Upload
lamtuyen
View
228
Download
1
Embed Size (px)
Citation preview
University of WürzburgInstitute of Computer ScienceProf. Dr. Phuoc Tran-Gia
Phuoc Tran-GiaInstitute of Computer Science
University of Würzburg
Trends towards Next Generation Internet
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
Impossible to see the future is.
(Master Yoda)
Searching the Future Internet
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
The IP bottleneck
IPIP
E-commerceE-commerce
VPNVPN
Call CenterCall
Center
Internet AccessInternet Access
ATMATM
WDMWDM
Intra-/Extranets
Intra-/Extranets
SDHSDH
TDMTDM
GSMGSM
VoIPVoIP
Web cacheWeb
cache
New applicationsNew applications
High-speed transportHigh-speed transport
Worldwide infrastructureWorldwide infrastructure
Traffic explosionTraffic explosion
Everythingover IPEverythingover IP
Peer to peer
Peer to peer
Interact. video comm.Interact. video comm.
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
IP & Post-IP
science or repair shop ?
future application
IP protocol
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
Usage (service?) evolution
Internet evolution is mainly driven by innovative usages (or services)
telnet
WWW
… Future Applications
Evolution
P2P
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
Projects toward Future Internet
USA
GENI : Global Environment for Network Innovations (USA)
FIND : Future Internet Network Design
u-IT839 : Future of the Internet for Korea (Korea)
European Union: FP6 & FP7:
Network of Excellence: Euro-NGI, Euro-FGI, …
EIFFEL
other EU Framework 7 projects to come
Japan: Collaborative Overlay Research Environment (CORE)
IKT 2020 (BMBF, Germany)
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
Trends towards Next Generation Internet
1. Trends and observationsProjects towards Next Generation Networks
Multi-Network Service & Edge-based Intelligence
2. Service Evolution
Overlay and self-organizing networks
Changes in user traffic behavior
Functional versus stochastic scalability
Example: VoIP signaling platform with an overlay network
3. Challenges
Quality of Service or Quality-of-Experience
Evolutionary path to the Future Internet
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
Trends
From Multi-Service Network to Multi-Network Service
Blurring of Link between Network Provider and Services
From Quality of service (QoS) to Quality of Experience (QoE)
Edge-based Intelligence
Content placement: boundary between providers and users is disappearing
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
Network-centric: Multi-Service Networks
Service
Service
Provider
Network
Provider
Application
Internet Service
Provider A
Carrier A Carrier B Carrier X
Telephony Service
Provider XY
Internet Service
Provider B
PSTN
TelephonyWebServices Other Services
Internet
Telephony
Browser InstantMessenger
Soft Client Phone
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
Application-centric: Multi-Network Service
Network
Provider
WLAN
Wired
UMTS
Application
Service
Service
Provider
VoiceInstant
Messaging
Data
Transfer
Presence
InformationVideo
Personal
Communication
Application
PSTN
Internet Service
Provider A
Internet Service
Provider B
Service
Provider C
Multi-Network Service
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
Services, Platforms and Networks
User A User BServices:- Phone call- Content Distribution
Platforms
Skype SIPPOTSBitTorrenteDonkey KaZaA
NetworksWired
SensorsUMTS
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
Trends
From Multi-Service Network to Multi-Network Service
Blurring of Link between Network Provider and Services
From Quality of service (QoS) to Quality of Experience (QoE)
Edge-based Intelligence
Content placement: boundary between providers and users is disappearing
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
Overlay Control Structure
NGNcore
transport network
overlaystructure
supportingnew service
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
Stepwise Traffic Change
web
7,9%
OTHER
23,3%
FTP
0,3%
1,2%
P2P applications
67,4%
source:Telefonica 2004Jose EnriquezCOST 279, Rome: traffic observed in a transit router
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
Trends in Services and Platforms
Until now: Service design by network provider
classic way to design service, provider and platform-dependent
Future Evolution: Edge-based service design
designed and deployed by user groups
transition from disruptive technology to business cases
edge-based intelligence & application-driven overlay structure
example: P2P content delivery
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
Trends
From Multi-Service Network to Multi-Network Service
Blurring of Link between Network Provider and Services
From Quality of service (QoS) to Quality of Experience (QoE)
Edge-based Intelligence
Content placement: boundary between providers and users is disappearing
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
Intelligence Placement & Service
Da
ta
Da
ta
Da
taD
ata
Da
ta
User InitiatedService
Server
Da
ta
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
Examples for Edge-Based Intelligence & Platforms
free VoIP-Platform
since July 2004
acquired September 2005 (4.1 Billion $) by eBay
Source: Wikipedia
video clips platform
since February 2005
acquired October 2006 (1.3 Billion $) by Google
Platform for pictures
since 2002, first as feature of an on-line game
Acquired March 2005 by Yahoo
