Niqa competitive alternative for non-intrusive voice quality testing (p.563)
Quality of Experience characterization and provisioning in ... semes… · + In-service,...
Transcript of Quality of Experience characterization and provisioning in ... semes… · + In-service,...
![Page 1: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/1.jpg)
Mobile and wireless networks
Quality of Experience characterization and provisioning in
wireless and mobile networks
Dr. Eirini Liotou14/12/2018
![Page 2: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/2.jpg)
Mobile and wireless networks
Subjective: Crowdsourcing
* T. Hossfeld, C. Keimel, M. Hirth, B. Gardlo, J. Habigt, K. Diepold, and P. Tran-Gia, “Best Practices for QoE Crowdtesting: QoE Assessment With Crowdsourcing,” IEEE Trans. Multimed., vol. 16, no. 2, pp. 541–558, Feb. 2014.
Q: Have you ever used such
a tool?
![Page 3: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/3.jpg)
Mobile and wireless networks
![Page 4: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/4.jpg)
Mobile and wireless networks
Crowdsourcing example 1 (BSc thesis)
• http://gain.di.uoa.gr/kyr/s2b/qoe.html
• http://gain.di.uoa.gr/kyr/s1c/qoe.htmlHands on
![Page 5: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/5.jpg)
Mobile and wireless networks
Crowdsourcing example 1
X
X
![Page 6: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/6.jpg)
Mobile and wireless networks
Crowdsourcing example 1
X
X
![Page 7: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/7.jpg)
Mobile and wireless networks
Crowdsourcing example 1
Video ID
Inst
ance
s o
f h
igh
er
pe
rcei
ved
qu
alit
y
1 stalling of 6s vs 2 stallings of 3s each
![Page 8: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/8.jpg)
Mobile and wireless networks
Crowdsourcing example 2
• http://gain.di.uoa.gr/kyr/357/qoe.html
• http://gain.di.uoa.gr/kyr/753/qoe.html
Descending vs ascending quality
Video ID
Inst
ance
s o
f h
igh
er
pe
rcei
ved
qu
alit
y
![Page 9: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/9.jpg)
Mobile and wireless networks
QoE monitoring example (MSc thesis)
![Page 10: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/10.jpg)
Mobile and wireless networks
QoE monitoring example
• Initial delay
• Number and duration of stalling events
• Total video duration
• Actual user viewing time
• Video size in bytes
• GPS info
• Internet connection type
![Page 11: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/11.jpg)
Mobile and wireless networks
Electroencephalography (EEG)
• Human influence factors (HIFs) characterize the user’s perception, emotional and mental state with respect to a service
– Facial expressions, body posture, voice, eye measurement, electrocardiography (ECG), electrodermal activity (EDA)
– EEG measures electrical activity in the brain
* R. Gupta, K. Laghari, H. Banville, T. H. Falk, “Using affective brain-computer interfaces to characterize human influential factors for speech quality-of-experience perception modelling”, Human-centric Computing and Information Sciences, 2016.
* http://myscienceschool.org/index.php?/archives/3208-What-is-Electroencephalography-EEG.html
![Page 12: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/12.jpg)
Mobile and wireless networks
Objective: evaluation methods
• Media-layer
• Packet-layer / Bitstream
• Parametric planning
NetworkMedia-layer
model
Input
Output
Reference
QoE
estimate
QoE
estimateNetwork
Parametric model
Input
Output
QoE
estimate
Input
NetworkPacket-layer
model
Input
Output
QoE
estimate
Q: Why is this difficult/impossible to implement in a real-time network?
