Graduate School of Information Sciences,
Tohoku University, Japan
Prof. Nei Kato
Perspective on
New Generation Networks
1
Introduction – Kato laboratory
Building of GSIS
http://www.it.ecei.tohoku.ac.jp/
Member Assistant Prof.: 1
Doctor course: 6
Master course: 13
Undergraduate: 4
Other: 3
Countries Vietnam,
Thailand, Egypt,
Bangladesh,
Venezuela,
America
2
Appearance of Convergent Network
IP-based networks
Ad hoc networks
Satellite networks
Cellular networks
Wired networks
WLAN
Ubiquitous, multimedia
services
3
Today’s Topics
Discussions about the Future Network
Next generation networks (NXGN)
New generation networks (NWGN)
Introduction of Our Research Activities
Researches for individual networks
sensor, ad hoc, satellite, wireless and wired networks
Researches about integration of networks
Mobility management, Cross-layer design, Protocol design
Researches from the aspects of applications
Application layer multicast, Contents delivery system
Next Generation Networks (NXGN)
New Generation Networks (NWGN)
Discussions of the Future Networks
5
Future Networks
1. Next Generation Networks (NXGN)
2. New Generation Networks (NWGN)
Two types of future networks
Generating new services based on
currently widely used networks
(NTT, KDDI, and many other network operators over the world)
Evolutionary approach
6
Next Generation Networks (NXGN)
ITU-T: International Telecommunication Union – Telecommunications standardization sector
QoS: Quality of Service
Features of NXGN (ITU-T recommendation Y.2001)
Wideband and capable of QoS control
Packet-based networks
Separation of service and transport
Unlimited access
Provision of universal Mobility and Ubiquitous
service
7
Overview of NTT NGN
Triple-Play Service
IPTV server
VoDserver
Other server
SNI
UNI
MulticastOne wayunicast
ISP
NNI
Bi-directionalunicast
PPPoE connection
NTT NGN
IPTVIP telephone (VoIP), TV telephoneInternet access
VoD
SNI: Server-Network Interface
NNI: Network-Network Interface
UNI: User-Network Interface
VoD: Video on Demand, VoIP: Voice over IP
PPPoE: Point-to-Point Protocol over Ethernet
Other NGN
8
QoS Control of NTT NGN
Bandwidth control using SIP
Similar to DiffServ
NTT NGN
SenderReceiver
SIP server
Data
Service
edge
ISP
Other NGN
PSTN /ISDN
Core router&
Edge router
Setting DSCP
Bandwidth allocation
Processing
based on DSCP
Internet
DiffServ: Differentiated Services
SIP: Session Initiation Protocol
DSCP: DiffServ Code Point
PSTN: Public Switched Telephone Networks
9
QoS Classification of NTT NGN
Unicast has 4 classes, Multicast has 2 classes
Only Best Effort connects can access to the Internet
Question How can actually ensure QoS with increased number of users?
How to collaborate with other companies?
What is the primacy over existing infrastructures (services)?
QoS classA
(Highest)
B
(High)
C
(Priority)Best Effort
latency short(70ms) mid(200ms) - -
jitter short(20ms) mid(200ms) - -
packet loss 0.1% - -
latency mid(400ms) -
jitter mid(200ms) -
packet loss 0.1% -
ISP (PPPoE) connection ○(From field trial document. Values: between UNI and SNI, -: no prescription)
Unic
ast
Multic
ast
10
Other
VoIP/Mobile network
Overview of KDDI Ultra 3G
Realization of FMBC(Fixed Mobile and Broadcast Convergence)
Transport
(next generation CDN)
IMS(MMD)
WiMAXWLAN
CDMA2000
New wireless system
Application servers
VoiceChat
PSTN
Serv
ice c
ontro
lIP
core
Access
Digitalbroadcast
VoD
FTTHxDSL
VoD: Video on Demand
IMS: IP Multimedia Subsystem, MMD: MultiMedia Domain
CDN: Contents Distribution Network, PSTN: Public Switched Telephone Networks
From KDDI News Release
TV tel. Online
game
11
Future Networks
Two types of future networks
Intending to create a new network infrastructure
with new services
(GENI, FIND, FP7, AKARI)
Revolutionary approach
1. Next Generation Networks (NXGN)
2. New Generation Networks (NWGN)
12
Toward NWGN
GENI&FIND (NSF, USA)
GENI (Global Environment for Network Innovations)
Promote research development of new network
architecture, service, application
Large test beds which take in the innovative
technologies (photonic, mobile, sensor, etc.)