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
Trends in Services and Platforms
Transition to Multi-Network Services & Multi-Platform Services
highly dynamic network topology and traffic growth
customer behaviour changed, selfish users and applications
QoS issues unclear: customer perceived or network provider defined QoS
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
Trends towards Next Generation Internet
1. Trends and observationsProjects towards Next Generation Networks
Multi-Network Service & Edge-based Intelligence
2. Service Evolution
Overlay and self-organizing networks
Changes in user traffic behavior
Functional versus stochastic scalability
Example: VoIP signaling platform with an overlay network
3. Challenges
Quality of Service or Quality-of-Experience
Evolutionary path to the Future Internet
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
Handover Decision Unit
Example: Multi-Network (Vertical) Handover
GGSN
SGSNMSC
RNC
RNC
RNC RNC
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
Peer
Overlay connection,e.g., Pastry
Example: Multi-Network (Vertical) Handover
GGSN
SGSNMSC
RNC
RNC
RNC RNC
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
Thinning the protocol architecture
Physical
Data Link
Network
Transport
Session
Presentation
Application
OverlayOverlay adapted to underlying layers
Increase efficiency and robustness
Supporting transport over any kind of
networks
Heterogeneity of network technologies
Interoperability between technologies
Application
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
Thinning the protocol architecture
Data Link
Network
Transport
Session
Presentation
Connectivity
Mediation
Application
Physical
Overlay
Application
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
Trends towards Next Generation Internet
1. Trends and observationsProjects towards Next Generation Networks
Multi-Network Service & Edge-based Intelligence
2. Service Evolution
Overlay and self-organizing networks
Changes in user traffic behavior
Functional versus stochastic scalability
Example: VoIP signaling platform with an overlay network
3. Challenges
Quality of Service or Quality-of-Experience
Evolutionary path to the Future Internet
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
Selfish application: positive feedback bitrate
pack
et/b
it ra
te
time30ms
QoS assessment period QoS assessment
low QoS : increase
packet/bit rate
still low QoS: introduce other counter measures (e.g.,
Codec change, app. layer re-route)
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
Emulating Dynamic Changes
Interneteth0
eth1eth2
IPIPIPIPIP
IP
packet loss
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
Bandwidth Adaptation Based on QoS
Packet sent times depend on codec, independent on packet loss
Variable bit rate by increasing packet size, i.e. more audio data
sender
receiver
0 0.5 1 1.5 2 2.5 3
x 106
0
0.5
Time [ms]
Pac
ket
loss
0 0.5 1 1.5 2 2.5 3
x 106
0
50
0 0.5 1 1.5 2 2.5 3
x 106
0
50
Tim
e b
etw
een
pac
kets
[m
s]
0 0.5 1 1.5 2 2.5 3
x 106
0
0.2
0.4
Time [ms]
Pac
ket
loss
0 0.5 1 1.5 2 2.5 3
x 106
0
200
400P
acke
t si
zes
0 0.5 1 1.5 2 2.5 3
x 106
0
200
400
sender
receiver
emulatedpacket loss
packetloss
packet size
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
Trends towards Next Generation Internet
1. Trends and observationsProjects towards Next Generation Networks
Multi-Network Service & Edge-based Intelligence
2. Service Evolution
Overlay and self-organizing networks
Changes in user traffic behavior
Functional versus stochastic scalability
Example: VoIP signaling platform with an overlay network
3. Challenges
Quality of Service or Quality-of-Experience
Evolutionary path to the Future Internet
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
Functional Scalability & Stochastic Scalability
Functional Scalability Stochastic Scalability
22 nodes
25 nodes
24 nodes,stable stationarystructure
24 nodes,higher churn rate(joins and leaves)
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
IP Call
InformationProvider
InformationSeeker
InformationPush
InformationPull
Chord Ring as Signaling Network
Example: P2P Voice-over-IP Signaling using Chord
Information:- nick-name- current IP-address- user profile
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
Performance analysis of a VoIP Signaling Platform
Voice-over-IP application with distributed P2P-based directories
Architecture: Signaling platform using Chord ring with distributed hash table
Scalability: how many customer can be supported by stochastically varying ring
size due to “churns”
Service Level: 99% of directory searches need less than 1 sec
Performance analysis with stochastic modeling approach
Research cooperation with Siemens
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
Model parameters
A
BTN
TN
TN one hop delay
X number of hops until searched peer is found
TA delay of the answer
T total search duration
X = 2
12
Wirednetwork
UMTSnetwork
WIMAXnetwork
TA
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
Phase Diagram of a Search Process
TN
TN
TN
TN TNTN
TN
TN
TN
TA
TA
TA
TA
p1
p0
p2
pi
pk
i - times
k - times
searchbegin
searchend
Probability, that the searched peer isi hops away
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
Stochastic Scalability
0 10 20 30 40 50 60
10-4
10-3
10-2
10-1
100
CT
N
= 0.5 1 1.5 2
t / E[TN
]
P(T
>t)
E[TN] = 50 ms10000 customers
Signalingplatformscales!
1 second
?