![Page 13: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/13.jpg)
Mobile and wireless networks
Objective: evaluation methodsModel Advantages Disadvantages
Media-layer:
Full Reference
(e.g. PESQ)
+ Do not require any a-priori knowledge or assumptions about
the underlying network
+ Highly accurate and robust (based on psycho-acoustics)
- Require the reference signal (intrusive)
- Very high computational effort
- Practically impossible to implement at network midpoint
- Do not enable insight into the internal system functionality &
degradation causes (black-box) => diagnosis not possible
- Neglect human dimensions, pure technical
Parametric planning:
E-model
+ Ease of use and respect of privacy
+ The network is characterized by the technical specifications
of its constituent elements, (non-intrusive approach )
+ Quantifies the human factor through the “Advantage
factor”, & contextual factor
+ Mouth-to-ear complete transmission chain => conversational
+ No restrictions on the network with respect to size,
configuration, hierarchy, technology used, nor on the
components of the network
- Intended only for the planning phase of a system (extended format)
- Good in theory, but difficult to include all the model parameters
online
- Accurate only under strict application scenarios: new subjective
tests and regression analysis needed for different conditions
- Speech independent
- A-priori information requirement
Packet-layer:
ITU-T P.564
+ Enables insight into the internal system functionality (glass-
box)
+ Light in terms of computational effort
+ Multiple monitoring points help identify the root of a
network impairment
+ Used not only for speech quality predictions but also for the
production of diagnostic outputs
+ In-service, non-intrusive (privacy)
+ Quality followed and pooled over time
- Not standardized, models need to be created that comply with
these recommendations
- The model doesn’t know the characteristics of speech content to
evaluate (speech level, echo, background noise etc.): assumes a
generic voice payload
- Only concerns impairments on the IP network (no end-to-end
evaluation)
- Large volume of QoE data
- Models deployed require strict conformance testing
![Page 14: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/14.jpg)
Mobile and wireless networks
Examples of objective models
➢ VoIP: = 𝟗𝟒. 𝟐 − 𝟎. 𝟎𝟐𝟒𝒅 + 𝟎. 𝟏𝟏 𝒅 − 𝟏𝟕𝟕. 𝟑 𝑯 𝒅 − 𝟏𝟕𝟕. 𝟑 −[𝟏𝟏 + 𝟒𝟎 𝒍𝒏 𝟏 + 𝟏𝟎𝒑 ]
➢ YouTube (TCP): 𝑸𝒐𝑬 = 𝟑. 𝟓 ∗ 𝒆− 𝟎.𝟏𝟓𝑳+ 𝟎.𝟏𝟗 ∗𝜨 + 𝟏. 𝟓
➢ HTTP Adaptive Streaming (TCP): 𝑸𝒐𝑬 = 𝟎. 𝟎𝟎𝟑 ∗ 𝒆𝟎.𝟎𝟔𝟒∗𝒕 + 𝟐. 𝟒𝟗𝟖
➢ Real-time video (UDP): 𝑽𝒒 = 𝟏 + 𝑰𝒄𝒐𝒅𝒊𝒏𝒈 ∗ 𝑰𝒕𝒓𝒂𝒏𝒔𝒎𝒊𝒔𝒔𝒊𝒐𝒏
➢ FTP: QoE = 𝜶 𝒍𝒐𝒈𝟏𝟎(𝜷𝑹) , 𝟏𝟎𝒌𝒃𝒑𝒔 < 𝑹 < 𝟑𝟎𝟎𝒌𝒃𝒑𝒔
packet loss ratedelay
#of stallsduration of stalls
time on highest quality level
data rate
FR, BR, PLR
![Page 15: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/15.jpg)
Mobile and wireless networks
A. ITU-T G.107 “E-model” for voice
• Computes the transmission quality of VoIP by estimating the mouth-to-ear conversational quality as perceived by the receiver
• A parametric model that produces the so-called Rating factor 𝑅:
𝑹 = 𝑹𝟎 − 𝑰𝒔 − 𝑰𝒅 − 𝑰𝒆−𝒆𝒇𝒇 + 𝑨
➢ Ro→ basic signal-to-noise ratio, Ro = 100
➢ Is→ impairments due to the voice signal travelling in the network
➢ Id→ impairments caused by delay from end-to-end travelling signal
➢ Ie-eff→ equipment impairment factor & impairments due to packet loss
➢ A → advantage/expectation factor, in exchange for some user benefits or other factors difficult to quantify
![Page 16: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/16.jpg)
Mobile and wireless networks
𝑅 = 𝑅0 − 𝐼𝑠 − 𝐼𝑑 − 𝐼𝑒−𝑒𝑓𝑓 + 𝐴
𝑅 = 94.2 − 0.024𝑑 − 0.11 𝑑 − 177.3 𝐻 𝑑 − 177.3 − 11 − 40 𝑙𝑛 1 + 10𝑝
E-model: simplified version
0 50 100 150 200 250 300 350 4000
20
40
One-way delay (msec)
Id
0 50 100 150 200 250 300 350 4003
4
5
QoE
(M
OS
)
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.160
20
40
60
Packet loss rate
Ie-e
ff
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.162
3
4
5
QoE
(M
OS
)
packet loss ratedelay
![Page 17: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/17.jpg)
Mobile and wireless networks
B. ITU-T G.1070 “E-model” for video