FIND (Future InterNet Design)
Build up future internet architecture
Examination of the architecture developed on GENI
13
Toward NWGN
Euro-NGI/FGI (EU)
Responding to variety of access networks
Establish the basis of multi-network service
(FMC, seamless mobility, etc.)
AKARI project (NICT, JPN)
Design the network architecture on a clean state
Unconstrained by the existing internet architecture
Anticipate Non-IP environment
14
Some hurdles we need to get over
Existent Internet What features NWGN should have?
How do we switch the network architectures?
problem of smooth transition from current networks
Can we relocate most of the applications?
problem of how to deal with additional cost
Lack of general agreement on NWGN? People in academia have different way of thinking
What is the driving force for new generation
networks?
15
What is the direction to set our targets
Currently, we have problems on bandwidth,
security and seamless communications, etc.
How seamless a convergent network should be?
Who can pay how much?
Clarify the real frustrations over the current
Internet
Find the places where the current networks are
real in the danger of collapsing
16
One step at a time is important for NWGN
Different people may have different thinking
NWGN should be more controllable and secure
We need to solve many problems for the current
networks
Consider the killer applications is indispensable
to trigger the NWGN
17
Our Research Approaches
Providing solutions to current networks(First step) Wired/wireless networks
Sensor/ad hoc networks
Satellite networks etc.
Integrating heterogeneous networks for further
efficiency(Second step) Mobility management
Cross-layer
Protocol design issues
Designing the infrastructure(Third step) Contents delivery system & monitoring system
Application layer multicast
18
Part 1 : Researches for Individual Networks
Sensor networks
New Data Gathering Scheme
Ad hoc networks
Reliable Topology Control Algorithm
Hybrid MANET Strategy
Satellite networks
Traffic distribution in Low Earth Orbit (LEO)
satellite system
Data Gathering Scheme Based on
Clustering Algorithm
for Mobile Sinks in WSNs
20
Background
WSNs (Wireless Sensor Networks)
The self-organized networks which are composed of
deployed sensor nodes
Applications
Environmental monitoring and military use
Resource restrictions
Limited batteries and computational resource
MULEs
Adopt the mobile sinks to gather data from deployed sensor
nodes
How to gather data from sensor nodes considering the
energy efficiency is crucial in WSNs
21
MULEs (Mobile Ubiquitous LAN Extensions)
Mobile Sink Scheme
Advantage : High energy efficiency, which is
consumed by the sensor nodes
Disadvantage: High latency due to the arrival delay of
the mobile sinks
MULEs can reduce consumed energy as the mobile sink gathers data
Static sink scheme Mobile sink scheme (MULEs)
22
Conventional Scheme on MULEs
Deterministic Mobility Scheme
The mobile sink moves along the deterministic path
Random Mobility Scheme
The mobile sink moves randomly in the field
Mobile sinkSensor node
Deterministic Mobility Scheme Random Mobility Scheme
23
Proposed Mobility Scheme
KAT mobility
The mobility which achieves high efficiency
Procedure
1. Cluster the nodes by the k-means algorithm
2. Calculate the TSP-path among the centroids of each cluster
3. The mobile sink moves along the path calculated in 2
Mobile sink
Sensor node
Voronoi edge
Centroid of a cluster
Cluster
Where k = 3
24
Performance evaluation
Simulator Qualnet 3.9.5
Simulation time 30 [minutes]
Field 5000 5000 [km2]
Velocity 10 – 30 [m/s]
Pause time 20 [s]
Buffer size 10 [MB]
Data rate 512 [B/s]
Number of nodes 20 – 200 [nodes]
Number of sinks 1 [sink]
Number of clusters 10 [clusters]
Fig. 1. Simulation result of energy efficiency
Fig. 2. Simulation result of energy efficiency
with 5% dead sensors
Table 1. Parameter settings
H. Nakayama, N. Ansari, A. Jamalipour, and N. Kato,
"Fault-resilient Sensing in Wireless Sensor Networks,"
Computer Communications, Special Issue on Security on
Wireless Ad Hoc and Sensor Networks, Sep. 2007.