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
Quality of Service: Delay Quantile
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 100000
5
10
15
20
25
0.95-quantile
0.99-quantile
0.9999-quantile
Chord size n
sear
chde
lay
boun
d/ E
[TN
]
E[TN] = 50 ms
cTN= 1
0.75 sec
2000 customers
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
Conventional Service Architecture
Service2
Servicen
User/Application
Service1
time
time
time
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
Future Service-Oriented Architecture
Services
1Service Elements/Subservices /Substrates
2 3 M54
2 n1
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
Future Service-Oriented Architecture
Service ElementsSubservices
Service Supply Chain
Servive Elements BusService Discovery Engine
User/Application/Service
1 2 3 M54
time
time
time
Start ofapplication
End ofapplication
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
Trends towards Next Generation Internet
1. Trends and observationsProjects towards Next Generation Networks
Multi-Network Service & Edge-based Intelligence
2. Service Evolution
Overlay and self-organizing networks
Changes in user traffic behavior
Functional versus stochastic scalability
Example: VoIP signaling platform with an overlay network
3. Challenges
Quality of Service or Quality-of-Experience
Evolutionary path to the Future Internet
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
Trends
From Multi-Service Network to Multi-Network Service
Blurring of Link between Network Provider and Services
From Quality of service (QoS) to Quality of Experience (QoE)
Edge-based Intelligence
Content placement: boundary between providers and users is disappearing
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
Scenario
tQ
traffic adjustment
selfish
altruistic
usersbottleneck node
observe QoEfor period
QoEnotification
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
0 0.2 0.4 0.6 0.81
1.5
2
2.5
3
3.5
4
4.5
ploss
MO
S v
alue
measurementIQX hypothesis
IQX Hypothesis : Exponential Relation between QoE and QoS
QoE = 3.0819 ⋅ e-4.6446 ⋅ ploss + 1.07QoS
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
Trends towards Next Generation Internet
1. Trends and observationsProjects towards Next Generation Networks
Multi-Network Service & Edge-based Intelligence
2. Service Evolution
Overlay and self-organizing networks
Changes in user traffic behavior
Functional versus stochastic scalability
Example: VoIP signaling platform with an overlay network
3. Challenges
Quality of Service or Quality-of-Experience
Evolutionary path to the Future Internet
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
Disruptive Delelopment & Mainstream
ISDNPOTSMainstream
DisruptiveDevelopment
Old Data Services
ATM Future?
Today
IP
Next InternetTechnologies
P2P, VoIP, QuadPlay
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
MediationIncubator
MediationIncubator
Service Evolution, Testbed & Funding Strategies
InternetInternet
ApplicationsApplications
Evolved InternetEvolved Internet
Fut.
App. 1
Fut.
App. 1
Fut.
App. 2
Fut.
App. 2
Fut.
App. 3
Fut.
App. 3
Fut.
App. 4
Fut.
App. 4
10 years
Funding?
ApplicationPlayground
ApplicationPlayground
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
MediationIncubator
MediationIncubator
ApplicationPlayground
ApplicationPlayground
Fut.
App. 1
Fut.
App. 1
Fut.
App. 2
Fut.
App. 2
Fut.
App. 3
Fut.
App. 3
Fut. App. 4
Fut. App. 4
Service Evolution and Testbed
Evolved InternetEvolved Internet
5 years
Beebl
e
Beebl
e
Who
opee
Who
opee
TestbedTestbedAdapted TestbedAdapted TestbedAdapted TestbedAdapted Testbed
MyP
ipe
MyP
ipe
Dist.
CDP
Dist.
CDP
ApplicationsApplications
Evolved InternetEvolved Internet
MyP
ipe
MyP
ipe
Distr.
CDP
Distr.
CDP
InternetInternet
10 yearsToday
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
Trends towards Next Generation Internet
1. Trends and observationsProjects towards Next Generation NetworksMulti-Network Service & Edge-based Intelligence
2. Service EvolutionOverlay and self-organizing networksChanges in user traffic behaviorFunctional versus stochastic scalabilityExample: VoIP signaling platform with an overlay network
3. ChallengesQuality of Service or Quality-of-ExperienceEvolutionary path to the Future Internet– Testbed, Planetlab, Geni
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
Virtualization: Overlay & Slice Concept
PlanetLab Testbed
Slice A
Slice ?
…
Slice B
Slice ?
…
Slice A
Slice B
Slice ?
…
Slice A
Slice ?
…
Slice A
Slice B
Slice ?
…
PlanetlabNode
PlanetlabNode
PlanetlabNode
PlanetlabNode
PlanetlabNode
Project in (private) PlanetLab
Project in (private) PlanetLab
A
B
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
Virtualization: Overlay & Slice Concept
PlanetLab Testbed
Slice A
Slice ?
…
Slice B
Slice ?
…
Slice A
Slice B
Slice ?
…
Slice A
Slice ?
…
Slice A
Slice B
Slice ?
…
PlanetlabNode
PlanetlabNode
PlanetlabNode
PlanetlabNode
PlanetlabNode
Project in (private) PlanetLab
A
3) Project deploystest software
1) Configure PlanetLab Central
2) PlanetLab Central deploys slices
4) Run test software
6) Collect results
5) Establish overlaynetwork and measuredesired data
University of WürzburgInstitute of Computer Science Phuoc Tran-Gia
End of Talk
Thank you !