Decompiling QoE - ETSI docbox

2 downloads 172 Views 2MB Size Report
Oct 22, 2015 - Department of Informatics IFI, Communication Systems Group CSG, ... Grupal Singh, "Performance Evaluation
ETSI workshop, Vienna, Austria, October 22, 2015

Decompiling QoE Christos Tsiaras, Burkhard Stiller Department of Informatics IFI, Communication Systems Group CSG, University of Zürich UZH [tsiaras,stiller]@ifi.uzh.ch Background QoE Models Proposed Solution Example

© 2015 UZH, CSG@IFI

QoE Fundamentals 

QoE is affected by multiple variables – Each variable has different importance



Each variable affects differently each user’s QoE – Each variable also affects the user differently in each service



The QoE concept is a mess! – Because is a user&service-centric concept



Mapping human brain into math is challenging – And fun  QoE  f (user, service, X  {QoS1, ,QoSk , ,QoSN })

© 2015 UZH, CSG@IFI

QoE in IPTV Services (Bandwidth, Price)

T. Hayashi, A. Takahashi, Nippon Telegraph and Telephone Corporation (NTT), Japan, "QoE Assessment Method for Video Quality and Pricing in IPTV Services", European Telecommunications Standards Institute (ETSI) Workshop, 17-19 June 2008, Prague, Czech Republic. © 2015 UZH, CSG@IFI

QoE in VoIP (Latency)

O3b Networks, Sofrecom, “Why Latency Matters to Mobile Backhaul” © 2015 UZH, CSG@IFI

QoE in VoIP (Hops in WMNs)

A. Chhabra, Grupal Singh, "Performance Evaluation and Delay Modeling of VoIP Traffic over 802.11 Wireless Mesh Network", International Journal of Computer Applications, Vol. 21, No. 9, pp. 0975 – 8887, May 2011. © 2015 UZH, CSG@IFI

2/n-D

IP packet Loss Ratio

QoV: QoE in Video (IPTD & IPLR)

IP packet Transfer Delay

S. Aroussi, T. Bouabana-Tebibel, A. Mellouk, "Empirical QoE/QoS correlation model based on multiple parameters for VoD flows," Global Communications Conference (GLOBECOM), 2012 IEEE, pp.1963-1968, 3-7 Dec. 2012 © 2015 UZH, CSG@IFI

QoE Models Market - IQX Hypothesis IQX : QoE    e  QoS   

1 degree of freedom – β: curve gradient





α and γ define the min and max MOS Describes antitonic MOS only

© 2015 UZH, CSG@IFI

QoE Models Market – QoV QoV : QoE    e 0 1QoS12 QoS2  

One degree of freedom per variable – β1, β2, ...

 

Describes multiple but only antitonic MOS β0 has to be defined based on the number of variables that are involved – The generic QoE equation do not reproduce the same QoE equation if only one QoS parameter is examined • Example that makes physics beautiful (Lorentz factor disappears) m

m0 1

© 2015 UZH, CSG@IFI

u2 c2

u  c 2 u  0  m  m0  c2

Proposed Solution An Axiomatic QoE Model (AQX) 

Formalizing QoE in steps 1. Identify the variables that affect QoE 2. Characterize those variables • Isotonic Variables (IVs) - The more you have the better it is • Antitonic Variables (AVs) - The more you have the worst it is

3. Select the ideal/desired/expected/agreed value of a variable 4. Considering the service specifications select the best and the worst values of the variable 5. Identify the effect of each variable’s variation • Influence factors

6. Identify the importance of each variable

© 2015 UZH, CSG@IFI

Expected value Step 3 Best and worst values Step 4 QoE equation for AVs

Influence factor Step 5

AQX

ea (x)  4e

 x m   ln 4 3  x0 

Variables characterization Step 2

 x m   ln 4  x0 

1

QoE equation for IVs

ei (x)  4(1 e

Generic QoE equation

 e ia  xk  1    E(X)  1 4  4   k1  N

Variables selection Step 1 © 2015 UZH, CSG@IFI

QoE

) 1 wk

QoE-related Importance factor Step 6 variables values

(Parenthesis) IQX : QoE    e QoS   μ: the minimum score

 

 h h: width of the given options

AQX : QoEa  h  e

( a QoS m )

h  M   a  QoS © 2015 UZH, CSG@IFI

m 0





 h   ℓn    e0   





λ: a reference point (QoS,QoE)

Is QoE Served?

