Upload
phungthuy
View
217
Download
2
Embed Size (px)
Citation preview
Quality of Experience
for Mobile Communications
Challenges and Solutions
Zukunft der Netze
Netze und Plattformen für Dienste ohne Grenzen
8. Fachtagung des ITG-FA 5.2
Dr.-Ing. Dirk Kutscher
2
Mobile Communications: Current Trends
© NEC Corporation 2009
• Mobile data communications is taking off
– Reasonably priced data plans, flat rate tariffs
– HSPA coverage increasing, LTE on the horizon
3
Reasons (1)
© NEC Corporation 2009
4
Reasons (2)
© NEC Corporation 2009
5
Problems (1)
• Revenue does not keep up with traffic increase
© NEC Corporation 2009
6
Problems (3)
© NEC Corporation 2009
• Seamless connectivity, always-on does have a cost
– Largely proportional to the capacity provided – or worse
– Still no real QoS provided
• Problem with flat-rate pricing
– Attractive with customers (good thing)
– Once introduced, difficult to give up
– Invites unrestrained use
with only marginal revenues
• Possible solutions
– Limiting flat-rate data usage: volume caps,
“fair-use policies”
– Cutting cost on OPEX
• Simplifying network infrastructure
(flat architecture, evolved packet core)
• Self-deployment, self-management
• Leveraging local access at low cost (Femtocells etc.)
• Lowering QoS requirements
Source: KTH
7
IMS Architecture Overview
© NEC Corporation 2009
Managed Services:
recovering network costs through services
8
This talk
• Dealing with increased bandwidth demand and cost-
saving requirements
• Leveraging local heterogeneous access and
opportunistic communications
• Looking into sample scenarios and technical solutions
• Linking to ongoing Future Internet and standardization
activities
© NEC Corporation 2009
9
Local Heterogeneous Access
• Specific access technologies for specific
environments
– Leveraging local service clouds: WLAN hotspots,
Femtocells, mesh networks
• Lowering QoS requirements
– Say goodbye to seamless connectivity and strict QoS levels
– Leveraging opportunistic communications
– Accepting less than best effort service
– Providing acceptable quality of experience
© NEC Corporation 2009
10
Local Access with Femtocells
© NEC Corporation 2009
• Increasing coverage and capacity at reduced cost
– Re-using customers’ Internet access
– Direct IP access to the Internet (“local break-out”)
• Challenges
– Unpredictable Internet access link utilization and performance
– Femto deployment models such as CSG may interfere with macro network
• Conclusion
– Femtocells more like alternative access
(together with macro 3G/LTE, WLAN, WiMAX)
– No easy answer to quality of service
– Cannot take “seamless” for granted
– unless we are prepared to spend moneySource: Femto Forum
11
• CHIANTI: Beyond Seamless Connectivity
• Scalenet-TV2Go: Personal Multimedia Leveraging
Opportunistic Networking
• IETF LEDBAT: less than best effort service
• Trilogy: TCP friendliness revisited
© NEC Corporation 2009
12
CHIANTI
CHIANTI
Challenged Internet Access Network Technology Infrastructure
www.chianti-ict.org
CHIANTI
FP7-ICT-2007-1 CHIANTI 216714
Dr.-Ing. Dirk Kutscher 13
UMTS/GPRS on German ICE trains
CHIANTI
FP7-ICT-2007-1 CHIANTI 216714
Dr.-Ing. Dirk Kutscher
German Rail Long-Distance Rail Network
14
Complete Mobile Broadband coverage
economically challenging
Even if you could cover all tracks with
Mobile Broadband – this will not eliminate
all difficulties
CHIANTI
FP7-ICT-2007-1 CHIANTI 216714
Dr.-Ing. Dirk Kutscher
Beyond Seamless
Seamless connectivity…
Great goal, but
Impossible
High mobility
Incomplete coverage
Multiple reasons for
disruptions: technical, social,
legal
Expensive
Last 10% can double the cost
15
Vodafone network in Germany ( 3G )
CHIANTI
FP7-ICT-2007-1 CHIANTI 216714
Dr.-Ing. Dirk Kutscher
CHIANTI Approach
16
PCMP Operation under Different Conditions
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
10000
0 360 720 1080 1440 1800 2160
Time (s)
Data
rate
(kb
it/s
)
0
50000
100000
150000
200000
250000
300000
350000
Cu
mu
lati
ve D
ata
Vo
lum
e (
KB
)
Data rate (kbit/s)
Data volume (KB)
WLAN
access point
Make good use of connectivity where available:
Seamless service perception
Sourc
e:
Drive-t
ru I
nte
rnet pro
ject
CHIANTI
FP7-ICT-2007-1 CHIANTI 216714
Dr.-Ing. Dirk Kutscher
Objectives
17
IP
TCP
HTTP
WLAN
IP
Loopback
Web
Browser
Mobile Node Router
Intermittently Connected Link
HTTP+X
CHIANTI
Client
Tunnel Prot.