• A computational model for point-to-point interactive videophone applications over IP networks (UDP-based - lossy video)
• Network, Application & Terminal parameters incorporated
• Video quality =
𝑽𝒒 = 𝟏 + 𝑰𝒄𝒐𝒅𝒊𝒏𝒈 ∗ 𝑰𝒕𝒓𝒂𝒏𝒔𝒎𝒊𝒔𝒔𝒊𝒐𝒏
➢ 𝐼𝑐𝑜𝑑𝑖𝑛𝑔 = the video quality affected by the coding distortion
➢ 𝐼𝑡𝑟𝑎𝑛𝑠𝑚𝑖𝑠𝑠𝑖𝑜𝑛 = the video quality affected by the transmission process
➢ Ultimately everything is a function of: • the video frame rate (fps) - FR
• the video bit rate (kbps) - BR
• the video packet loss rate - PLR
• 12 coefficients
![Page 18: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/18.jpg)
Mobile and wireless networks
C. QoE for YouTube
• Video on Demand (VoD), TCP-based connection (no losses)
• Quality influence factors (by crowdsourcing & lab tests):
➢ Number of stalling events, N
➢ Duration of stalling events, L
➢ Total video duration, T (total stalling duration over video duration)
➢ Initial delay (video start-up delay) → cache redirections’ impact
Q: Why does a stalling happen?
![Page 19: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/19.jpg)
Mobile and wireless networks
QoE for YouTube
• Some conclusions:
➢ The user demographics have no significant influence (!)
➢ Initial delays have almost no influence on MOS for videos of duration 60s and 30s compared to the influence of stalling length
➢ The user ratings are statistically independent from video motion, type of content, the usage pattern of the user, access speed, etc.
➢ The number of stalling events together with the stalling length are clearly dominating the user perceived quality
➢ The video duration only plays a role if there are only a very few stalling events
Q: Is this your impression too?
![Page 20: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/20.jpg)
Mobile and wireless networks
QoE for YouTube
IQX hypothesis validation:
𝑸𝒐𝑬 𝑳, 𝑵 = 𝜶 ∗ 𝒆−𝜷 𝑳 ∗𝜨 + 𝜸,
α = 3.5, β(L) = 0.15L+ 0.19, γ = 1.5
* T. Hossfeld, R. Schatz, E. W. Biersack, and L. Plissonneau, “Internet Video Delivery in YouTube: From Traffic Measurements to Quality of Experience,” in Data Traffic Monitoring and Analysis, Eds. Springer Berlin Heidelberg, pp. 264–301, 2013.
![Page 21: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/21.jpg)
Mobile and wireless networks
D. HTTP Adaptive Streaming (HAS)
Comparison of HTTP video streaming and HTTP adaptive video streaming
Q: Why is this better?
![Page 22: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/22.jpg)
Mobile and wireless networks
HTTP Adaptive Streaming (HAS)
Based on how fast the current (and previous)
segments are downloaded, the bit rate
of the next segment is selected
* M. Seufert, S. Egger, M.Slanina, T. Zinner, T. Hoßfeld, and P. Tran-Gia, “Survey on Quality of Experience of HTTP Adaptive Streaming”, IEEE Communication Surveys & Tutorials, Vol. 17, No. 1, First Quarter 2015.
![Page 23: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/23.jpg)
Mobile and wireless networks
HTTP Adaptive Streaming (HAS)
• Other influence factors: Adaptation frequency (number of switches), adaptation amplitude, adaptation direction, segment length, buffer size, etc.
𝑸𝒐𝑬= 𝟎. 𝟎𝟎𝟑 ∗ 𝒆𝟎.𝟎𝟔𝟒∗𝒕 + 𝟐. 𝟒𝟗𝟖
𝒕 = time on highest layer
![Page 24: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/24.jpg)
Mobile and wireless networks
E. QoE for file download services
• Elastic service, for which the utility function is an increasing, strictly concave, and continuously differentiable function of throughput
• The user satisfaction of a file transfer service is solely dependent on the provided data rate
• Logarithmic relationship between MOS and throughput:
𝑴𝑶𝑺 = ቐ
𝟏, 𝑹 < 𝟏𝟎𝒌𝒃𝒑𝒔𝜶 𝒍𝒐𝒈𝟏𝟎(𝜷𝑹) , 𝟏𝟎𝒌𝒃𝒑𝒔 < 𝑹 < 𝟑𝟎𝟎𝒌𝒃𝒑𝒔
𝟒. 𝟓, 𝟑𝟎𝟎𝒌𝒃𝒑𝒔 < 𝑹
➢ R is the data rate of the service ➢ α and β obtained from the upper and lower
user perceived quality expectations
* S. Thakolsri, S. Khan, E. Steinbach, and W. Kellerer, “QoE-Driven Cross-Layer Optimization for High Speed Downlink Packet Access,” J. Commun., vol. 4, no. 9, 2009
![Page 25: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/25.jpg)
Mobile and wireless networks
F. Netflix – challenges
• Understanding the impact of QoE on user behavior (compression artifacts, scaling artifacts, rebuffering rate, bitrate, etc.)