An Efficient and Reliable Topology Control
Algorithm in Wireless Ad-Hoc Networks
26
Background
Wireless Ad-Hoc Network
A decentralized wireless network without any
infrastructure such as base stations
Resource constraints
Low battery capacity
Low processing ability
Unstable links
Link failures frequently occur
Energy consumption must be minimized
Multipath transmission can enhance reliability
27
Reduction of energy consumption
Control each node’s transmission power properly
Challenges
Large transmission radius Small transmission radius
High interference
High energy consumptionNetwork may partition
The key is providing adequate topology control
28
Topology Control
Features
Global connectivity from local information
Each node operates independently
Low interference and low energy consumption
Problem: Resultant topology is unreliable because of single paths
Local operation Global connectivity
29
Proposed Method
Local Tree-based Reliable Topology (LTRT)
Overview
Improve on the reliability problem of existing topology control
Features:
Using only 1-hop information
Low calculation cost
Multi-connectivity of the topology is proved mathematically
Normal topology control LTRT
30
LTRT Algorithm
Operation of each nodes
1. Obtain local network from the information of adjacent
nodes
2. Calculate the reliable graph locally
3. Set the transmission radius to the level that can reach
the farthest neighbor
31
Performance Evaluation
Simulation environment
Comparative algorithms
CBTC: traditional approach
FLSS : nearly optimal, much higher computational cost
Metrics
Transmission radius: longest link length of each node
Measure of the transmission cost
Number of nodes 50 ~ 150
Simulation area 1000 [m] 1000 [m]
Maximum transmission radius 250 [m]
CBTC: Cone-based Distributed Topology-Control
FLSS: Fault-tolerant Local Spanning Subgraph
Compare the topologies derived under comparative algorithms
32
Simulation Result
• LTRT outperforms CBTC
• LTRT achieve close value to near-optimal algorithm
None
FLSS LTRT
CBTC
Transmission radiusTopologies
Kenji Miyao, Hidehisa Nakayama, Nirwan Ansari, Yoshiaki Nemoto, and Nei Kato, “A Reliable Topology for Efficient Key Distribution
in Ad-Hoc Networks”, Proc. of IEEE Workshop on Security, Privacy and Authentication in Wireless Network, 2008(Invited Paper)
Gateway selection scheme
for Hybrid MANET
34
Integration of MANET into the Internet
Interconnection through Gateways (GWs)
Expands mobile users’ activities
Presence of Sensitive Data
Private or Confidential
MANET: Mobile Ad-hoc Network
GW must be trustful under tight control
of a trusted Network Admin.
What is Hybrid MANET?
Gateway
InternetEx) disaster area, event place, etc.
Ex) biomedical information
business secret, etc.
Phone call Rescue
35
The larger a MANET gets,
the more GWs are required…
Problem
High cost for Admin. to tightly
control all the GWs
Security risk for sensitive data
Not all GWs
can be fully trusted
Challenges for ScalabilityNetwork Admin.
Trustful
Loosely controlled
Choose trusted GW for sensitive data
Need some modification to routing protocol
Countermeasure
36
MANET Routing Protocol
DYMO (Dynamic MANET On-demand)
Newest reactive protocol Representative of future MANET routing protocol
Successor of AODV Route discovery with RREQ & RREP messages
New/modified functionality
Simplified RERR behavior
Adaptation to unified packet format
Multiple interfaces*
Internet connectivity (GW deployment)*
Path accumulation* etc.
