data- rate network standards (WCDMA, LTE). However, the de-mands on the mobile network imposed by millions of subscribers added with every passing day presents formidable challenges. Since a portable communication device is employed as an entertainment platform rather than a mere phone, ensuring acceptable quality of service becomes crucial to ensuring market dominance.

In mobile radio networks user experience is usually quantified in terms of call drop rate, number of users per cell, spectral efficiency etc. The traffic in a given mobile network is not constant throughout the time of the day or locality. Hence, an analysis of the network traffic scenarios will allow us to identify the parameters for optimization. It is essential to avoid unacceptable degradation of critical network performance parameters while implementing the optimization.

In today’s mobile network, preponderant portion of data traffic is constituted by Inter-net access for web browsing and video streaming. Since video files contribute the major-ity of the traffic over today’s Internet, the subscribers qualmajor-ity of experience (QoE) depends on certain factors like buffering latency of the video, download speed and resolution of the video/images. Hence, in this work, QoE metrics for these applications have been considered. In order to model the web browsing and video streaming applications, our work utilizes the real Radio Network Controller (RNC), simulated core network and simulated User Equip-ment (UE) consequently we refer to our test environEquip-ment as ’partly simulated’ through out the thesis report. The main focus of the thesis is to model the web browsing and video stream-ing applications in the partly simulated environment. Also, to identify the key performance indicators (KPI) that affects the user experience for these applications. This work also identify and tune the parameter settings inside the radio access network to improve user experience.

We were able to set up a Service Level Index (SLI) model for measuring QoE in the partly simulated environment. The test set up were developed using internal tools at Ericsson. The correctness and reliability of the model is validated by comparing the QoE of iPhone 3 and iPhone 6 for video streaming and web browsing applications. Based on the results we were able to conclude that iPhone6 has significantly better QoE than iPhone3 which conforms the expected result. Also evaluated the impact of KPI and overall QoE for web browsing and video streaming applications while using different UE capabilities, varying network load

(13)

1.1. Background and motivation

conditions and changing RNC parameters. With the relevant results we were able to show the QoE as a function of processor load. Also able to identify different RNC parameters which improves/ degrades the end user experience.

1.1 Background and motivation

The proliferation of smart phones, tablets, laptops and other wireless communication devices has become a defining feature of our technology-driven age. The demand for ever increas-ing data rates in such mobile networks from a rapidly growincreas-ing subscriber base puts a heavy strain on the network infrastructure. The delivery of high quality services to all subscribers while dealing increased number of users with limited amount of bandwidth available is the most challenging issue faced by the network operators. With every passing day, the popu-larity of mobile broadband (MBB) services is increasing. According to recent reports, MBB users constitute the vast majority of broadband users [25]. The development and spread of MBB networks has been accompanied by the concomitant deployment of numerous services such as web browsing and video streaming on mobile terminals. Application such as Face-book, Twitter and Youtube which emblazon the zeitgeist of this age are rapidly migrating to mobile terminals. Such a paradigm shift in the accessibility and content of MBB services has profound implications for MBB service providers. It is no longer sufficient to provide rudimentary communication services to the subscribers. Whether the MBB network/service provider remains a global market leader or falls by the wayside is determined solely by its ability to satisfy the QoE specifications, a key indicator of customer satisfaction. When dis-satisfied customers leave a network, it results in loss of revenue and often has a deleterious impact on brand value [25]. QoE also determines the manner in which MBB services are charged. Consequently, a provider who can ensure reliable QoE over its MBB network will earn greater profits.

Under these circumstances, enhancing user experience in a widely adopted mobile net-work such as 3G requires ingenious solutions.

1.2 Aim

This work aims to perform the traffic analysis in a 3G network and determine key perfor-mance indicators which can have impact on end user experience and measure the user satis-faction level. As part of the project a comprehensive model of the network will be built using standard development tools. Subsequent analyses will be used to propose network tuning techniques to improve the quality of service.

1.3 Research questions

Based on the partly simulated system approach described in Section 1, this work has sought to address the following research questions

• Is the partly simulated environment good enough to measure QoE for users in RNC node ?

• Is it possible to model the real world web browsing and video streaming applications in the partly simulated environment?

• How to measure QoE in the partly simulated environment ?

1.4 Related work

In the literature several works focusing on different benchmarking for QoE improvement and proposed different KPI for applications like web browsing and video streaming have been

(14)

re-1.4. Related work

ported [36, 10, 30, 35, 29, 32]. Harris et al [32] analyze user satisfaction level based on different web metrics like number of requests per second and delay. The proposed assessment method also considers the psychological measures while assessing the QoE. The Casas et al [10] dis-cussed the impact of download rate with QoE. The authors proposed different test cases with web browsing application and youtube application to model the throughput fluctuations and measurement results have been reported in the paper. Sackl et al [35] analysed the effect of different types of user equipment over the radio access network. The analysis is done related to dropped call for each UE type, latency observed for each UE type, traffic analysis per radio conditions etc. Even though numerous work describes analysis of quality of experience based on different network conditions, load conditions, KPIs and proposed different improvement techniques, improvement of QoE by tuning RNC features has not been sufficiently addressed. To the authors best knowledge this work is the first instance of using a simulatable model in contrast to real network implementations for QoE analysis. In this work the QoE estimates have been obtained for the simulatable model for a mobile broad band network. Ericsson analytics expert group, has done a study on customer experience where most important KPI of the web services and video streaming were evaluated in the real time environment. Based on the mean opinion score (MOS), satisfaction level of customers per service usage has been evaluated. In this work, we have focused on modeling the QoE of web browsing and video streaming applications in the partly simulated environment.