TCPIP
WLAN / …
IP
HTTP+X
CHIANTI
Proxy
Tunnel Prot.
TCP
IP
TCP
HTTP
Access Link
IP
TCP
HTTP
Web
Server
Wireless
Link
Internet Internet
Fixed Server Appl. Server
Improving the experience for mobile users
Hiding disruptions from the application layer
Developing a system architecture providing optimal disruption tolerance
prototype implementations of user devices and infrastructure components
service platforms and middleware components
CHIANTI
FP7-ICT-2007-1 CHIANTI 216714
Dr.-Ing. Dirk Kutscher
CHIANTI Summary
Dealing with intermittent connectivity
Providing specific support functions at the problematic link at
the edge of the network (here: proxy-based architecture)
Useful perceived quality of service without requiring
seamless connectivity
Question Moving form CHIANTI proxy-based architecture to end-to-end approach
What kind of applications can actually leverage intermittent connectivity?
18
19
Scalenet-TV2Go
• Scalenet-TV2Go
– Mobile TV in converged environments
– Alcatel-Lucent Bell Labs and TZI
• BMBF-sponsored (Scalenet project)
© NEC Corporation 2009
Mitnehm.TV
[Kutscher, Meyer, Schefczik, Soellner; ScaleNet-TV2Go: A Personalized Mobile
Multimedia Service; Wireless Communication and Information 2008 Proceedings]
20
Mobile-TV
21© 2008 TZI, Alcatel-Lucent
Time window for managed Mobile TV is closed
Multiple distribution networks, multiple applications Applications and user behavior becoming independent of specific distribution approach
(“convergence”)
“Long-tail” content becoming increasingly important, cannot broadcast everything
Users want to decide what, when and where to watch Public content (“TV”) and private media assets
Fixed (living room) and mobile usage
Personalized TV:access live TV and archived content from personal devices
22© 2008 TZI, Alcatel-Lucent
Converged Multimedia Services: IPTV and Mobile-TV
Use Case:“Session Mobility for a Convergent IPTV Service”
IPTV service over unified wireless or wireline IP network with convergent control (resource and mobility management)
Seamless session continuity and handover
Robust transmission for personalized mobile service
DataVoIP
HomeOn the move
Access-BorderGW
xDSLAM
IPTV
Office
CPE
MicroBSR
MicroBSR
BSR
Seamless Mobility and Session Continuity
VDSL
UMTSHSxPA
UMTSHSxPA
Application Server
(IMS/Web/TV)
Femto BSR Home
Gateway
Opticalfiber
VoIP Data
DataVoIP
HomeOn the move
Access-BorderGW
xDSLAM
IPTV
Office
CPE
MicroBSR
MicroBSR
BSR
Seamless Mobility and Session Continuity
VDSL
UMTSHSxPA
UMTSHSxPA
Application Server
(IMS/Web/TV)
Femto BSR Home
Gateway
Opticalfiber
VoIP Data
23© 2008 TZI, Alcatel-Lucent
Robust Video Transport
PCMP Operation under Different Conditions
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
10000
0 360 720 1080 1440 1800 2160
Time (s)
Data
rate
(kb
it/s
)
0
50000
100000
150000
200000
250000
300000
350000
Cu
mu
lati
ve D
ata
Vo
lum
e (
KB
)
Data rate (kbit/s)
Data volume (KB)
WLAN
access point
From Real-time-Streaming to Video-Chunk-Delivery
File-transfer paradigm for video-streaming
Reliable transport sessions (here: HTTP/TCP)
Opportunistic network usage and aggressive buffering
Use network resources as they become available
Try to buffer as much data ahead of time as possible
Suitable codecs and video file formats
Video file formats instead of RTP payload formats
Not all traditional formats are suitable for streaming
Here: Matroska universal audio/video
container format
24 ScaleNet-TV2Go, WCI 2008, Berlin, October 2008
Client Implementations
• Set-top box client
– Based on Freevo open source personal video system platform
– Hi-resolution content
– Web based GUI
• Mobile phone client
– J2ME-based
– MPEG4-Part2 @ 352x288, 15 fps, MP4 format
– Feature set limited by platform constraints
• PC platform mobile client
– MPEG4-Part2, Matroska
– Local pre-fetching proxy
RNC
GGSN
SGSN
NodeB
25
Scalenet-TV2Go Summary
• Converting traditional real-time-based
services to data-oriented approach
– Real-time requirements can be relaxed
for most content
• Context-based transmission:
fill buffers when it’s cheap
– E.g., In your local femto zone
• Background transmission of push content
– Question: how to differentiate background
traffic from other traffic?