Our proposal
Establish a route to satisfy the security requirement
of data
*: optional
AODV: Ad hoc On-demand Distance Vector,
RREQ: Route Request, RREP: Route Reply, RERR: Route Error
Integrate
into DYMO
37
MANET
Integration of the ProposalDestinations
n-Datas-Data
Approach
Classification according to data type
GW, routing message, routing entry
n-Data s-Data
Security
requirementLow High
Route discoveryn-RREQ
n-RREP
s-RREQ
s-RREP
Relay by n-GW ○ ×
Relay by s-GW ○ ○
7
35
2
4
9
6
8
n-GW s-GW
• GW responds to RREQ with RREP
• Only s-GW can forward s-Data
Route discovery is based on
Hop count & Data type
Internet
n: normal, s: special
1… …
… …
38
Proposed
(s-GW: 3, 5, 7, 9)DYMO
Performance Evaluations
21
2
1 Simulation by QualNet 4.0
1. No difference to each data type
2. Only s-GW forwards s-Data
Enables sensitive data to be
delivered through trusted route
Simulation MANET topologyT. Matsuda, H. Nakayama, S. Shen, Y. Nemoto, and N. Kato, “On the Gateway Selection
Protocol for DYMO-based MANET”, IEEE WiMob 2008(Best Student Paper Award)
A Routing Protocol for Multi-Hop
NGEO Satellite Networks
40
Multi-Hop Satellite Networks are:
Low Earth Orbit (LEO) satellite networks
Medium Earth Orbit (MEO) satellite networks
Features:
Short propagation delay
Suited for real-time sessions
Low energy consumption
Higher mobility
Global coverage area
To mitigate the digital divide
Multi-hop satellite networks can be quite attractive as
“Global Wireless Networks”
Background
LEO
MEO
41
User distribution variance is very high
Satellites covering urban areas are more likely to be congested
Network resource over rural area or sea has more margin
High
densityLow density
High
densityLow density
High
density
There is need for an efficient routing protocol for traffic load
balancing over multi-hop satellite networks
Issues of Multi-Hop Satellite Networks
42
ELB: Explicit Load Balancing
ELB can be implemented over any routing protocols
This scheme produces an explicit exchange of current congestion status among neighboring satellites
A satellite with high traffic load requests its neighboring satellites to reduce their data forwarding rates
Neighboring satellites reduce their transmission rates of traffic originally destined to the “soon-to-be congested” satellite and search for other alternative paths that do not include the satellite
Our Proposal (ELB)
43
A B C D
1
2
3
4
Flow 1 Flow 2
Congestion
1. Satellite C3 enters Busy
State
Overview of ELB-Algorithm
44
A B C D
1
2
3
4
Flow 1 Flow 2
BSA
Detoured
Detoured
1. Satellite C3 enters Busy
State
2. Satellite C3 sends Busy
State Advertisements
(BSAs) to its neighboring
satellites (B3, C2, C4, D3)
3. They start traffic detouring
in order to decrease traffic
through Satellite C3. A
portion of the traffic is
detoured based on Traffic
Reduction Ratio (TRR), c
Overview of ELB-Algorithm
45
A B C D
1
2
3
4
Flow 1 Flow 2
Detoured
“Free”
Detoured
1. Satellite C3 enters Busy
State
2. Satellite C3 sends Busy
State Advertisements
(BSAs) to its neighboring
satellites (B3, C2, C4, D3)
3. They start traffic detouring
in order to decrease traffic
through Satellite C3. A
portion of the traffic is
detoured based on Traffic
Reduction Ratio (TRR), c
4. As a result of this detouring,
the congested satellite C3
enters Free State due to
decrease of its traffic load • Congestion alleviation
• Better traffic distribution
Overview of ELB-Algorithm
46
The ELB scheme achieves lowering packet drops
and increasing the total throughput
T. Taleb, D. Mashimo, A. Jamalipour, K. Hashimoto, N. Kato, and Y. Nemoto, “Explicit Load Balancing Technique for NGEO Satellite
IP Networks with On-Board Processing Capability,” IEEE Trans. of Networking. (Accepted)
Performance Evaluation
47
Part 2 : Integration of Networks
Mobility management scheme
Cross-layer approach for TCP in WLAN
Protocol design in transport layer
A Mobility Management Scheme
for Mobile IPv6 Networks
49