Balachandran et al[9] defined the web metrics as a function of radio network metrics. The selected work considered the user level metric and the authors shown that end user ex-perience changes according to the processor load variation. The authors proved by using machine learning that different radio network characteristics has influence on web browsing QoE. However in our work focused on partly simulated environment where we modeled a web browsing and video streaming applications on the top of TCP/UDP layer and we also analyzed the web browsing QoE by tuning some Radio Network Control features. The pro-posed model allows to add more performance metrics for a particular application without much modification and it can be used to test different kind of applications such social media, VoIP, gaming application etc. It will be great advantage for network providers since the dif-ferent versions of RNC can be tested with the same model. The authors of [40] done analysis on QoE and QoS as a combination of user, application and network input. The selected study done a combination of qualitative and quantitative analysis where metrics considered for the analysis are user centric. User centric QoE evaluation lacks the ability to accurately assess the QoE particularly when the user experience is degraded by non-optimal application design. In this method [40] inferior application design can also eclipse the high QoS available in the network. Our work also focused on quantitative analysis of data where user experience value is generated based on the QoS parameters and QoE metric considered for the analysis is ap-plication centric. QoE perceived by the users depends on different apap-plications as well. The proposed approach helps to identify the KPIs related to each application and the overall QoE is measured based on the important KPIs for different applications. Hence, the generated QoE values will be more closer to the provided QoS.

(15)

2 WCDMA radio access network

2.1 WCDMA

Background

Wideband Code Division Multiple Access (WCDMA) is the third generation mobile commu-nication system. WCDMA provides higher speed which is utilized to transfer high volume of data. The first WCDMA standard were launched in 2002. Since it s launch in 2002 WCDMA has been adopted in more than 170 countries with the subscriber base of roughly 2.2 billion. WCDMA continues to dominates the network access standard for smartphones which bears testimony to its numerous advantages.

Features

• It supports bandwidth on demand.

• Two modes of operations are being employed is Frequency Division Duplex(FDD) • Due to wide carrier band, it supports high data rates and increased mutipath diversity. • Efficient packet access mechanism.

• Supports inter frequency handover.

• Reduces the interference and transmission power.

• It supports (i)High Speed Downlink Packet Access (HSDPA), the technique has been in-troduced in 3GPP release 5 which allows higher data speeds and capacity in the down-link. (ii) Enhanced Uplink (EUL) which has been introduced in 3GPP release 6. It is extended with additional transport channel and control to increase the performance in the uplink. Both features helps to increase the speed in the downlink and uplink.

2.2 Basic architecture

When the second generation mobile communication system Global System for Mobile com-munications (GSM) was introduced it only supported a voice service together with some

(16)

2.2. Basic architecture

Figure 2.1: WCDMA architecture

supplementary services such as SMS. At that time the voice service was by default a Cir-cuit Switched service. Later on a packet data services, General Packet Radio Service(GPRS), was added to the GSM standard and by that access to the Packet Switched Core Network (Internet) was enabled. The third generation mobile communication system, WCDMA, was defined to use the same CN architecture as GSM. A WCDMA network therefore consists of a WCDMA Radio Access Network (WRAN) which is connected to a Core Network (CN)[1]. Due to the inheritance from GSM, the CN comprises Circuit Switched Core Network (CS CN) and a Packet Switched CN (PS CN). Figure 2.1 shows the overall architecture of WCDMA. In the early days of the standard it was believed that beside speech also different CS data services would be of importance. But today the only use of the CS domain is for speech and all data is handled in the PS domain. A more detailed view of a WRAN connected to CS and PS core network is illustrated in Figure 2.2. The main components in a WRAN is the Radio Network Controller, RNC, the Radio Base Stations, RBS, Cells and UEs (e.g.,Smartphones). A cell is the coverage area of a radio frequency (RF carrier) where the size and direction of the coverage area is defined by the antenna’s shape and the transmitted power. A typical RBS has a three sector antennas, where each sector defines a cell and where each cell could have one or more RF carriers.

Basic functions

Concepts

• User Equipment (UE):

UE includes all kind of end user devices. The user equipments contains two part, Mo-bility Equipment(ME) and UMTS subscriber Identity Module (USIM). MoMo-bility Equip-ment is used for the radio communication between the UE and RBS. USIM contains the identity of the subscriber and also performs the authentication process based on the subscription information of the particular UE [1].

(17)

Figure 2.2: WRAN connected to CS and PS Core Networks

UTRAN provides wireless connectivity between the UE and CN. It contains Node B (RBS) and RNC (Radio Network Controller). NodeB is responsible for managing the resources needed for radio communication to and from the user equipment. RNC can control one or more NodeB. RNC manages and controls the mobility features. It is also responsible for maintaining the optimal usage of radio resources in a secure way and also handles the handover functions during the cell change [8].

• Core Network (CN):

CN is responsible for switching, routing and connectivity to the external networks [1]. CN contains different network elements described below.

– Serving GPRS Support Node (SGSN)

It is responsible for packet switched delivery to and from UE. SGSN forwards the incoming IP packets to and from the UE that is attached to a SGSN service area. SGSN also provides mobility management, session management, authentication and billing.

– Gateway GPRS Support Node (GGSN)

GGSN handles communication between packet switched UMTS network and ex-ternal network. When GGSN receives packets destined to particular user, it checks whether the user activate and forwards the packets to SSGN associates with that

(18)

UE. GGSN also collects the billing information for external data network usage related to each UE.

– Mobile Switching Centre(MSC) Server

MSC handles the circuit switched services to and from the UE. MSC is responsible for switching function such as call set up, release and routing. As mobile moves between different locations it is important to know the location information to handle the communication between them. The location and other relevant infor-mation about each UE is stored in HLR and lots of network resources are required to access HLR. Hence, all the operators maintains a small database called Visitor Location Registry (VLR) which is integrated with in the MSC. MSC also plays a significant role in handover functions which involves multiple RBS and multiple MSCs.

– Gateway MSC server (GMSC) Server

GMSC provides an interface to the external networks.

