Cognitive Networks: Autonomous Customer Experience

Cognitive Networks:
Autonomous Customer Experience and
Resource Management
Péter Szilágyi, Nokia
BUTE 5G Technologies
25th April, 2017
1
25/04/2017 © Nokia 2016
Today’s challenge: massive growth of mobile Internet traffic
Main characteristics
€
$
f
Diversity and massive volume
From VoIP through social networking to high resolution multimedia.
Uniqueness and personalization
Context, location, time, device, OS, load, user specific content.
Always on (never sleeps)
Continuous demand and dynamic load.
Photos shared by
Facebook Messenger
Service Quality sensitiveness
Poor service deteriorates business. Users don’t accept bad quality.
216,302
Encryption
– Data Never Sleeps 4.0,
https://www.domo.com/blog/2015/08/data-never-sleeps-4-0/
1
2
25/04/2017 © Nokia 2016
The amount of encrypted content has an increasing trend.
The importance of customer experience (Quality of Experience, QoE)
How do people experience latency? 1
☺
0.1 second
Feeling of immediate
response

1 second
10 seconds
3
75%
OF AUDIENCE
ENDURES
17%
4%
OF AUDIENCE
ENDURES
OF AUDIENCE
ENDURES
5-9
10-29
0-4
MINS
MINS
MINS
4%
OF AUDIENCE
ENDURES
30+
MINS
How patient are people with poor video streaming experiences? 2
Feeling at the mercy
of the system
25/04/2017 © Nokia 2016
Amazon: every 100 ms latency:
 -1% in sales. 3
Bing. 2 second slowdown:
 queries per user -1.8%,
 revenue per user -4.3%. 4
Shopzilla: performance hike
(from ~7 to ~2 seconds):
 +25% in page views,
 +7-12% in revenue,
 -50% hardware need. 4
No. 1 CHOICE
Sensing delay but
still in control

How does user experience impact
business?
1–
Usability Engineering, Jakob Nielsen, 1993
– How Consumers Judge Their Viewing
Experience, Conviva, 2015
2
THE VIDEO SERVICE
INTERNET SERVICE
CONTENT DELIVERY
Who do users blame after poor video streaming experiences? 2
3
– G. Linden, Amazon, Make Data Useful;
4
– O’Reilly Velocity 2009 Conference
QoE management: needed already for today’s Internet applications
End-user
lifestyle
Application
diversity
Available
resources
Revenue
impact
• ubiquitous Internet
access via mobile
devices
• highly dynamic and
versatile traffic profiles
(per session)
• over-provisioning is not
economical or feasible
• good QoE is revenue
generator
• expectation: seamless
Quality of Experience
• resource requirements
for good QoE span a
wide scale
• congestion may occur
on the radio interface or
at the mobile backhaul
• subscribers mandate
high quality services and
react to degradations
Requirements for network side QoE management:
real time
4
context based
25/04/2017 © Nokia 2016
target individual application
sessions (granularity)
actions are based on the resource
requirement of each application session
Cognitive stress by QoE degradation
Acute stress
47% of smartphone
users loose focus
Cognitive load/stress
Dorsolateral
prefrontal cortex
EEG technology
to record brain
activity
Orbitofrontal
cortex
70% loose focus
30% rise in stress
even at 1 second delay
0
1
2
3
peak in
stress
4
5
6
81% loose focus
critical delay: all
users lose interest
7
8
9
10
Streaming video delay [seconds]
5
+44%
30%
72%
increased heart rate
by a 2 second delay
in selfie upload
frustrated by 1st occasion
of delay while uploading
pictures to Facebook
millennial video users
lose interest after 4
seconds of delay
25/04/2017 © Nokia 2016
EXIT
Source: Content in the blink of an eye, Ericsson & Vodafone, 2017
+10%
!
increased stress in
age group 18-24
compared to 35+
Future service verticals and applications
AUGMENTED
Augmented
gaming
INTERCONNECTED
Personal
robot
8k Video
beamer
Real time
work in cloud
REDEDICATED
Work & game
while traveling
Assisted driving
Augmented
dashboard
4k Video
Smart
watch
VR gaming
People & Things
6
Remote
Diagnosis
Real-time
remote control
3D
3Dprinting
printing
Advanced
monitoring
Touch & steer
TACTILE
Industry 4.0
Traffic Mgmt.
Smart grids
Safety & Security
HD Cams NW
Real-time
remote control
REVOLUTIONIZED
Traffic steering &
management
Connected
home
Health
25/04/2017 © Nokia 2016
Tracking / inventory
systems
Factory
automation
Toll collection
Real time
cloud access
VIRTUAL
Maintenance
optimization
Logistics
Travel &
commute
Virtual 3D
presence
Automotive
Smart
clothes
Mobile living
Communication
Time shift
Augmented
shopping
AUTONOMOUS
Self driving
Waste mgmt.
Utility & Energy
Reliable emergency
communications
SUPEREFFICIENT
Future service verticals and applications – disruption from Internet traffic
Communication patterns may be fundamentally different from currently known applications
Vehicular
Tactile/AR/VR
M2M, IoT
• car to car / roadside
communication
• remote object
manipulation
• massive number of
distributed devices
• self-driving cars
• remote surgery
• smart metering
• platoon steering
• robotics/control
• security/surveillance
• crossroad coordination
• virtual/augmented
reality
• sensor networks
Future
?
New requirements include: availability, reliability, multi-/local-connectivity, ultra-low latency, security, etc.
7
25/04/2017 © Nokia 2016
Challenges and requirements
Challenges to overcome
Future proof system
Do not consider today’s applications only –
adaptation to new use cases should not require a
major redesign or a new telco generation
Dynamic and adaptive
Do not trust QoE and service quality to always
having plenty of bandwidth – resources should be
managed and used on demand, even under limitations
Programmable
Reduce system defaults – pre-defined configurations
and static actions lack the flexibility needed for
optimal operation
8
25/04/2017 © Nokia 2016
Real Time
Self-learning
Embed cognitive functions in the system that are capable of
autonomous self-measuring/monitoring, self-learning and selfprogramming (parameterization, configuration, optimization).
Cognitive network concept
Autonomous Network & Service
Optimization/Management
Dynamic & Adaptive
Customer
Experience Mgmt
Autonomous
Network & Service
Optimization/Mgmt
Dynamically adapt the network to the
actual or predicted demand/context.
Dynamic & Adaptive Customer
Experience Management
Use the available resources better to
maximize the customer experience
and system efficiency in real time.
Physical/virtual infrastructure and
resources
9
25/04/2017 © Nokia 2016
Optimal network operation from both
end-to-end performance and
customer experience point of view
Motivation for autonomous cognitive functions also at the RAN
Application detection
and packet marking
CORE  it is YouTube...
Content
servers
UE
CORE
RAN
...but how to serve it? still need to
infer the specific requirements of
this particular video session
RAN – Radio Access Network
UE – User Equipment
10
25/04/2017 © Nokia 2016
RAN
the traffic may be routed locally to
keep it physically close to the
communicating devices – but it may
need dynamic and adaptive service
specific to the use case, content,
application / vertical
Autonomous cognitive functions
Content
servers
UE
CORE
RAN
Cognitive resource
(radio, physical/virtual)
and QoE management
V2V
V2V
RAN
Cognitive resource (DC, transport,
physical/virtual) and QoE management
Embedded cognitive decision
and action functions
Self-configuration, -optimization,
-learning, -monitoring
RAN – Radio Access Network
UE – User Equipment
11
25/04/2017 © Nokia 2016
Disruption from GPRS/EPC paradigm: CFs are deployed
through the e2e system (not limited to the core).