Bridging QoE and QoS for Mobile Broadband Networks - · PDF fileBridging QoE and QoS for Mobile Broadband Networks ... (NSN) -Head of Solutions ... System QoS, Alarms and Events

HUAWEI TECHNOLOGIES CO., LTD.

Security Level:

47pt

www.huawei.com

Bridging QoE and QoS for Mobile Broadband NetworksDr. David Soldani VP Huawei European Research Centre

21-22 September 2010

http://www.etsi.org/WebSite/NewsandEvents/QoSQoEUserExperience.aspx

HUAWEI TECHNOLOGIES CO., LTD. Page 2

Curriculum Vitae – Dr. David Soldani

• 15 years in ICT industry

• 2009 – present: Huawei Technology Düsseldorf

- VP European Research Centre

- Head of IP Transformation Research Centre (IPTRC)

• 2007 – 2009: Nokia Siemens Networks (NSN)

- Head of Solutions & Services Innovation

- Head of Customer Networks & Solutions (CTO office)

• 1997 – 2007: Nokia (Finland and Italy)

- Various technical & research management positions

• 1995 – 1997: Military Navy, Sirti SpA, Rohde & Schwarz

- Various technical positions

Areas of Expertise (not exhaustive)

• Lead R&D and Customer Services organizational units

• Unit/area strategy formulation and implementation

• Technology and Innovation Management for ICT industry

• Conduct lectures at Universities, Military Academy and ICT Companies

• Perform advanced research in the fields of own expertise

• Provide consulting functions to profit and nonprofit organizations

• Supervise any type of R&D deliverables

• Published/presented many international papers

• Editor in chief and one of the main contributors to several books

• Holder of several international patents

Relevant Experience (not exhaustive)Professional Background

• Solutions for Traffic Management in Mobile Broadband Networks

• Mobile Broadband Networks (TETRA, GSM, EDGE, WCDMA, HSPA, LTE/SAE and WiMAX)

• E2E QoS, QoE and Policy Based Management Solutions

• E2E Service and Network Performance, Network Planning, Optimization and Automation

• Transport Network Layer Technologies (IP/MPLS/Ethernet)

• Fixed Broadband Networks (xDSL, xPON)

Dr. David SoldaniVP European Research Centre

HUAWEI TECHNOLOGIES CO., LTD.Huawei Technologies Duesseldorf GmbHRiesstr. 25, D - 3.0G, 80992 Munich, Germany

Tel: +49-89-1588344095Fax: +49-89-1588344446Mobile: +49-1622047695E-mail: [email protected]

Outline

• Definitions and industry trends

• Objectives of monitoring QoE indicators

• Solution layered architecture

• Bridging QoE and QoS

• Reference case

• Conclusions


Definition: Quality of Experience (QoE)

• Measurement of how well that

network is satisfying the end user's requirements

• Includes the complete end-to-

end system effects (client,

terminal, network, services infrastructure, etc.)

• QoE is also a consequence of

a user’s internal state (e.g.

expectations), the characteristics of the

designed system (e.g.

functionality) and the context(or the environment) within

which the interaction occurs5. Excellent

4. Good

3. Fair

2. Poor

1. Bad

“The overall acceptability of an application or service, as perceived subjectively by the end-user.”by ITU-T P.10/G.100


QoS and QoE parameters – Mapping Model

SubjectiveMeasurements

ObjectiveMeasurements

QoS QoE

SQoS(Network based)e.g. Packet Error Loss Ratio

ESQoS(Service based)e.g. Web Page Loading Time

MOS(Mean Opinion Score)

Objectiveevaluation

Subjectiveevaluation

Viable solutionSQoS = System QoSESQoS = E2E Service QoS


2020 Global Broadband Revenues Forecast

Source: Mobile, Fixed and Wholesale Broadband Business Models, Telco 2.0, June 2010

• Increase of 52% in revenue and more than half the revenue growth will come from wholesale and “two-sided” fees for improved access capacity and quality

• Current forecast expects mobile data traffic to grow by 300x to 500x over the next 10 years: mobile broadband will be worth $138bn (32% of total broadband revenues)

• New ‘upstream’ customers are forecast to generate over $90 billion in broadband revenues globally by 2020