– Home Location Register (HLR)

Th database which stores location and other relevant information about each sub-scriber. The UMTS network able to route the calls to appropriate NodeB based on the HLR information. When UE moves the different location it registers with different network, from this information it is possible to determine the Node B so the calls can be routed appropriately.

• Routing Area (RA):A routing area is made up of a number of cells, and the identity of the RA is broadcast by all cells belonging to the RA. When a UE enters a new RA it has to update the network about its new positions by making a Routing Area update [8]. • UTRAN Registration Area (URA):An RA could be divided into smaller areas, URAs

[5].

• Attach/detach: As soon as an UE is powered on, it start scanning WCDMA band to find the strongest cell which broadcast the system information along with the identity of the operator to which the user is subscribed to. As scanning is successful, it will do Routing Area update with cause equals to attach which provides the associated credentials and successfully run the authentication process in order to be accepted by the network[5]. After attaching procedure, UE has to activate the PS services by invoking an Activate Data Profile Request (Activate PDP request) [6]. During that time UE will be assigned IP address and is ready to communicate through the Internet. When UE is turn off the power it will send a Deactivate PDP request thereby its IP address will be releases and update the status of the UE as detached.

• Connection handling: In order to support most commonly used services such as speech and Internet access, WCDMA allocate different Radio Access Bearers (RAB) with dif-ferent characteristics.A RAB is established between UE and CN where it utilizes fixed bearer services (IU-bearer) between the RNC and CN. The Radio Bearer(RB) uses radio link between RBS and the UE as shown in the Figure 2.4. Speech service is mapped to conversational RAB and Internet services are mapped to interactive or background RAB’s. At the same instant, a UE can have up to 4 RAB’s as an example one for speech and 3 for PS/ interactive/background.

A lot of signaling and processing is involved in establishing RAB and it takes some time also for establishing the RAB. A PS service is characterized by periods of activity and inactivity period. It would be very inefficient to set up and release the RAB for every active and inactive period. This have influence on latency and processing load. Hence,

(19)

the IU-bearer associated with a RAB is kept alive for up to 30 minutes if no ongoing activity. However, RAB that is carried by radio bearer utilizes the radio resources in the Radio Base Station (RBS). The capacity of RBS is limited in terms of hardware, trans-mitted power and interference. Hence, it is import to allocate the radio resources to the users that are involved in active sending or receiving the data. Different RRC states [7] for UE is defined in order to handle the radio resources in an efficient way. The Figure 2.3 describes different RRC states.

An idle UE attached to the network requests RNC for a PS connection by sending cell update message. The RNC and the PS CN will respond and PS interactive EUL/HS RAB will be established. The EUL/HS RAB is commonly used PS RAB for smartphones. EUL stands for Enhanced Uplink and could provide bit rates up to 11 Mbps for sending data from a UE. HS stand for High Speed Downlink Packet Data Access (HSDPA) and provide bit rates up to 84 Mbps on the receiving side of UE.

When the UE is assigned a EUL/HS RAB the connection will be in the sate cell_DCH. It will remain in this state as long as it has data to send. UE will be switched down to the state cell_FACH if it is inactive for more than HSInactivityTime and the RAB will be switched to a PS interactive RACH/FACH RAB. This RAB uses only less resources and provide bit rates up to 32 Kbps for sending and receiving the data. The HsInactivity timer is typically around 1 to 3 seconds.

UE can send and receive small amount of data in the state cell_FACH. If the UE send more data say 256 bytes then it will be switched to state cell_DCH and the EUL/HS RAB will be re-established. If the UE stays inactive for more than FACH inactivity time the state will be changed to PCH URA and the RAB will be PS Interactive URA. UE is not able to send or receive the data directly in the PCH URA. In order to send the data UE has to initiate the request. If the request is successful, it has to switch the state to cell_FACH and use RACH/FACH followed by a state transition to cell_DCH and EUL/HS RAB if it is necessary due to the amount of data [7].

RNC is responsible for handling the decisions related to state transitions and change of connection state, EUL/HS, RACH/FACH or URA/URA. In order to do that RNC is allocated a set of timers and buffers with different sizes for all the cells inside the RAN. RNC controller timers and buffers are configurable which can be utilized to optimize the performance in different directions.

• EUL/HS principles: A UE using a EUL/HS RAB is connected to a serving cell where the UE will share the down link resources with other UE’s also using EUL/HS. The RBS that handles the serving cell has a complex traffic scheduler used to distribute the DL traffic among the UE:s in the cell. Factors that impacts the distribution is the perceived radio quality by UE’s, the capacity of the UE’s, service priority (Interactive higher priority than background) etc. For the up link (EUL) every UE that is connected to the serving cell are granted a minimum bit rate. However, there is also a scheduler that will allow UE’s to use higher bit rates limited by the interference headroom in the cell.

• Mobility: A UE which is attached to the network and in the idle state will update its position every time it changes the RA. It also includes a periodic update which is controller by a timer of 10 minutes. A UE in the state PCH_URA will update its position every time as it enters a new URA. Similarly, UE in the state cell_FACH will update its position every time as it enters a new cell. The UE in the state PCH_URA will update its position every time it enters a new URA.

For a UE in state Cell_DCH a change of position is a process called handover aiming to maintain the connection and flow of data. The handover process includes an active measurement by the UE where it measures the signal levels and signal qualities from

(20)

2.3. Channel switching

Figure 2.3: UE states

the serving cell’s neighbour cells. When a cell better than the serving cell is detected by the UE, it reports the finding to the RNC. The RNC than makes the decision whether a new serving cell is needed or not [7].

• Admission and Congestion Control The admission control is done based on the load conditions in the air, up link interference, down link power and number of requested users. Admission control is able to distinguish between guaranteed service and best effort service and thus, will solve the problem of overload situations. Using these avail-able services it will be allocating resources inside the cell. Thus, it does not do admis-sion control on common channel. Congestion control will reduce the load situation by taking the following actions

– Activate congestion control and minimize the bit rate of non real time applications.