• Creating a personalized streaming experience
• Determining what movies and shows to cache on the edge servers based on member viewing behavior
• Improving the technical quality of the content using viewing data and member feedback
* https://medium.com/netflix-techblog/optimizing-the-netflix-streaming-experience-with-data-science-725f04c3e834
Q: How could Netflix infer that something is wrong?
![Page 26: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/26.jpg)
Mobile and wireless networks
Other QoE metrics
PSNR=31dB
PSNR=34dB
DMOS=82
DMOS=96
DMOS=27
DMOS=58
DMOS is 100 for the reference video
Q: Does this or… Q: this look better?
![Page 27: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/27.jpg)
Mobile and wireless networks
Netflix Video Multimethod Assessment Fusion (VMAF) quality metric
* https://medium.com/netflix-techblog/toward-a-practical-perceptual-video-quality-metric-653f208b9652
Q: Which one is better and why?
![Page 28: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/28.jpg)
Mobile and wireless networks
SOS – The MOS is not enough
• Standard deviation of Opinion Scores (SOS)
• Reflects the level of rating diversity
• A square function of MOS → SOS hypothesis
• No diversity at the edges and maximal diversity at MOS = 3
* T. Hossfeld, R. Schatz, and S. Egger, “SOS: The MOS is not enough!,” in 2011 Third International Workshop on Quality of Multimedia Exper ience, 2011, pp. 131–136.
![Page 29: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/29.jpg)
Mobile and wireless networks
QOE MANAGEMENT
![Page 30: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/30.jpg)
Mobile and wireless networks
QoE management framework
QoEMANAGER
(1) Instructions controlling the QoE-input data generation are sent to the network(2) Input data from all data sources are collected by the QoE-Controller(3a) Processed QoE-data per flow are sent to the QoE-Monitor(3b) Information regarding the current network state is sent to the QoE-Manager(4) Estimated QoE scores are reported to the QoE-Manager per flow (5a) Customer Experience Management procedures are performed(5b) Corrective actions are triggered, if required(6) The QoE-Controller actualizes these corrective actions
QoEMONITOR
(4)
(3a) (3b)
(5b)
NETWORK
QoECONTROLLER
Periodical
Conditional
(5a)
Network specific control & optimization
Network specific collection & feedback
mechanisms
Appropriate QoE estimation models
OBJECTIVE:
Enable a QoE-centric network management framework to:
1. Monitor the end-users’ QoE
2. Enhance their experience
3. Improve the network’s efficiency (spectrum, energy)
* E. Liotou, D. Tsolkas, N. Passas and L. Merakos, “Quality of Experience management in mobile cellular networks: Key issues and design challenges,” IEEE Communications Magazine, Network & Service Management Series, July 2015.
![Page 31: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/31.jpg)
Mobile and wireless networks
QoE research stages & management
Goal: Optimize end-user QoE, while making efficient use of network resources & maintaining a satisfied customer base
Function 1
QoE Controller(data collection)
Function 2
QoE Monitor (modeling)
Function 3
QoE Manager(control)
DIAGNOSIS
MONITORING
ACTIONS
What to collect?
From where?
How?When?
Which model?
How to react?
Where?Where?How
often?
How to transfer?
How to diagnose?
How to deliver?
![Page 32: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/32.jpg)
Mobile and wireless networks
Realization issues
• Selection of the physical/virtual location of the QoE managementframework inside the operator’s infrastructure
• Identification of required QoE data sources, configuration of datacollection periodicity, signaling between network and QoE-Controller
• Selection of appropriate QoE models and KPIs for the QoE-Monitor
• Traffic/service classification performed in the QoE-Monitor,especially in the content-encrypted domain
• Network-specific type of decisions taken by the QoE-Manager andtheir actualization through the QoE-Controller
![Page 33: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/33.jpg)
Mobile and wireless networks
Driving network decisions – D2D
D2D1
D2D2
QoED2D
D2D data Cellular data
QoEcell
➢ QoE awareness can control the operational mode of users in LTE-A:
• Drive cellular Device-to-Device (D2D) mode transitions
➢ Enhance QoE, ↑ throughput, offload network, ↓ power, allow for profits
Q: Why is D2D beneficiary?