1. Offloading to WiFi2. Offloading to Femtocell3. Radio network enhancements

1. Signaling management2. Macro network offload 3. Radio packet scheduling and prioritization

4. Compression, adaptation and transcoding Compression Rate-adaptation Network-sharing

5. Device-based traffic management techniques 6. Contention management & tuning TCP/IP 7. Deep packet inspection, policy-based traffic shaping & differential

charging 8. End-to-end service assurance and monitoring9. Caching, multicast & CDNs10. Congestion APIs

Key Technologies for Mobile Broadband Traffic Management

Source: Disruptive Analysis, June 2010

In many operators, there is often no single individual who "owns" the issue of data traffic, who can develop a holistic solution

In many operators, there is often no single individual who "owns" the issue of data traffic, who can develop a holistic solution

8. End-to-end service assurance and monitoring


QoE/QoS aspects addressed by all important standardization bodies

MobileVoice

VOIP MTVIPTV Data

Framework

E2E Servicemetrics

Systemmetrics

Measurement Method

GB934 GB938

TR 26.944

TS 102 250 -3/4TS 101 329 - 5

P.NAMS PEVQG.1070

P.861-862 P.563 G.107seriers

Y.1541

E.800

TS 32.410TR 32.814

MDI

Y.1541

TS 32.425

Improving QoE of RT

Communication Services

ETSI Specialist Task Force 354

UE satisfaction criteria

TS 102 250-2

KQI for PS services

ETSI STQ

TR-126

EG 202 670


Audio QoE: the state of art

P.564

G.107E-Model

1996 1998 2002

ETSI ETR 250

TS 101 329-5

P.861PSQM

P.862PESQ

P.563

ITU P.800

P.562

P.561

P.OLQAP.CQO

2005

• PESQ: intrusive method that compares the degraded signal to the original signal• P.563: non-intrusive method and does not need the original signal• P.564: model for assessing voice over IP transmission quality• E-Model: non-intrusive method and its extensions in under study (2009-2012)• P.OLQA: evolution of PESQ and it is still under research (2009-2010)• P.CQO: new model based on P.561, P.563, P.564 and E-Model currently under study (2009-2012)

• PESQ: intrusive method that compares the degraded signal to the original signal• P.563: non-intrusive method and does not need the original signal• P.564: model for assessing voice over IP transmission quality• E-Model: non-intrusive method and its extensions in under study (2009-2012)• P.OLQA: evolution of PESQ and it is still under research (2009-2010)• P.CQO: new model based on P.561, P.563, P.564 and E-Model currently under study (2009-2012)

• E-Model is supported by Huawei SQM/QoE products• E-Model is supported by Huawei SQM/QoE products

E-ModelExtension


Video QoE: the state of art

Reduced Reference

Non ReferenceFull ReferenceSubjective

Estimation

Image resolution

ITU-T J.246

ITU-T J.147

ITU-T J.249

VQEG:

MM Project

VQEG:

RRNR-TV

HDTVITU-T SG12: P.NAMSP.NBAMS

G.OMVASITU-T P.910

ITU-T P.911

ITU-R BT.500

ITU-T J.140

ITU-T J.245

QCIF

CIF

VGA

HDTV

SDTV

ITU-T J.247

ITU-T J.144

ITU-R BT.1683

Completed Ongoing projects

• P.NAMS (from packet header) and P.NBAMS (from payload information): non-reference multimedia, audio and video quality estimation methods

• G.OMVAS defines a Quality Planning Tool (E-Model) for IPTV services (by 2011)• ITU-T also has several IPTV QoE projects, such as G.IPTV_MMRP, G.IPTV_PMP.

• P.NAMS (from packet header) and P.NBAMS (from payload information): non-reference multimedia, audio and video quality estimation methods

• G.OMVAS defines a Quality Planning Tool (E-Model) for IPTV services (by 2011)• ITU-T also has several IPTV QoE projects, such as G.IPTV_MMRP, G.IPTV_PMP.