– If the reduced bit rate is not enough then congestion control mechanism triggers the inter or intra frequency handover which switch some of the subscribers to less loaded frequencies.

– Handover to GSM helps to discontinue to connections and protect the quality of the on going connections.

2.3 Channel switching

In order to handle the flow of information WCDMA employs different types of channels. The two basic channel types are common channel and dedicated channel. Channel switching feature is utilized to change the resource allocated to the UE depending on the amount of

(21)

2.4. Transport in WCDMA radio access network

Figure 2.4: Connections

data flow in the uplink and downlink. The dedicated channel is used when the amount of data is large such as voice communication, downloading the web page etc. Due to the support of power control feature and soft handover feature dedicated channel utilizes the radio resources efficiently. Common channel is used to reduce the delay since number of users shares the same resource. Hence, this channel is used to transfer small volume of data.

2.4 Transport in WCDMA radio access network

WCDMA Radio Access Network nodes communicates through the transport network. Trans-port network is IP based. Different transTrans-port channel protocols are specified in 3GPP. Accord-ing to 3GPP all the radio network related functionality and protocols are separated form the functions and protocols in transport network layer (TNL). The WCDMA RAN transport net-work are responsible for transmitting data and control messages between the RNC and Node B(RBS) or between RNC’s. Each UE is provided with a single serving RNC which terminates user and control plane protocols. The transport network layer provides is responsible for pro-viding the signaling bearer for radio network protocols between RAN nodes and also include the functionality of establishing and releasing bearers as informed by the radio network layer [17]. Figure 2.5 shows the functionality of transport network.

2.5 Radio interface

The protocol stack of radio interface involved in communication between WCDMA radio access networks and handset contains 3 layers [17]. Each layer service the layer above. Fol-lowing are the functionality of the 3 layers:

• Layer 3: This layer is used to control the signaling between the handset an RNC. Func-tionalities of this layer includes controlling radio bearers, mapping of different channel types, handover and other mobility functions.

• Layer 2: It includes MAC and RLC.RLC offers the transparent, acknowledged and un-acknowledged transfer of data. It also supports segmentation and reassembly. MAC

(22)

2.6. HSPA

Figure 2.5: Transport network

protocol maps the logical channel data to transport channel. It also adds the addressing bits required to differentiate the data traffic to different handset.’

• Layer 1: This layer is responsible for transporting the data across the radio link. The main responsibilities included are interleaving, channel quality measurement, transport channel multiplexing etc.

2.6 HSPA

The technology built on UMTS/WCDMA architecture to provide high data rates along with good spectral efficiency. Number of users are competing for high data rates and some appli-cations demands high data rates, consequently HSPA was developed.HSPA is a combination of two mobile telephony protocols HSDPA and EUL.

HSDPA

HSDPA provides greater download speed for the data. High Speed Downlink Shared Channel(HS-DSCH) is used to provide the high data rate. HSDPA allows the resource sched-uler to take into account both channel conditions and traffic conditions.

• Fast link adaption allows maximum channel usage and there a rapid adjustment to data rate in every TTI (2ms) instead of varying radio conditions by means of power control. It also allows NodeB to operate at close to maximum cell power.

• Fast hybrid ARQ(Automatic Repeat Request) with soft combining allows to reduce the overall delay during transmissions. For hybrid ARQ there is no need of signal-ing between NodeB and RNC and subsequently lur/lub delays can be avoided dursignal-ing retransmission. Prior to decoding, UE combines the information of the original trans-misssions with retransmissions which leads to improvement in capacity and robustness [4].

(23)

2.6. HSPA

Figure 2.6: WCDMA channels

• Fast channel dependent scheduling allows the resource utilization efficiently. It mini-mizes the number of resource required per user and allows as many users in the system while satisfying the quality of service requirements [4].

Enhanced uplink (EUL)

EUL is also known as High Speed Uplink Packet Access (HSUPA) is used to improve the performance and uplink capabilities of WCDMA. The design of EUL targeted to reduce the delays, increase the bit rates and capacity of the link. The enhancement features is imple-mented in WCDMA through a new transport channel, Enhanced dedicated channel (E-DCH) [2]. The features of EUL includes

• Fast scheduling feature allows the NodeB to decide on when and amount of data rate transmitting since scheduler is located in NodeB. Without fast scheduling, burst high rate data packets can occurs because of parallel communication with the multiple users.

(24)

2.7. Packet routing and transfer of data

Fast scheduling rapidly adapt to the interference variations and allocate the resources and capacity evenly. Hence, the end users will experience and increase in data rate. • Hybrid ARQ with soft combining feature allows robustness against the transmission

errors. During soft handover time, hybrid ARQ protocol is terminated in many cells and subsequently at least one Node B will receive the data but not all. The NodeB involved in transmission will decode the data and sent ACK or NAK. UE will consider it as successful transmission if it receives a successful reply from at least one NodeB. Hence, hybrid ARQ with soft combining feature also improves the link efficiency and coverage of given data rate.

• Short Transmission Time Interval (TTI) of 2ms is supported by E-DCH to reduce the end user delay during packet transmission and rapid adaptation of transmission pa-rameters.

2.7 Packet routing and transfer of data

GTP is a tunneling protocol as well as signaling protocol utilized for PDP context activa-tion, deactivation and modification. GTP tunnel management messages help to activate and release PDP contexts and their related GTP tunnels between SGSSN and GGSN. IP is the routing protocol inside the packet switched core networks between RNC and SGSN, GGSN to GGSN, SGSN to GGSN. Even though IP routing is employed between GGSNs, the GTP protocol is used in between SGSN and GGSN and between RNC and SGSN. User data pack-ets are tunneled between SGSN and GGSN, GGSN and GGSN and RNC and GGSN. Packet transport between the UE and GGSN depends on PDP context created for a specific UE [3].