Movement for All IP Network
Cellular/WiMAX
High demand for seamless communication through
heterogeneous networks
IETF proposed a packet-based mobility management
protocol called Mobile IP version 6 (MIPv6)
IP Networkwired
IETF: Internet Engineering Task Force
WLAN
50
Internet
Mobile nodes
Correspondentnode
MAP
BS
HMIPv6: Hierarchical MIPv6
MAP: Mobility Anchor Point
Mobility Management (MIPv6, HMIPv6)
Over concentration
Ex.) Hierarchical Mobile IPv6 MIPv6
Support for mobile nodes
HMIPv6 Introducing MAP
Reduce handoff latency
Advance seamless handoff
Problem
Load imbalance among
MAPs
Access Router
51
Our proposal (ADMAPS)
Objectives
Achieve the minimum load variance of all MAPs
Considering both load distribution and handoff delay
Three Main Steps
MN AR MAP1 MAPk MAPM
1. Load notification
3. MAP selection
・・・ ・・・
RS
RA
2. Load variance computation and updatehandoff
52
Example of MAP Selection
Select new MAP to minimize load variance
MAP1 MAP2
MAP4
AR1 AR2
CNEx.) MAP1 MAP2 (load difference)M : number of MAPs
i : current MAP
j : next MAP
b : data transmission rate
of a MN
B : link bandwidth
l : load of a MAP
C : processing speed
(MAP)
: weight
MN
53
Performance Evaluations
BUs to HA ratio Lowest ratio of BU message
Load TransitionWell traffic distribution among
MAPs
0
20
40
60
80
100
2 4 6 8
BU
s to
HA
ra
tio [%
]
100
60
40
20
0
80
BU
s to H
A r
atio [
%]
33.3
63.2 65.9
26.2
HMIPv6 HMIPv6-UP DEMAPS ADMAPS
Load [
%]
60
80
0
40
20
200 600 1000 1400 1800
Time [sec]
HMIPv6
Load [
%]
60
80
0
40
20
ADMAPS
T. Taleb, Y. Ikeda, K. Hashimoto, Y. Nemoto, N. Kato, "An Application-Driven Mobility Management Scheme for Hierarchical Mobile IPv6 Networks," IEEE ICC 2007.
T. Taleb, A. Jamalipour, N. Kato, and Y. Nemoto, "A Load-Transition Based Mobility Management Scheme for an Efficient Selection of MAP in Mobile IPv6 Networks," IEEE Trans. on Vehicular Technology. (Accepted)
Cross-layer approach for TCP window
control in multi-rate wireless LAN
55
Background
IEEE 802.11 PHY provides
multiple data rates.
802.11a/g support
8 data rates (6–54Mbps).
Rate adaptation algorithm
(ex: ARF) at the MAC layer
selects one of the data rates.
Cross-layer mechanism can
provide better performance.
Exchanges information among layers.
PHY: Physical
MAC: Medium Access Control
ARF: Automatic Rate Fallback
AP: Access Point STA: Station
48Mbps
36Mbps
24Mbps
18Mbps
12Mbps
9Mbps
6Mbps
tux@linux#
tux@linux#
54Mbps
AP
STAtux@linux#
Physical
MAC
Network
Transport
Application
Cro
ss-layer
mechanis
m
56
Challenges
“Performance anomaly” in multi-rate WLAN
When stations with a different data rate exist in BSS,
aggregate throughput in the BSS is degraded.
Stations transmitting at the lower data rate occupy
the medium for a long time.
WLAN: Wireless Local Area Network
BSS: Basic Service Set
DCF: Distributed Coordination Function
Channel occupancy time
Standard DCF mode (Fair access opportunity)
Time required to transmit/receive a frame @ low data rate
Ideal (Fair access time)
16 frames
25 frames
Time required to transmit/receive a frame @ high data rate
57
Existing solution and its problem
Several approaches [1] etc. controls MAC parameters.
Increasing contention window size (decreasing the probability of
transmitting frames) of STA with low data rate.
Shortening frame length of STA with low data rate.
Most research works do not consider the ARF.
TBR: Time-based Regulator
frame @ low data rate
frame @ high data rate
Channel occupancy time
Channel occupancy time
58
Our Proposal
Cross-layer design
TCP sender controls the maximum window size
based on the estimated throughput.
PHY
MAC
IP
TCP
APP
Computes the channel
occupancy time available• Number of active stations
• Channel occupancy time
• Transmission/reception data rate
• Round trip time
• Use delayed ACK?, etc…
• Number of connections
• Average packet size
Estimates the maximum throughput
59
Simulation setup
N stations download a file from different FTP servers.