• Huawei has a non-reference video quality estimation method, called MOS-V, it is a contribution to P.NAMS, and it is supported by Huawei SQM/QoE products

• Huawei has a non-reference video quality estimation method, called MOS-V, it is a contribution to P.NAMS, and it is supported by Huawei SQM/QoE products


IEEE special issue on QoE- March 2010 -

• IEEE Network – Special Issue on Improving Quality of

Experience for Network Services

Guest Editors Dr. Jahan A. Hassan

University of New South, Wales, Australia

Prof. Sajal K. DasUniversity of Texas at Arlington, USA

Prof. Mahbub Hassan,University of New South Wales, Australia

Dr. Chatschik BisdikianIBM T. J. Watson Research Center, USA

Dr. David SoldaniHuawei Technologies Co.,Ltd., Germany

http://dl.comsoc.org/livepubs/ni/public/2010/mar/index.html

Outline





• Reference case

• Conclusions


Operators’ challenges• Very little work available on end user perception of the

services, especially from Internet, that are being offered

• Gaps between network measured quality and user perceived experience

• Mechanisms, procedures and tools to continuously monitor, operate and report QoE indicators

Ban

dw

idth

End to End

Effects!

1

Cause!

User plane

Control plane

2


Main objectives

1. QoE monitoring as support for strategic company decisions

2. QoE monitoring for network optimization, supervision and operation

3. Support for business and customer intelligence processes

Outline





• Reference case

• Conclusions


Functional architecture

Accessnode

Accessnode

Core Network node

Core Network node

Aggregationnode

Aggregationnode

Terminal device

Terminal device

Embedded Agents

Network Elements and Interfaces Probes

Network Elements PM and FM Data

Corenode Corenode

Service corenode

Service corenode

Data Acquisition Data Acquisition LayerLayer

Data Analysis Data Analysis LayerLayer

Data Presentation Data Presentation LayerLayer

Data Query, Filtering, Refinement,Correlation and Processing

System QoS, Alarms and Events

Network Model, Configuration and Topology

E2E Service QoS and Events Data CollectionData Collection

LayerLayer

MBB QoE Indicator reports

MBB QoE Network Optimization, Supervision

and Operation

Customer and Business Intelligence

Functional Functional Objectives/Use CasesObjectives/Use Cases

Data Visualization and Report Generation

Mediation layer

Mediation layer (Optional)

Network Based Network Based ApproachApproach

Terminal Based Terminal Based ApproachApproach


Functional elements (layers)

• Data acquisition layer All relevant information related to QoE monitoring and measuring

Network and Terminal based approaches

• Data collection layer Storage for vertical solutions with continuous monitoring

Real time (seldom) and periodic monitoring

Scale of samples across network and customers

• Data analysis layer Format and data conversion required for different data sources

Data query, filtering, refinement, correlation and processing

Data aggregation, correlating and processing layer

Storage for horizontal solution

• Presentation layer Tools for visualization and report generation

Continuous tracking of mobile broadband experience

Customer experience as a combination of network indicators


Solution segmentation

ServicesServices

• Web surfing• Streaming• File DL/UL• E-mail• …

Network ScopeNetwork Scope

• Whole Network

• Per region

• Per NE/Cell

• Per RAN Vendor

•…

DeviceDevice

• All devices

• Device class

• Device type

• Device group

• …

TechnologyTechnology

• All technologies

• 2G

• 3G

• …

CustomerCustomer

• All customers

• Segment

• Tariff

• …

• Filtering based on all possible combinations


Layered approach

………Throughput of service C of user X,Packet drop ratio,…

Throughput of service B of user X,Packet drop ratio,…

…

Streaming

…

etc

…

VoIP

Email SessionSetup Success RatioEmail SessionSetup time…

Email

Session Setup Success RatioActive Session Throughput…

FTP download

Web Page Loading Time

Active Session Throughput

…

SDU Throughput, SDU Error Ratio, SDU Transfer Delay …

Per User,QoS class, etc.

Look at the pipe(a PDP context)

Monitored ParametersSegmentation / Granularity

Gross and net throughput at each interface (cell, Iub, …),Wide band received and transmitted power (cell), etc.

Per Cell, Itf,NE, etc.

Look at theinfrastructure

Throughput of service A for user X,Packet drop ratio,…

Per service,User, etc.