2.8 Protocol stack

Protocol structure is shown in the Figure 2.7. GTP provides a way to exchange data using the GTP tunnel. The GPRS support nodes communicate through GTP tunnel where PDP PDU are encapsulated inside the GTP header and exchanged using UDP/IP protocol. The first two layer comprises the user payload which identify the flow between mobile station and UE. The GTP protocol layer identifies the GTP tunnel session. The UDP protocol is used to identify the GTP protocol. The IP header identifies the session flow between the GGSN and SGSN. Each user data packet is encapsulated before transmitting to packet domain network. The encapsulation adds a GTP header which is provides tunneling information to every user data packet. Tunneling information contains an identifier which is responsible for two purposes. It allows to identify the PDP context inside the tunnel end point. It also allows multiplexing of user data which destined for different address to follow single path which is identified by IP addresses [3]. Our web browsing and video streaming application implemented over the GTP protocol stack.

(25)

2.8. Protocol stack

(26)

3 QoE and its relation to QoS

This chapter includes the discussion about the factors influencing QoE in mobile networks and its relation to QoS.

Mobile subscribers always expect their devices to provide high quality of connectivity and performance. Subscribers quality of experience while consuming the mobile content is not strictly based on the high speed which is achieved through high speed wireless technologies like LTE. Subscribers make a subjective assessment on QoE in mobile devices based on some factors such as bandwidth, latency, smoothness etc. Improving the quality of experience of the mobile devices significantly increases the number of costumers consuming the mobile data service. Better the QoE larger will be the number of subscribers utilizes mobile data for long duration. For network and service providers, a good understanding of the relation between QoS and QoE is required in order to provision and enhance the service offerings so that end users will be satisfied with the quality levels.

3.1 QoE Vs QoS

The term QoE can be defined in users perspective and QoS in network perspective. QoE and QoS are interdependent i.e in order to achieve the best QoE in an efficient and cost effective manner, network providers should optimize the use of available services(eg: bandwidth)in relation to subscriber demand [24]. QoE is generally expressed in human feelings like ’good’, ’bad’ or ’excellent’, on the other hand QoS is expressed and measured in terms of network elements.

The latest LTE and future 5G network is designed to support huge capacity, connections for billions of devices and support individual user experience with low latency and response times. The growing demands imposed by the mobile traffic put heavy strain on the cellular networks. One of the key challenges faced by the operators is to meet the change in demand of customers in terms of quality of experience and quality of service. From the operators point of view, network resources and services should be optimized for maximizing the profit. In order to provide high quality of service, operators should be able to manage and control independent service elements. An acceptable quality of experience is achieved by assessing the functionalities of radio access network such as radio scheduling and loading algorithm utilized, QoS mechanism etc. In this way limited resources can be utilized in an efficient way to improve the quality of service. From most of the network providers point of view putting

(27)

3.2. Protocols that impacts QoE

Figure 3.1: QoE vs QoS

investment on network monitoring and management transition may not be a cost effective options. They focus on adding investment on new services rather than on the operation in-frastructure [23]. Hence the service centric management challenge can be properly addressed by providing complete, end to end, service based solutions. These solutions expands the scope of operation management beyond the boundaries of traditional networks. Therefore network providers shift the focus from the network based QoS and Service Level Agree-ment(SLA) to the customer service centric parameters. In this way Service Level Agreements are driven by end user quality metrics rather than network quality metrics and end user ex-perience quality of metrics will be improved because of the following reason:

• Performance is measured based on the service quality perceived by the end user. • Service problems can be detected at an early stage.

3.2 Protocols that impacts QoE

In the proposed approach web browsing and video streaming behavior is simulated over the TCP protocol. To get the more accurate measurement for QoE, it is important to incorporate the delay imposed by HTTP protocol as well. The implementation of HTTP protocol also has influence on end user experience. As an example, HTTP 1.1 implementation without pipeline will perform worse than an HTTP connection which utilizes multiple connections and with different bandwidth utilization. Due to pipelining, HTTP/1.1 changes the performance of HTTP protocol for re-validating cached items and the resulting application will perform in a different way. HTTP/1.1 changes the behavior of the traffic pattern on the Internet which results in large mean packet size, more packets per TCP connection. Since the HTTP protocol contributes to increased network traffic, it is important to simulate the application scenario using HTTP protocol to get more accurate QoE measurements. HTTPS has significant influ-ence on volume of data consumed due to the size of TLS handshake and its lack of capability to handle in network caches and compression proxies. The transition from HTTP to HTTPS imposes new challenges on network providers [31]. Hence it is important to consider HTTPS for the partly simulated environment. Moreover implementation of HTTP protocol on the

(28)

3.3. Impact of signaling in the mobile networks

TCP/UDP layer helps to test different HTTP status code so that more key performance in-dicators can be considered while modeling the QoE. However implementing additional pro-tocol on top of the TCP inside the GTP packet would increase the data sent because of the header overload during the layer segmentation.

Several multi media applications are highly sensitive to end to end delay and jitter, but TCP protocol does not guarantee the end to end delay variation. As an example if some part of the video scene is lost, the TCP will retransmit the packet again which will overlap with the new video scene resulting in worst user experience. Hence jitter is important key parameter to consider while measuring the video QoE . The role of RTP protocol is to compensate for jitter so that receiver application will play the video smoothly. The use of TCP protocol for measuring the QoE of video streaming application in partly simulated environment results in only the worst possible case. Hence incorporating RTP protocol in the partly simulated environment will help to get the real QoE measurement of video streaming application.