Simulator: QualNet 4.0.1
PHY: 802.11a (5.2GHz)
Two-ray propagation model
Constant shadowing model without fading
Data rate: 54/48/36/24/18/12/9/6Mbps
Cell radius: ca. 36m@54Mbps
ca. 360m@6Mbps
RTS/CTS option: enabled
Velocity of stations: 0–10m/s
Distances from AP to STAs: 0–360m
MSS: 1460bytes
Number of STAs: 1–20
tux@linux#
N FTP servers
N stations
AP
Stations randomly roams in the BSS.
FTP
…tux@linux#
…
100Mbps
20ms
FTP: File Transfer Protocol
MSS: Maximum Segment Size
RTS/CTS: Request/clear To Send
360m
60
Simulation results
The proposed scheme exhibits:
Improving of the aggregated throughput.
Decreasing of packet drops.
Fairness in terms of channel occupancy time
among the competing stations.
K. Kashibuchi, N. Kato “Performance Enhancement of TCP over Adaptive Multi-Rate IEEE 802.11 Wireless LANs”, IEEE IWS’08
Aggregated throughput in BSS Fairness in channel occupancy*
N
i i
N
i i
tN
tf
1
2
1*
: channel occupancy time
used by STAi
it
Transport layer protocols:
Router-supported approach
62
Fundamental issues of TCP
AIMD window control in TCP AI: Window is increased upon
succeeding a packet transmission
MD: Window is decreased upon
detecting a packet loss by
duplicate ACKs or timeout
Limitation of AIMD The degree of increment and reduction is not always suitable to
network congestion level.
→ Lower link utilization, Too packet drops
The increasing speed of window depends on RTT
→ Unfair bandwidth allocation
The cause of packet loss is unknown
→ Throughput degradation in wireless environments
TCP: Transmission Control Protocol AI: Additive-Increase MD: Multiplicative-Decrease RTT: Round Trip Time
63
Approaches to improve TCP performance
End-to-End approaches
Advantage: Simple mechanism and easy implementationDrawback: Just a little performance improvement
TCP Vegas (L.S. Brakmo et al., 1995)
HighSpeed TCP (RFC3649, 2003)
TCP Westwood/Westwood+ (S. Mascolo et al., 2004)
TCP New Jersey (K. Xu et al, 2005)
Router-supported approaches
Advantage: Drastic performance enhancementDrawback: Requiring an additional function to a network
eXplicit Control Protocol (XCP) (D. Katabi et al., 2002)
Explicit Window Adaptation (EWA) (L. Kalampoukas et al, 2002)
Enhanced TCP (M. Savoric, 2004)
T-REFWA+ (H. Nishiyama et al., 2006)
T-REFWA: Terrestrial-Recursive, Explicit, and Fair Window Adjustment
64
Router-supported approach (T-REFWA+)
Proper window size is fed back from the bottleneck router to a source via RWND field in an ACK packet
Advantages of our proposal, T-REFWA+, compared with
other router-supported schemes
NOT require additional field in TCP/IP header
Differentiate bandwidth allocation is available as well as fair
allocation (QoS can be supported)
Throughput degradation in wireless environments can be improved
RWND: Receiver’s advertised window
Source
DestinationProper window size
Traffic convergence
65
Performance evaluation (Wired)
Proposed methods achieve:
High link utilization
Fair bandwidth allocation
Sources Destinations
T. Taleb, H.Nishiyama, A. Jamalipour, N. Kato, and Y. Nemoto, “A Fair TCP-Based Congestion Avoidance
Approach for One-to-Many Private Networks,” in Proc of IEEE International Conference on Communications, 2006.
66
Performance evaluation (Wireless)
Proposed scheme achieves high throughput even in
environments with high link-error related losses
[Invited Paper] H. Nishiyama, T. Taleb, N. Ansari, Y. Nemoto, and N. Kato, "On the Performance of Congestion Control
Protocols in Lossy Wireless Networks," in Proc. of Wireless Rural and Emergency Communications Conference, 2007.