Look shallowlyinto the pipe

Web Browsing

Service Quality

Look deeplyinto the pipe

“Pipe” = Bearer service

Mobile Broadband Network


Huawei Solution for 3GPP Packet Switched Networks

Backhaul

MSBTS

NodeB RNCSGSN

Internet/Intranet

HLR

Iub Iu-PSGn

GiGrGb

Datacard

eNB

eNB

X2S1-MME

S1-U

S4

S5 S6a

S7

S11

S3

SGi

LTE UE

Uu

Data acquisition from NE & OSS

FM / PM and 3GPP NE

embedded probes

Data acquisition from Interface

Data acquisition from transport NE

Transport NE embedded

and interface probes

Data acquisition from Terminal

Device

manager

Service Quality Manager

3r party

Backbone

GGSN

MME

Serving GW PDNGW HSS

PCU

PCRF

Abis

Outline





• Reference case

• Conclusions


User and service satisfaction criteria

Gold, Silver and Bronze usersG, S, B• User active throughput > User_BW AND

• Session success rate > User_Target_SSR

User bandwidth

8

7

6

5

4

3

2

1

QoS class* (i)

• Air interface (IP) packet error loss rate < User_PELR(i) AND

• 98-percentile delay < User_PDB(i) – 20 ms AND

• Active throughput > User_target_rate(i) AND Video (Buffered Streaming), TCP-based

(e.g., www, e-mail, chat, ftp, p2p file sharing, progressive video, etc.)

Voice, Video (Live Streaming), Interactive Gaming

Non-GBRVideo (Buffered Streaming), TCP-based (e.g., www, e-mail, chat, ftp, p2p file sharing, progressive video, etc.)

IMS Signalling

Non-Conversational Video (Buffered Streaming)

Real Time Gaming

Conversational Video (Live Streaming)GBR

Conversational Voice• Air interface (IP) packet loss error rate < User_PELR(i) AND

• 98-percentile delay < User_PDB(i) – 20 ms AND

Example Services and UsersSatisfaction criterionResource Type

10.0Mb/s

*) Unique combination of QoS parameters identifying the bearer or service data flows, e.g. DSCP


Target performance

8

7

6

4

3

2

5

1

QoS class* (i)

< 384 kb/s, < 64 kb/s

4-13 kb/s, < 384 kb/s, < 60 kb/s

< 384 kb/s, < 128 kb/s

4-13 kb/s

20-384 kb/s

60 kb/s, 30 kb/s

32-384 kb/s

4-25 kb/s

Data Rate (DR)

Video (Buffered Streaming),TCP-based (e.g., www, e-mail, chat, ftp, p2p file sharing, progressive video, etc.)10-6300 ms

Voice,Video (Live Streaming),Interactive Gaming10-3100 ms

Non-GBR

Video (Buffered Streaming),TCP-based (e.g., www, e-mail, chat, ftp, p2p file sharing, progressive video, etc.)

10-6300 ms

IMS Signalling10-6100 ms

Non-Conversational Video (Buffered Streaming)

10-6300 ms

Real Time Gaming, Telemetry10-350 ms, 250 ms

Conversational Video (Live Streaming)

10-3150 msGBR

Conversational Voice10-2100 ms

Example ServicesPacket Error Loss Rate (PERL)b

Packet DelayBudget (PDB)a

Resource Type

a) PDB (radio) = PDB (BS-PCEF) - 20 ms (Delay Correction BS-PCEF)b) PERL (non congested BS-PCEF) = 0*) Unique combination of QoS parameters identifying the bearer or service data flows, e.g. DSCP

Reference [3GPP TS 23.203, TS 22.105]


Relationship between QoE and QoS1. Constant optimal QoEE.g. x1 (sharp threshold) is the boundary below which a user

feels the system reacting instantaneously

2. Sinking QoEThe higher the QoE the higher the impact of a certain additional QoS

disturbance on QoE

3. Unacceptable QoEE.g. Reached x2 (user-

dependent threshold) the transmission might become unacceptably bad in quality

QoSeQoE

QoEQoS

QoE

QoSfQoE

)(

)(


VoIP: Internet low bit rate codec as used by Skype

• Voice quality affected by loss, jitter and reordering

• Type-p reordered ratio is used to quantify the jitter

98th percentile delay < 100 msPERL < 10-2


Web Browsing: perceived quality

• QoS in terms of response and download times

• Web session: (a) requesting a search page; (b) typing and submitting a query; and (c) retrieving the results