Mobile voice assisted services currently addressing exponential growth. IMS (IP Multi-media core network Subsystem) is recognized as the signaling architecture for offering multi-media and voice over IP services (VoIP). In order to enhance the service quality regardless of access network and device , IMS supports signaling. IMS procedure is based on SIP (Session Initiation Protocol). SIP is an application layer protocol designed by incorporating many el-ements of Hypertext Transfer protocol and the Simple Mail Transfer Protocol. SIP is playing an important role in communications applications due to its ability to support all features such as voice, video and data together in one session. One of the key parameters impacting the end user satisfaction is signaling delay and session set up time. Therefore signaling per-formance of SIP plays an important role in end user experience. The perper-formance metrics of SIP protocol impact the QoE of IMS based telephony service [15].

.

3.3 Impact of signaling in the mobile networks

The explosive growth for mobile data traffic is mainly due to the widespread adoption of the smartphone as a communication and entertainment platform. However it is worth noting that specific features of smartphones have deleterious impact on quality of the mobile data traffic. Huge number of smartphones and numerous embedded applications creates signal-ing storm on the mobile networks which has significant impact on number of subscribers. The peak traffic scenarios and network scenarios caused by continuous requests for con-nection establishment from millions of devices generate massive amount of signaling. The smartphone subscriptions are expected to double by 2020 which contribute 8 times more traf-fic to the mobile networks [12]. The design of the smartphones and applications burdens the network operator with signaling traffic and also creates an integrity issues in mobile network. Most of the popular applications like social networking, instant messaging generates thou-sand of signaling keep-alive while the device is on. Also the popularity of cloud application is increasing day by day where the signaling involved in synchronization between local data server is high. Hence scalable network solutions are required to increase the operator revenue and improves the customer’s experience.

Today’s mobile networks offers various solutions to handle the overload signaling situa-tions [21]. Some of the solusitua-tions provided by todays network is discussed as below

• Discard the signaling messages to address the overload situation caused by signaling. In this way overload traffic can be quickly rejected assuming that the overload is a tem-porary situation so that on next attempt the message can deliver successfully. However this strategy is not very efficient one to handle today’s explosive traffic in mobile net-works. Even if the small percentage of messages are discarding, the amount of services successfully delivered to the customer is very less. Moreover in some cases terminals

(29)

3.4. Emerging solutions for QoE improvement

try to establish the connection and as a result of periodic reattempts the intensity of overload situation increases [21].

• A server receiving the signaling level exceeding its capacity leads to rejection of mes-sages which annoys the end user. Hence a load protection and regulation mechanism should be employed in user database management system to protect the system from crash and to ensure the end to end quality of the data. User databases are the promi-nent part of telecommunication networks to deliver the end user services. The user data should not be compromised while solving the signaling issues. User data base is at the tail of the signaling chain. Any degradation in database compromise the whole network which can have significant impact on end user experience. Hence user data manage-ment system needs an intelligent solution while handling the overload conditions to maximize the throughput. one of the solution is by minimizing the resource utilized for discarding the excess traffic and there by improving the end to end throughput. The overloaded signals generated from the user database is continuously monitored and reduce the traffic accordingly. The level of traffic reduction is adjusted in order to make sure that database is performing up to its maximum capacity and thereby avoiding the throughput degradation caused the resource allocated for rejected traffic [21].

However, recommended solutions [18] for handling the explosion in the signaling with out over dimensioning network is by building a robust and scalable network architecture. This approach reduce the amount of traffic and signaling required to offer the network service and also incorporate overload protection mechanism in several part of the network elements to maximize the throughput in the overloaded scenarios [18]. Even though network operators made substantial progress in addressing service degradation due to signaling, the large scale network failure driven by the increase in smartphones and new traffic scenario generated by IoT are major challenges that remain to be solved.

3.4 Emerging solutions for QoE improvement

Recently researchers have proposed a holistic approach in QoE improvement in mobile net-works by facilitating explicit interaction between radio and transport domain [19]. The re-lentless growth of different mobile applications and services are the major cause for network congestion. The traffic congestion will result in packet delay and dropped packets which degrade the QoE. The occurrence of traffic congestion can vary according to time, location and nature of the traffic. Pooling the information from radio domain and transport domain will offer important means for mitigating traffic congestion. Proactive congestion avoidance is such a technique in which the radio domain makes intelligent handover decisions by uti-lizing the congestion information from the transport domain. Another scenario where the radio transport domain interaction benefits QoE is load balance. In order to achieve load balancing information from the radio domain is used by the transport network [19]. For the latest LTE and future 5G standard to provide extremely high data rates traditional means of radio access will prove to be inadequate. To meet extremely high data rates major changes in existing radio access networks are required. The state of the art technique such as cloud RAN technologies helps to reduce latency, wide band spectrum utilization and improves the net-work capacity. The improvement of these performance metrics definitely improves the user experience [19]. Explosive growth in data traffic has huge impact on power consumption and with high cost burdens. Around 80% of contributors for power consumption in radio access network is base station [34]. Recently a QoE guaranteed and power efficient framework for cloud RAN technologies has been proposed [39].

(30)

3.5. QoE considerations in future networks

3.5 QoE considerations in future networks

The enhancement in latest 4G and future 5G technology that supports very high data rate ex-pected to deliver more innovative multimedia service and applications which has benefits for both end user and business people. With dramatic increase in data rates of the 5th generation mobile networks, mobile cloud technology will assume a pivotal role in its successful imple-mentation. One of the key challenge of mobile cloud is the demand for excess storage and computing power which is limited in the mobile itself. However battery life time, communi-cation latency, QoS/QoE and seamless mobility are the import issues which could be solved with network virtualization techniques. The future mobile network 5G should be built with logical resources instead of physical which will allow operators to provide networks based on the service usage. Network slicing is one of the technologies which yields flexibility needed for allocating and reallocating resources on demand [22]. This technology utilizes cloud tech-nologies along with software defined networking and network function virtualization where network functions are built in software packages and deployed in the virtualized infrastruc-ture which increases the scalability of telecommunication networks. Hence by using only the necessary amount of network resources, an energy efficient network system can be built which provides QoS and thus improves the resulting QoE. The advantage of having Software Defining Network(SDN) features is the ability to provide an abstraction of physical layer us-ing network wide programmus-ing behavior which is able to change the behavior of the network that simplifies network management system [22]. The next-generation 5G standard aims to service the Internet of Things (IoT) where billions of devices interact seamlessly over ultra high-speed networks. For the successful implementation of 5G, formidable challenges such as latency, reliability, energy-efficiency and unrivalled QoS, QoE requirements engendered by the myriad applications and services have to be surmounted. Development of a such a robust, scalable and responsive network standard necessitates new air interfaces, protocols and network models to handle the gargantuan volumes of traffic. Software Defined Network (SDN) holds great potential in this regard since it enables the operator to achieve network resilience and adequate QoE at diminished hardware and software costs [16].

