The QoE Provisioning-Delivery- g y -- Hysteresis and its - - PDF document

NGI 2011

The QoE Provisioning-Delivery-

• g

y Hysteresis and its Importance for Service Provisioning in the F t I t t Future Internet

Tobias Hoßfeld, Markus Fiedler and Thomas Zinner

Hysteresis Hysteresis …

• Hysteresis refers to systems that may exhibit path

dependence, or "rate-independent memory".[1] In a d t i i ti t ith d i h t i deterministic system with no dynamics or hysteresis, it is possible to predict the system's output at an instant in time given only its input at that instant in instant in time given only its input at that instant in

• time. In a system with hysteresis, this is not

possible; the output depends in part on the internal possible; the output depends in part on the internal state of system and not only on its input. [www wikipedia com] [www.wikipedia.com]

and Quality of Experience … and Quality of Experience

[http://frombogotawithlove.com/wp-content/uploads/2009/10/Bored-Computer-User.jpg]

Hysteresis and Quality of Experience Hysteresis and Quality of Experience

PDH

• Input

QoE

• Goodput

p

• Output
• Internal state

p

• User satisfaction
• To distinguish:
• Internal state
• To distinguish:

– Provisioning Delivery problems – Delivery problems

Two branches of the hysteresis Two branches of the hysteresis

• Up = provisioning

– More packets per unit of time – More bandwidth and goodput – Typ. QoE = + log (QoS)

QoE

• Down = delivery issues

QoS

y

– Less packets per unit of time – Loss reduces goodput

QoS

Loss reduces goodput – Typ. QoE = + exp( QoS)

It’s a matter of signs It s a matter of signs …

Q E Q S

• Satisfaction rating

– The higher, the better Mean Opinion Score (1 5)

• Resource measure

– The higher, the better Throughput goodput

QoE

r

QoS

– Mean Opinion Score (1..5) – Throughput, goodput

• Success measure

– Availability (e g 99 99 %)

s

QoS

Not considered here:

• Failure rating

– Availability (e.g. 99.99 %) – Packet success ratio

QoE

• Failure rating

– The higher, the worse – Cancellation rate

• Failure measure

– The higher, the worse P k t l ti

QoE

f

QoS

Cancellation rate – Churn rate – Packet loss ratio – Delay jitter – Reordering Reordering

and a matter of being in control … and a matter of being in control.

typically on application layer typically due to network impairments

The PDH is seen for voice [14] The PDH is seen for voice [14] …

for live video streaming [21 23] … for live video streaming [21, 23] … and for web surfing [17] … and for web surfing [17]

What to do? What to do?

Simple rules for preserving reasonable QoE:

• 1. Avoid trouble – through management
• 1. Avoid trouble

through management

– Skype’s approaches

• Increase of consumption for voice [10, 12]

p [ , ]

• Decrease of consumption for video

– Communication between providers

• 2. Save resources in a controlled way –

through dimensioning through dimensioning

– Gain formulae

Skype’s increase of consumption for voice Skype s increase of consumption for voice

FEC: x2

Skype’s decrease of consumption for video Skype s decrease of consumption for video

Communication between providers (!) Communication between providers (!)

• Frameworks available, e.g. FP7 SmoothIT; IETF

ALTO; P4P; PaDIS; …

Potential for savings web example Potential for savings, web example Gain curves Gain curves

• 10

2 3 3

exp MOS G

• 19.9

ln(MOS) 18.3 G

Lessons to learn for future provisioning Lessons to learn for future provisioning

• Better reduce resource consumption

– Slightly less quality – Slightly higher response times – Eventually replace some fidelity by redundancy – Cooperation between application and network provider – Reduced energy consumption per user

than suffer from problems

– Avoid losses, freezes, …, churn Avoid losses, freezes, …, churn – Do not squeeze traffic in an uncontrolled manner

Conclusions & future work Conclusions & future work

• Design future services, applications, enablers
• etc. such that users do not suffer unnecessarily

– Systematic changes in provisioning keep QoE on an acceptable level – Accidental changes due to problems drag QoE down

Further work required w r t understanding

• Further work required w.r.t. understanding

fundamental relationships between

U l t d Q E t – User-related QoE parameters – Technical QoS parameters

Acknowledgements Acknowledgements

• To the QoEWeb team

– University of Würzburg y g – Orange Labs/France Télécom Warsaw University of Technology – Warsaw University of Technology

• QoEWeb was a Specific Joint Research

Project from the First Call (2008–2009) Project from the First Call (2008 2009)

Thank you very much! Thank you very much! Q & C Q

markus fiedler@bth se markus.fiedler@bth.se