67
Part 3 : From the Aspects of Applications
Application Layer Multicast (ALM)
Contents delivery monitoring system
Robust and Efficient Stream Delivery
for Application Layer Multicast
69
The rapid growth in network speed and bandwidth
Increase of contents delivery
Unicast communication increases the traffic load
of the server and network
Introduction of IP multicast
Reducing the load of server by IP multicast
Replicate the contents at the intermediate routers
Drawbacks
Realizing IP multicast incurs
high cost
Infrastructure
Protocol
Background
DeliveringServer
DeliveringServer
ALM (Application Layer Multicast)
70
ALM: Application layer multicast
Multicast communication is realized
in the application layer
The duplication and relay of packets
is performed by the end-host
Communication between end-hosts
is unicast
ALM constructs multicast trees and
delivers the stream through this tree
Problem of ALM
A node leaves from multicast tree
The stream cannot be delivered to the descendant nodes
Multicast Tree
71
Existing solution
Multiple-tree Protocol
This method splits the stream into several sub-streams and deliver each
streams by using multicast trees in parallel
ex.) CoopNet, SplitSteam, and THAG
・・・
Stream 1
Stream 2
Stream K
Independent trees
Tree 1 Tree 2 Tree K
THAG:
THAG constructs the independent trees
Independent trees can be constructed
by making a node which is a parent
node in the specific tree to be the leaf
node in all other trees
Construction of independent trees
guarantees that the departure of any
node will only affect data delivery in at
most one multicast tree.
THAG does not manage the bandwidth contribution of each node
72
THAG does not consider node’s bandwidth condition
All nodes join all multicast trees
If the node’s bandwidth is narrow,
the node does not transfer all streams
Problem of THAG
Link bandwidth will vary with each user in a real network
Internet
FTTH Nodes
~100Mbps
ADSL Nodes
1.5~50Mbps
Wireless Nodes
11~54Mbps
・・・
Tree 1 Tree 2 Tree K
JoinQoS of stream is compromised dramatically
Server
Low bandwidth node
73
Our Proposal
The node join the appropriate number of multicast trees based on
node’s bandwidth
Each node receives the streams properly
Proposed method locate the node with high speed link in higher tree
position
Improvement efficiency of streaming delivery
・・・Tree 1 Tree 2 Tree K
Join
: High bandwidth node
: Low bandwidth node
Proposed method adaptive to the variation of link bandwidth
Low bandwidth nodeTree 1 Tree K
・・・
74
Performance Evaluation
We evaluate performance of our proposal method in real network
environments by simulation
Throughput Delay
The proposal method provides high throughput and low delay
M. Kobayashi, H. Nakayama, N. Ansari, and N. Kato, “Robust and Efficient Stream Delivery for Application
Layer Multicasting in Heterogeneous Networks” IEEE Trans. on Multimedia(Accepted).
Traitor Tracing Technology of
Streaming Contents Delivery
using Traffic Pattern
76
Digital Rights Management (DRM)
Basic Idea
The access right for the content is provided only to the
user who has a license
Contents
server
Contents
User
Decrypt
[4] Distribute encapsulated contents
License
server
[2] Call on license issue
[3] License assignmentCooperate
[5] Decrypt and play[1] Encryption
Encapsulate
77
Traitor Tracing Technique
Problem of DRMUnable to stop ongoing abuses
Traitor Tracing Monitor the use of the content
Detect and stop the abuse
Embed copyrights
into the content
Report the usage of the content
based on embedded data
Check misuse
modify embedded data
to avoid the report
Unable to check
Monitoring Problems
Embed copyrights
into contents
78
Monitoring Real-time Contents Distribution
Use only information about traffic amount on the network
no process is needed for the user node
Matching to traffic pattern for assessment (watching or not)
Contents
server
Management
server
User
[1] Observe the traffic
[3] Match patterns
[4] assessment
Pattern transformation
Dynamic determination threshold
Adopted to the network condition
Server User
[2] Send the traffic to
Management Server
79
Example of Pattern Matching
Experimental Results in Real Network
(graph of the cross-correlation coefficient)
Watching Not watching
Threshold changes dynamically according to the network condition
If the user watches contents, there is a value higher than threshold
M. Dobashi, H. Nakayama, N. Kato, Y. Nemoto, and A. Jamalipour, “Traitor Tracing Technology of Streaming Contents Delivery
using Traffic Pattern in Wired/Wireless Environments,” IEEE Globecom2006
80
Conclusion
Discussions about future networks
Introduction of our research activities
I believe those accumulative efforts will finally
make change for reaching a concept of new
generation networks.
Thank you.