(31)

4 Estimation of QoE

This chapter describes the method employed to generate the Service Level Index (SLI) value to measure the QoE and its benefits.

4.1 Modeling QoE

QoE is defined by International Telecommunication Union (ITU) as the overall acceptability of an application or service as perceived subjectively by the end-user.

QoE is perceived by user depends on technical factors and individual preferences. Differ-ent factors influencing QoE are the following:

• Different application types such as web browsing, streaming video, encoding type, transport layer characteristics,transaction delay, protocol used.

• Content type

• Specification of UE’s such as battery life, screen resolution, operating system etc. • Network characteristics like jitter, delay, packet loss, bit rate etc.

• Service level factors like quality, coverage, reliability, cost etc • User demands like expectation and perception

• Depends on various factors like location(urban/remote/rural), access type (station-ary/moving)

Our work focused on modeling QoE based on network characteristics and different appli-cations such as web browsing and video streaming. To model the QoE better understanding of Key Performance Indicator(KPI) which impacts the user quality is required. In our work QoE is measured in the form of Service Level Index(SLI) value. Even though our work fo-cused on 3G, the measurement method is general which can be applicable to LTE and 5G. SLI value indicates the user satisfaction level and is defined in the range between 0 and 10. The SLI values are divided in order to express good, bad and average QoE which is shown in the Figure 4.1.

(32)

4.2. Service level index (SLI)

4.2 Service level index (SLI)

Service Level Index (SLI) is a customer experience score that measures the experience for each individual user. It is a general value that can be used for analyzing the user experience of different applications like web browsing, video streaming, social networks,etc. The value can be compared to itself or to each other to quantify user perception.

• Key Performance Indicator (KPI)

A KPI is an application specific metric which directly impacts the quality of the ap-plication as perceived by the user. An example of KPI considered in web browsing application is download time.

• Reference model framework

To generate the SLI value, a reference model frame work is used. Minimum reference and maximum reference values are generated based on conducting test on different logs. A Linear function is used to approximate the score by comparing the obtained value from the simulator to minimum and maximum reference values. Score has a scale between 0 and 10 and Score is generated for each KPI.

4.3 SLI approach in the partly simulated environment

The QoE analysis for web browsing and video streaming application in a real world environ-ment has been performed by a specialized group with in Ericsson. They measured customer experience based on service usage in the real environment by taking Mean Opinion Score (MOS). Based on the user perception rating and user generated data, the most relevant KPIs were evaluated. Based on these KPIs, the SLI formula has been generated. In this work we have utilized the same formula. In the proposed approach, computed KPI values include the effect of both network performance as well as the applications used by the customers. It allows to add more performance metrics for a particular application without much modifica-tion and it can be used to test different kind of applicamodifica-tions such as social media, VoIP, gaming application etc. However simulator tool does not consider the limited radio resources in the cell for example the scheduler controlled by RBS. Hence the end to end QoE measurement does not consider the delay imposed by those radio resources. The SLI model can also be used to compare different versions of the RNC in order to ensure that implemented feature improves the SLI value.

4.4 Algorithm for generating SLI value

(33)

4.4. Algorithm for generating SLI value

• Based on previous study available in the literature, weights were assigned to each KPI which reflect their influence on QoE for a given application.

• By analyzing the KPI values obtained in large number of simulation, the minimum and maximum reference values were selected.

• A scale between 0 and 10 for ranking the KPI was developed by using a linear function in the range defined by minimum and maximum values obtained in the previous step. • QoE is given by the formula

n ¸ i=1 SiWi/ n ¸ i=1 Wi, (4.1)

where Si is the score, Wi is the weight generated in the previous steps and n indicates the

(34)

5 Measurement setup

Measurement setup for our thesis work consists of two parts such as preprocessing part and post processing part. In the preprocessing part log files where generated and tested using the simulator tool. The script is created for different application scenario with different UE capabilities. Log files were generated based on the defined application script. In the post processing phase, an analyzer tool is utilized for the analysis of the log files . During this phase different KPIs were measured and computed the SLI value. The simulation is done on the top of TCP/UDP.

5.1 Preprocessing phase

As a part of preprocessing phase simulated object has been created and configured. While configuring simulated objects different properties such as IMSI, IP address, traffic patterns, cell information, and log usage depend on the type of objects were considered. The prepro-cessing phase consists of the following steps which are sequentially executed in the partly simulated environment:

• Set up the simulated core network side • Set up the RAN

• Create the UE’s

To create the UEs in partly simulated environment following steps were done: • Load the UE capability parameters (iPhone 3 and iPhone 6).

• Create the IMEI subscriptions for the UE’s which is going to simulate.

• Create the web browsing and video streaming applications by using data descriptions. • Create the traffic behavior and assign to the application defined in the previous step. • Run the command for creating UE which accepts the traffic behavior defined in the

previous step. UE capability, cell information, subscription information and other sta-tistical information is already defined in the application script.

(35)

5.2. 3Gsim

Figure 5.1: Simulator tool

• Analyze using Wireshark the packet flow and monitor the output of the simulator tool called 3gsim which will be used as input for the post processing.