![Page 34: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/34.jpg)
Mobile and wireless networks
Driving network decisions – D2D
* E. Liotou, E. Papadomichelakis, N. Passas and L. Merakos, “Quality of Experience-centric management in LTE-A mobile networks: The Device-to-Device communication paradigm”, in 6th International Workshop on Quality of Multimedia Experience (IEEE QoMEX), Singapore, September 2014.
![Page 35: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/35.jpg)
Mobile and wireless networks
Improving QoE – Admission control
1
1.5
2
2.5
3
3.5
4
4.5
Progress of time
Inst
anta
neo
us
syst
em Q
oE
(M
OS)
Real-time operation of the QoE management framework
With QoE-management framework
Without QoE-management framework
AREA OF NON-ACCEPTABLE QoE
AREA OF ACCEPTABLE QoE
QoE Manager triggerscorrective actions in
the NetworkMonitored system QoE
less than theacceptable QoEthreshold (=3.5)
![Page 36: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/36.jpg)
Mobile and wireless networks
Improving QoE – Admission control
* E. Liotou, D. Tsolkas, N. Passas and L. Merakos, “Quality of Experience management in mobile cellular networks: Key issues and design challenges,” IEEE Communications Magazine, Network & Service Management Series, July 2015.
80 90 100 110 120 130 140 150 160 170 180 190 200 210 2201
1.5
2
2.5
3
3.5
4
4.5
Number of total concurrent VoIP flows in the system
QoE
: Me
an O
pin
ion
Sco
re (
MO
S)
c) QoE-driven admission control
Without QoE-management,UEs served by the small-cell
With QoE-management,UEs served by the small-cell
With QoE-management,UEs non-admitted by thesmall-cell, served by themacro-cell
QoE-basedadmission
triggerred by theQoE Manager
130 flows onwards
60flows
90flows
30 flows1flow
![Page 37: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/37.jpg)
Mobile and wireless networks
Saving resources through QoE awareness
➢ QoE awareness can drive a Power-Controlled Interference Managementscheme in femto-overlaid LTE-A
➢ How: Reduce HeNB’s transmit power, with no loss in femto-UEs’ QoE,provided that this is optimal
Optimum point of
operation
MUE
HeNBeNB
victimMUE
VictimFUE
MACROCELL
FEMTOCELL
Communication linkInterference link
IQX hypothesis
Q: Why is this the optimum operation point?
![Page 38: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/38.jpg)
Mobile and wireless networks
Saving resources through QoE awareness
20 40 60 80 100 120 140 160-6
-4
-2
0
2
4
6
8
10
12
HeNB-MUE distance (m)
HeN
B T
ransm
it P
ow
er
(dB
m)
3GPP PC
QoE-aware PC
* D. Tsolkas, E. Liotou, N. Passas, and L. Merakos, “The Need for QoE-driven Interference Management in Femtocell-Overlaid Cellular Networks ”, in 10th International Conference on Mobile and Ubiquitous Systems: Computing, Networking and Services (Mobiquitous), Tokyo, December 2013.
![Page 39: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/39.jpg)
Mobile and wireless networks
Extra: Another idea to save resources
QoE gain
5
Area 1Area 3
x1 x2
4
3
2
1
Area 2
ΔQ
oS1
ΔQ
oS2
ΔQoE1
ΔQoE2
q1 q2 q4q3
Qo
S g
ain
Qo
E (
MO
S)
QoS degradation
Preferred QoS reduction!
Q: Which ΔQoS reduction is preferred?