Web page loading time < 4s

User throughput > 64 kb/s


Example of Video clips: quality affected by packet loss

• QoE scores of a video clip repaired with two loss concealment schemes at various packet loss rates

PELR < 10-6PELR < 10-6


• Encoder H.264 with one IP packet = 7 × 188 Bytes

• With 100 kb/s and PERL 10-6 PLF = 1.2x10-4

Example of video quality of IPTV Services: quality affected by packet loss


Example of service satisfaction criteria

Web Service Web Paging Loading Time AND

Active Session Throughput

Video Streaming Streaming Mean Active Throughput AND

Streaming Media Active Transport Jitter AND

Streaming Media Loss

Video via HTTP Active Session Throughput

VoIP Call Setup Success Ratio AND

Call Setup Time AND

Call Cut-off Ratio AND

Speech Quality


Satisfaction estimation using Fuzzy Logic

0

1

a ba KPI

Truth (KPI)

0

1

Truth (WPLT)

0

1

Truth (WPLT)

Air Interface Packet Error Rate (%) 98th-Percentile of Delay (ms)

Truth (PD)Truth (PELR)

1 100

Satisfaction = min [Truth (Air Interface Packet Loss Error Rate), Truth (98th-Percentile of Delay )]

Truth Value of a KPI (denoted by

weight) is given by

• Maximum of truth values, if logical OR is used

• Minimum of truth values, if logical AND is used

Example for VoIP:


Bridging QoE and QoS

① Generic Truth function of

a QoS parameter:

② Ex. Truth value (weight) of QoE for Web Browsing:

0.9384Active Session Throughput (kb/s)

0.30.33Web Page Loading Time (s)

Weight (Satisfaction)TruthValueKPI

QoE = Exp(- QoS) +

Satisfaction = min [Truth (Web Paging Loading Time), Truth (Active Session Throughput)]

Configurable parameters: , , , x1 and x2

Needs to be normalized

Outline





• Reference case

• Conclusions


Service quality differentiated G/S/B portfolio

33

Service Quality Differentiation in Mobile Broadband portfolioService Quality Differentiation in Mobile Broadband portfolio

Bronze – Tariff A

Silver – Tariff B

Gold – Tariff C

„Ultimate“(e.g. HD Video)

„Best for Social Media“(e.g. facebook, youtube etc.)

„Best for Surf & Chat“

Service Quality Differentiation

Max Mbps DL/Mbps UL

Avg. Mbps DL/ Mbps UL

FUP / SSD

Montly fee


Bandwidth levels

(In loaded networks, e.g., throttling of Streaming and/or HD Video services for Bronze will make it possible to allocate the freed capacity to Gold, or Silver, so that a higher priority will enable a better user experience.)

Empty network

Bronze Silver GoldBronze Silver Gold

Streaming SurfingHD videoAllocated capacity:

Tariff

Speed

Tariff

Speed

Loaded networkEmpty network

Bronze Silver GoldBronze Silver Gold

Streaming SurfingHD videoAllocated capacity:

Tariff

Speed

Tariff

Speed

Loaded network

(1) Surfing

(2) http Streaming

(3) RTP Streaming

(4) E-Mail

(5) File DL

(6) Other

Example of Mapping for Gold:


QoS benefits

Dissatisfied Users (%)

Traffic (variable)

10%

With QoS on (at cost X)

All on BE (no QoS, with Resources R1)

QoS Gains = T2-T1 (at cost X, for paying the QoS features,

with resources R1)

Traffic = T2

Traffic mix = Tm(held constant)

0%

Traffic = T1

Worst performing

service Satisfied Users (%)

Resource (variable)

90%Over provisioning

Additional Resources R2, at cost Y, for accommodating

the same traffic as with QoS on, satisfactorily

Traffic = T2(both held constant)

Traffic mix = Tm

0%

Profit = Y-X > 0

Over provisioning = R2-R1Worst performing

service

Total profitTotal profit

1) Benefit of radio, transport and gateway QoS functions for HSPA and LTE/SAE

2) Corresponding cost savings with respect to over provisioning

+ additional revenues!