5.2 3Gsim

The 3Gsim tool used to recreate the traffic in a simulated environment shown in the Figure 5.2. 3Gsim includes simulated UE, simulated or real core network and real RNC. In our work we considered the simulated core network. 3Gsim is used to send traffic from both simulated UE’s and simulated core network. 3Gsim is connected to Iur, Iub, Iu-CS,Iu-PS,Iupc interface of RNC. The features of simulated nodes in 3G described in capability set and behaviors. Capability set is used to define what a node is capable of and the behavior indicates how a node behaves during simulation. Simulations are done through command line interface and the simulated results are examined from the statistics generated by the tool. It is worth noting that this work does not include the tuning of radio base station parameters and their resulting impact on QoE.

5.3 Application simulator tool

An application layer simulator (ALS) is utilized to generate the traffic behavior of real world applications like web browsing, file transfer, video streaming and so on. Simulation of differ-ent applications are done on the top of TCP/ UDP. ALS works along with the 3G simulator tool. The traffic behaviors can be divided into two parts such as defining the data descriptions and application script.

• Data Description

Data description describes how the flow of data looks like. It can be defined as a quadruple which includes

(36)

5.4. Limitations of 3Gsim

Figure 5.2: Data flow

– number of uplink bytes : defines the number of bytes in the uplink.

– number of downlink bytes : defines the number of bytes in the downlink.

– Time duration for packet transmission.

– Protocol : defines the protocol used in sending the packets.

3gsim create data description dd1(2500,0,40,1), 2*(0,1134,0,1) shows an example of send-ing one uplink packets and two downlink packets by the application.

Ps data(dd1) shows an example of sending the data with values described in the data description dd1. Figure 5.2 shows the example of splitting the uplink bytes into two packets by TCP protocol while transferring through the network.

• Application script

Application script is the second part of the application layer simulator which describes what to do with the data like send the data or wait for some event etc.

Wireshark which is installed between RNC and SGSN node is utilized to analyze the real log file. This process is utilized to correlate the packets in the partly simulated environment with the real environment in order to make sure that partly simulated environment has al-most same flow of data as in the real environment using an authentic UE. Figure 5.4 shows the location where wireshark is installed.

5.4 Limitations of 3Gsim

Following are the limitations encountered in the simulator during our work

• It was difficult to create an accurate real time scenario due to synchronization issues between client and server.

• Each data description creates new TCP connection which imposes delay in the network. • It was not possible to change the data descriptions dynamically.

(37)

5.5. Post processing phase

Figure 5.3: Post processing phase

Figure 5.4: Wireshark set up

5.5 Post processing phase

Analyzer tool is Eclipse based graphical tool for capturing and decoding traces. The tool contains a framework which takes the log events , decodes them and forward to the analyzers. In our work an analyzer caller UxAnalyzer has been developed to generate the SLI values for the given log files. The figure 5.3 shows the post processing phase.

(38)

6 QoE in web browsing application

This chapter describes web browsing application and the factors influencing QoE for web browsing application. Web browsing application is one of the most commonly used applica-tions in mobile devices. Hence achieving high quality of user experience for this application is of great significance.

6.1 Web browsing performance

Web browsing application uses HTTP protocol on the top of TCP/IP protocol to download the web content. During web browsing the following process occurs:

• While initiating the DNS query to resolve the IP address, UE should transit out of the idle state or PCH state.

• UE establishes the TCP connection with the destination IP using three-way handshake mechanism.

• After establishing the connection , the UE will request the web page object using HTTP GET and server will respond with the required information.

• Multiple objects included in the original page is downloaded using the same HTTP GET/200 OK scheme. The download of objects can be done in parallel or sequentially depending on the browser.

In order to give the user good satisfaction level, all these process should happen quickly. Two main challenges to achieve very responsive web browsing experience is connection la-tency and connected lala-tency. In 3G, Connection lala-tency is defined as time taken to transit from IDLE/PCH to DCH and connected latency occurs when it stays in DCH. The connected latency influence the speed of the acknowledgment packets. Lower latency leads to faster TCP ACK process and thereby overall increase in the download time. In high speed network the packet latency can be a crucial factor for applications like web browsing.

The connection latency plays prominent role in the obtained web experience. The follow-ing process occurs in 3G. Initially the device will be in IDLE OR PCH state. The device takes some time for the switching. During web browsing transaction is initiated by the UE , the first packet in the uplink direction triggers radio resource request. The process of switching

Enhance user experience based on traffic in operator network

Linköping University | Department of Computer science

Master thesis, 30 ECTS | Computer Science

2016 | LIU-IDA/LITH-EX-A--16/054--SE

Enhance

user

experience

based on traffic in operator

network

Sruthi Kodoth

and

Juan Manuel Jiménez

Upphovsrätt

Copyright

To my brother Krishnaraj Kodoth

Sruthi Kodoth

To my aunt and godmother Lola, my grandfather Luis, to the

siblings I never had Benji and Pablo and to those who did not

believe in me.

Juan Manuel Jiménez

Contents

List of Figures

List of Tables

Terminology

1

Introduction

1.1

Background and motivation

1.2

Aim

1.3

Research questions

1.4

Related work

2

WCDMA radio access network

2.1

WCDMA

Background

Features

2.2

Basic architecture

Basic functions

2.3

Channel switching

2.4

Transport in WCDMA radio access network

2.5

Radio interface

2.6

HSPA

HSDPA

Enhanced uplink (EUL)

2.7

Packet routing and transfer of data

2.8

Protocol stack

3

QoE and its relation to QoS

3.1

QoE Vs QoS

3.2

Protocols that impacts QoE

3.3

Impact of signaling in the mobile networks

3.4

Emerging solutions for QoE improvement

3.5

QoE considerations in future networks

4

Estimation of QoE

4.1

Modeling QoE

4.2

Service level index (SLI)

4.3

SLI approach in the partly simulated environment

4.4

Algorithm for generating SLI value

5