![Page 40: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/40.jpg)
Mobile and wireless networks
QOE CHALLENGES IN MOBILE NETWORKS
![Page 41: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/41.jpg)
Mobile and wireless networks
Key challenges in the QoE domain
Technical challenges
Economic challenges
Legal issues
QoE integration in communication networks
QoE needs to be managed on a per-user, per-application, and per-terminal basis in a real-time way
![Page 42: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/42.jpg)
Mobile and wireless networks
Technical challenges
• Monitoring: Network-centric vs. user agent-based approaches
➢ Agent-based:
+ Capture the HUMAN, CONTEXT, and WIRELESS medium aspects
- Do not offer diagnosis information
- Depend on manufacturer, not scalable
- Privacy, security, energy concerns
• Scalability & complexity issues
➢ QoE feedback, control and modeling per user session
• Network diversity
➢ Different operators or vendors, networks, mobile technologies (e.g., 2G or 3G), or even different countries or continents
• Energy consumption
➢ QoE-awareness and provisioning: monitoring, signaling, processing, memory requirements, new network entities
![Page 43: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/43.jpg)
Mobile and wireless networks
Heterogeneity (LTE-A example)
![Page 44: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/44.jpg)
Mobile and wireless networks
Economic challenges
QoE estimation may heavily depend on the expected price itself!
Charging for QoS vs. charging for QoE
𝑝𝑟𝑖𝑐𝑒 𝑓𝑢𝑛𝑐𝑡𝑖𝑜𝑛: 𝑝 = 𝑝 𝑞𝑑𝑒𝑚𝑎𝑛𝑑 𝑓𝑢𝑛𝑐𝑡𝑖𝑜𝑛: 𝑑 = 𝑑 𝑝𝑄𝑜𝑆 𝑓𝑢𝑛𝑐𝑡𝑖𝑜𝑛: 𝑞 = 𝑞 𝑑
𝑑𝑒𝑚𝑎𝑛𝑑 𝑓𝑢𝑛𝑐𝑡𝑖𝑜𝑛: 𝑑 = 𝑑 𝑝𝑄𝑜𝑆 𝑓𝑢𝑛𝑐𝑡𝑖𝑜𝑛: 𝑞 = 𝑞 𝑑𝑝𝑟𝑖𝑐𝑒 𝑓𝑢𝑛𝑐𝑡𝑖𝑜𝑛: 𝑝 = 𝑝 𝑥𝑄𝑜𝐸 𝑓𝑢𝑛𝑐𝑡𝑖𝑜𝑛: 𝑥 = 𝑥 𝑞, 𝑝
* P. Reichl, P. Maillé, P. Zwickl, and A. Sackl, “A Fixed-Point Model for QoE-based Charging,” in the ACM SIGCOMM 2013 Workshop on Future Human-Centric Multimedia Networking (FhMN2013), Hong Kong, 2013.
![Page 45: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/45.jpg)
Mobile and wireless networks
Legal challenges
• Network Neutrality
➢ “Quality” may be considered as a public good
➢ Differentiation/prioritization may be legally not allowed
• Double selling
➢ Sold as an add-on service to existing network connections?
➢ How profits will be distributed to involved parties?
• Service/Experience Level Agreements (SLAs/ELAs)
➢ Define the delivered quality in terms of QoE
➢ Find a “common vocabulary”
• Agreements among operators
➢ Collaborations, especially at interconnections
➢ Violations’ responsibility and handling
• Privacy and fidelity
➢ Transfer of user-sensitive information in an E2E path
![Page 46: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/46.jpg)
Mobile and wireless networks
MNO versus OTT
For mobile network operators (MNO) For Over-The-Top players (OTT)
• MNOs face a tremendous increase of data traffic
• Much of the traffic is originating from OTT Apps
• MNOs lose money: profits are unaffected while cost is higher
• Increased App types with large QoS diversity “compete” for the same pool of resources
• OTT Apps are served without QoS guarantees, over the default best-effort bearer
Then, why not go for a win-win paradigm of MNO-OTT “interfacing”, where:
• OTT Apps are served with better QoE →
(introducing some OTT control)
• QoS differentiation per App type
• App prioritization possible
• Resources shared more efficiently
• MNOs get into the revenue loop
![Page 47: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/47.jpg)
Mobile and wireless networks
SYNOPSIS
![Page 48: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/48.jpg)
Mobile and wireless networks
Revision
• The multidimensional definition of QoE
• The relationships between QoS and QoE
• QoE modeling evaluation methods
• QoE management required building blocks
• QoE exploitation possibilities
• Main challenges
![Page 49: Quality of Experience characterization and provisioning in ... semes… · + In-service, non-intrusive (privacy) + Quality followed and pooled over time - Not standardized, models](https://reader034.fdocuments.us/reader034/viewer/2022050202/5f5614b6a4b4075b0a0bde89/html5/thumbnails/49.jpg)
Mobile and wireless networks
Thank you!