RNC

SGSN&GGSN

Web/Streaming/E-mail/FTP/MMS server

Traffic Generator

Downlink traffic

Node B


Traffic Mix

70%30%Without QoS

60%30%10%With QoS

BronzeSilverGoldUser Mix

35%5%3%7%15%35%

OtherFile

downloadse-mailRTP

streaminghttp

streamingSurfing

Service Mix


User satisfaction criteria for HSPA

NANANAWeb page delay < 10s

128 kb/s

Bronze

Throughput GBR

Throughput 90 kb/sPDL < 300 msPERL < 10-3

Throughput 192 kb/s

Web page delay < 8s

256 kb/s

Silver

Throughput GBR

Throughput 90 kb/sPDL < 300 msPERL < 10-6

Throughput 384 kb/s

Web page delay < 4s

512 kb/s

Gold

File downloads

RTP streaming

http streaming

Web Surfing

Av. Speed (GBR*)

User Type

*) Minimum target rate


Without QoS - RTP Streaming

RTP streaming: % of Satisfied users

Number of users

Threshold = 90%

A = 3300

0

20

40

60

80

100

1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 6500

G

S


With QoS - RTP Streaming

G = (B-A)/A = 30%

0

1000 2000 3000 4000 5000 6000 7000

RTP streaming: % of Satisfied users

Number of users

Threshold = 90%

20

40

60

80

100

G

S

B = 4300


Gains in terms of % of satisfied users• The more users that can be satisfied, at a given offered load, the more

efficiently the spectrum is utilized by the operator

Limit = 90%

BS1 BS2

BS3

Traffic volume(kb/s)

Offered traffic(“Held constant”)

% Satisfied users BS = Bearer Service

Fitness

Fitness = |QoE Actual - QoE Target|


Example: QoS optimization for 5 traffic mixes

Mix 1 Mix 2 Mix 3 Mix 4 Mix 5

-10

0

10

20

30

40

50

60

1 10 100Iteration

Best

fitn

ess

Too much traffic

Too low traffic

Target

Traffic volume: optimisation

really effective!

Outline





• Reference case

• Conclusions


What could be standardized?

QoE as a combination of QoS parameters (e.g. ETSI TS 102

250-2) and truth functions (fuzzy logic)

NGBR and GBR bearer satisfaction criteria and user

satisfaction criteria as a function of them

Call Detail Records (CDR) for service data flows and relevant

interfaces / elements, which shall support those

Requirements for embedded probes and terminal agents for

multi-vendor support


References

1. D. Soldani, M. Li and R. Cuny, QoS and QoE Management in UMTS Cellular Networks, John Wiley & Sons (2006)

2. ITU-T Recommendation P.10/G.100 (incl. Amendment 2), “Vocabulary for performance and quality of service,” July 2006 (2008)

3. A. Takahashi, K. Yamagishi, G. Kawaguti, “Recent Activities of QoS/QoE Standardization in ITU-T SG12,” NTT white paper, 2008.

4. ETSI TS 102 250-2 V1.7.1 “QoS aspects for popular services in GSM and 3G networks; Part 2:Definition of Quality of Service parameters and their computation,” October 2009

5. ETSI EG 202 670 V1.1.1 “Human factors (HF); user experience guidelines for real-time communication services expressed in quality of service terms,” March 2010

6. P. Brooks, B. Hestnes, “Being objective and quantitative about user measurements of QoE,” IEEE Network Comm. Magazine, March 2010

7. M. Fiedler, T. Hossfeld and Phuoc Tran-Gia, “A generic quantitative relationship between quality of experience and quality of service,” IEEE Network Comm. Magazine, March 2010.

8. Kuan-Ta Chen, et al., “Quadrant of Euphoria: a crowd sourcing platform for QoE assessment,”IEEE Network Comm. Magazine, March 2010

9. K. Yamagishi, T. Hayashi, “Parametric packet-layer model for monitoring video quality of IPTV services,” IEEE, ICC, 2008

10. D. Soldani, Hou Xiao Jun, B. Lück, “Strategies for mobile broadband growth: traffic segmentation for better customer experience,” IEEE VTC Spring 2011 (submitted)

Thank youwww.huawei.com

Documents

Bridging QoE and QoS for Mobile Broadband Networks - · PDF fileBridging QoE and QoS for Mobile Broadband Networks ... (NSN) -Head of Solutions ... System QoS, Alarms and Events