© 2015 UZH, CSG@IFI

Open Questions 

Is it possible to define importance factors (w) without: – Surveys? – Curve fitting?

w © 2015 UZH, CSG@IFI

Example – Steps 1 and 2 



Scenario: Internet plans of an ISP for home customers in some places in Switzerland Step 1: Variables identification – Uplink bandwidth – Downlink bandwidth – Price



Step 2: Variables characterization – IVs • Uplink bandwidth • Downlink bandwidth

– AVs • Price © 2015 UZH, CSG@IFI

Example – Step 3 

Step 3: Select the ideal/desired/expected/agreed value of a variable – Assume a customer selected the “Internet 50” option – Ideal values based on the SLA • Uplink bandwidth: 5 Mbit/s • Downlink bandwidth: 50 Mbit/s • Price: 59 CHF/month

© 2015 UZH, CSG@IFI

Example – Step 4 

Step 4: Select the best and worst values per variable – Best values • Uplink bandwidth: 15 Mbit/s • Downlink bandwidth: 250 Mbit/s • Price: 0 CHF/month

– Worst values • Uplink bandwidth: 0.2 Mbit/s • Downlink bandwidth: 2 Mbit/s • Price: 89 CHF/month

© 2015 UZH, CSG@IFI

Example – Steps 5 and 6 

Step 5: Identify the effect of each variable’s variation – When a customer is starting to get annoyed/getting pleased? • Estimate/Assume/Extract this information from the Customer Care department statistics about report of problems – E.g., 50% less than expected bandwidth dissatisfies a customer – E.g., 25% discount would satisfy a dissatisfied customer



Step 6: Identify the importance of each variable – How a customer selects a plan in this scenario? • Estimate/Assume/Extract through a survey: – 50% based on the price – 30% based on the downlink bandwidth – 20% based on the uplink bandwidth

© 2015 UZH, CSG@IFI

Q&A AQX Model paper LCN 14

AQX in VoIP paper IFIP Networking 15

http://tiny.uzh.ch/da

http://tiny.uzh.ch/db

© 2015 UZH, CSG@IFI

Back-up Slide

IQX

© 2015 UZH, CSG@IFI

DQX

QoE Models Market – DQX (1) AQX : QoEa  h  e

( a QoS m )

h  M   a  QoS 

m 0



 h   ℓn    e0   

2 degrees of freedom – m: curve gradient – λa, QoS0: reference point



h and μ define the min and max MOS

© 2015 UZH, CSG@IFI

QoE Models Market – DQX (2) AQX : QoEi  h  (1 e

 i x m

h  M   i  QoS

m 0

 

)

  h  ℓn    h  e0   

Describes also increasing MOS 2 degrees of freedom – m: curve gradient – λi, QoS0: reference point

© 2015 UZH, CSG@IFI

QoE Models Market – DQX (3) X  {QoS1, ,QoSk , ,QoSN }  eia (QoSk )    E(X)    h      h k1 N



wk

Describes multiple mixed variables – Increasing/decreasing MOS



The multiple variables has an additional degree of freedom – wk: QoS importance factor

© 2015 UZH, CSG@IFI

QoE Models Market – DQX (4) Let :

N  1  w1  1

 eia (QoSk )     eia (QoS1 )        h    eia (QoS1 ) E(X)    h          h h k1 N



wk

1

The generic equation boils down to the specific equation – The beauty of a generic model • Lorentz would be happy 

© 2015 UZH, CSG@IFI