Henrik Christiansen

download Henrik Christiansen

of 22

Transcript of Henrik Christiansen

  • 7/28/2019 Henrik Christiansen

    1/22

    End-user QoSWhat, why and how

    Dr. Henrik Christiansen, CTO

    CommWyse A/S, Denmark

    [email protected]

    3G optimization

    Rome March 28th March 30th, 2006

  • 7/28/2019 Henrik Christiansen

    2/22

    2

    Motivat ion

    Data services new opportunities

    Everybody wants optimum QoE

    What is QoE? How to measure QoE?

    What to do?

    3G threats

    Competing technologies -Wimax, WLAN, EDGE Services must work from day one

    What tools do operators need?

  • 7/28/2019 Henrik Christiansen

    3/22

    3

    A quick test

    Can you answer these questions?

    What will be the impact on existing services

    when:

    Traffic increases?

    A new service is added?

    The QoS configuration is changed?

    and how will that impact your business?

  • 7/28/2019 Henrik Christiansen

    4/22

    4

    Presen tat ion overview

    The main challenge

    Application level QoS

    What it is How to measure it

    Use of simulation

    Case studies Summary and conclusions

  • 7/28/2019 Henrik Christiansen

    5/22

    5

    The main chal lenge

    Planning

    Meet requirements: cost, coverage, quality,capacity

    Coverage / capacity interlinked Multiple services and QoS

    Planning per service

    Variable bit rates Multiple QoS classes

    Main challenge:

    Planning a multi-service, multi-datarate network

    Capa-city

    QualityCove-

    rage

    Cost

  • 7/28/2019 Henrik Christiansen

    6/22

  • 7/28/2019 Henrik Christiansen

    7/22

    7

    End-to-end

    RRC

    GMM

    SM

    RANAP

    transport

    TCP

    WCDMA(Radio)

    ATM

    GTP-U

    UDP/IP

    WCDMA(Radio)

    ATM

    GTP-U

    Uu IuPS Gn GiUE Node B SGSN GGSNUser

    EquipmentRadio Network

    ControllerServing GPRS

    Support Node

    Gateway GPRS

    Support Node

    UDP/IP

    Application

    TCP

    IP

    Server

    transport

    Application

    AAL5

    UDP/IP

    AAL5

    transporttransport

    GTP-U

    UDP/IP

    RANA

    P

    GMM

    SM

    PDCP

    MAC

    RLC

    IP

    RNC

    ATM

    RLC

    MAC

    PDCP

    RRC

    GMM

    SM

    Iub

    GTP-U

    GMM

    SM

    MAC Relay

    AAL

    ATM

    AAL

    IP Relay IP RelayIP Relay

    Node BUE RNC

    SGSN

    GGSNInternet

    Servers

  • 7/28/2019 Henrik Christiansen

    8/22

    8

    What impacts QoE ?

    The user(s)

    Handset

    Protocol settings

    Device configs

    Network dimensioning Application usage

  • 7/28/2019 Henrik Christiansen

    9/22

    9

    Appl ication d i f ferences

    Request / response pattern

    Protocol overhead

    Chattiness

    Message sizes

    Inter request times

    QoE depends on the application:Application type and usage must be taken into account

  • 7/28/2019 Henrik Christiansen

    10/22

    10

    Opt im izing QoE

    Optimize what?

    Optimum QoE means

    Happier users? More revenue?

    Better KPIs?

    More customers? The big question is:

    How to measure application level QoS

  • 7/28/2019 Henrik Christiansen

    11/22

    11

    A hol is t ic v iew on netwo rks

    Be proactive!

    Let QoS drive planning & optimization efforts

    Drive tests Actual coverage

    actual QoS - for specific users

    Reactive

    Simulation Detailed protocol insight

    QoE fortypicalusers

    Proactive

    Customers Help desk overload

    Positive / negative feedback

    Reactive

    Churn Unhappy customers

    Bad reputation

    Reactive

    QoE ?

  • 7/28/2019 Henrik Christiansen

    12/22

    12

    Disc rete event s imu lat ion

    Represents everything that happens to a

    packet as one event skips periods where

    nothing happens Simulation approach

    Set goal must be specific

    Set up scenario

    Run simulation(s)

    Analyze / interpret results

    Reiterate if goal is not reached

  • 7/28/2019 Henrik Christiansen

    13/22

    13

    Case s tud ies

    A QoE view on networks by using advanced

    protocol simulation

    What is the impact of adding streamingusers?

    Specific goal: how many streaming users can

    be added if the response time for 80% of a

    group of web browsing business users may

    not increase by more than 20%?

  • 7/28/2019 Henrik Christiansen

    14/22

    14

    Add ing a new serv ice

    - impact on exist ing serv ices

    0

    0.1

    0.2

    0.3

    0.4

    0.5

    0.6

    0.7

    0.8

    0.9

    1

    0 5 10 15 20 25

    With existing services only

    After deploying new service

    CDF of service response times

  • 7/28/2019 Henrik Christiansen

    15/22

    15

    QoE dr iven plann ing

    0

    0,5

    1

    1,5

    2

    2,5

    3

    3,5

    2 3 4 5 6Number of streaming users addded

    Response time (seconds)

    90 % of selected group

    80 % of selected group

    20 %

  • 7/28/2019 Henrik Christiansen

    16/22

    16

    UMTS

    QoS enabled

    Service classes

    Prioritization Packet scheduling

    RRM algorithms

    Other

    Soft handover

    Radio planning

    Need for isolation between cells

    These handles must be setcorrectly in order for the services

    to work as expected

  • 7/28/2019 Henrik Christiansen

    17/22

    17

    Analys ing QoS

    Multi service networks

    QoS goal: preferential treatment of some

    services

    Multiple service classes

    with / without delay guarantees

    Optimizing QoE Improvement of service

    Deterioration of other services

  • 7/28/2019 Henrik Christiansen

    18/22

    18

    QoS d if feren t iat ion

    00.1

    0.2

    0.3

    0.4

    0.5

    0.6

    0.7

    0.8

    0.9

    1

    0 1 2 3 4 5 6

    0

    0.1

    0.2

    0.3

    0.4

    0.5

    0.6

    0.7

    0.8

    0.9

    1

    0 1 2 3 4 5 6

    Web service E-mail service

    All services in same class

    Web higher priority than e-mail

    CDF of service response times CDF of service response times

  • 7/28/2019 Henrik Christiansen

    19/22

    19

    Impact o f t raf f ic grow th

    0

    0,1

    0,2

    0,3

    0,4

    0,5

    0,6

    0,7

    0,8

    0,9

    1

    0 5 10 15 20 25

    light load

    medium load

    heavy load

    CDF of service response times

    QoE target

  • 7/28/2019 Henrik Christiansen

    20/22

    20

    Use QoE as the op t im izat ion target

    QoE is complex

    Measurements and network counters

    tell you about today and yesterday butwhat about tomorrow?

    QoE can be easily predicted by usingsimulation

    QoE impacts your usersand yourbusiness

  • 7/28/2019 Henrik Christiansen

    21/22

  • 7/28/2019 Henrik Christiansen

    22/22

    22

    Thanks!

    Visit our booth during the conference

    Visit: www.commwyse.com anytime!

    http://www.commwyse.com/http://www.commwyse.com/