© NEC Corporation 2008
26
IETF LEDBAT
• Less than best effort service
• Problem: usage of P2P technology, bulk data
transfer can have negative effects on interactive
applications and other TCP sessions
– TCP fills buffer at routers, delay for all applications
increases
– Home uplinks especially affected: asymmetric bandwidth,
most traffic is P2P
• LEDBAT: less than best effort service without
requiring help from routers
– Alternative congestion control schemes
– Saturate the bottleneck, but quickly yield to other TCP traffic
– Make using multiple TCP connections more friendly
– In general: reacting to queue growth earlier than standard
TCP would
© NEC Corporation 2008
27
LEDBAT solution space
• Application and transport layer
• Application layer:
– E.g., changing receive window size of standard TCP
• Transport layer
– Delay-based: interpreting observed delay as early congestion
indication (TCP Vegas like)
– Non-delay-based: apply weights to congestion control
• Behave like N TCP-friendly streams
• Adjust sending rate based on TFRC concept
– Problem: danger of starvation of low-priority streams
© NEC Corporation 2008
28
Trilogy Project
© NEC Corporation 2008
• General idea: designing for tussle
– Managing contention among parties with
conflicting interests
– Future Internet architecture should be
stakeholder-neutral, while supporting
tussle between parties
– Example: mobile operator vs. P2P
application/user
• In particular: resource sharing enabling
cost fairness
– Observation: traditional TCP fairness
concept has no resource usage/cost
model
– Account for resource usage when
network is congested [Bob Briscoe, BT]
29
there are better solutions than fighting
• light usage can go much faster
• hardly affecting completion times of heavy usage
NOTE: weighted sharing doesn't imply differentiated services
• can be weighted aggressiveness of end-point rate control
bit-rate
time
bit-rate
time
bit-rate
time
base case:TCP sharing
throttlingheavyusage
weightedsharing
30
core of solution
congestion-volume metric
• congestion-volume
• your volume weighted by link congestion
when each packet is served
• intuition
– some ISPs count volume only during peak
– like counting (100% x volume) during peak
and (0% x volume) otherwise
– congestion-volume counts p · xi over time
• measurement
– the amount of data discarded from your traffic
– or marked with explicit congestion notification
(ECN)
– end-point function in current architecture
1. cost to other users of your traffic
2. the marginal cost of upgrading equipment
• so it wouldn’t have been congested
• so traffic wouldn’t have affected others
• competitive market matches 1 & 2
metric for customers to judge ISPs,and ISPs to judge customers
congestion = too much traffic meets too little capacity
loss (marking)
fraction p(t) [%]
note: diagram is conceptual
congestion volume & capital cost of equipment would be accumulated over time
x1(t) [b/s]
x2(t) [b/s]
bit rate
most interesting when 'congestion' = marking, not loss
31
Diff
serv
ECN
R
E IPv
4
he
ad
er
Feedback path
Data packet flowSender Receiver
-1+1-1+1+1+1 Routers
Networks
1. Congested queue debit marks some packets
2. Receiver feeds back debit marks
3. Sender re-inserts feedback (re-feedback)
into the forward data flow as credit marks
4. Outcome:
End-points still do congestion control
But sender has to reveal congestion it will cause
Then networks can limit excessive congestion
5. Cheaters will be persistently in debt
So network can discard their packets
(In this diagram no-one is cheating)
1
23
54
re-ECN = standard ECN + re-inserted feedbackor re-feedback
• No changes required to data forwarding• Realisation of network control & economics research
stretching back to 1991 [Kelly05]
32
Re-Feedback Summary
• NGN/IMS: service layer on top of transport layer
– Charge for services to recover network operation cost
• Re-Feedback: adding cost accountability to IP
– ISP perspective
– Congestion-volume metric:
meter traffic that causes congestion/cost
– Cheap (best-effort) traffic will have
to yield to other traffic
– No fundamental changes
to Internet infrastructure required
• Quality of service perspective
– Detecting and accounting for congestion
– Limited resources: better service can be paid for
– Allowing for better user experience
by abandoning traditional TCP fairness concept
– Bandwidth can be added accordingly
© NEC Corporation 2009
Acceptable Use Policy
Your 'congestion volume' allowance:
1GB/month (= 3kb/s continuous)
Only limits excess traffic above the
Internet 'high-water-mark'
Under typical conditions this will allow
you to transfer about 70GB per day.
[Bob Briscoe, BT]
33
Summary
• Bandwidth demand is growing– (both mobile and fixed)
– Just adding more bandwidth does not cut it
– Will be expensive – and be filled up fast
• “Traditional” solution space:– capping bandwidth volume
– forbidding & controlling apps by policy functions in the network
(QoS, IMS, NGN)
• Here: address issue at the network edge – on
Internet/transport/application layer
© NEC Corporation 2008
34
Summary
© NEC Corporation 2009
Robustness for
transport and
application
protocols
Rethinking
quality of
service
requirements
Enabling less
than best-effort
on transport
layer
Making IP
accountable for
better resource
sharing
Danke