Transcript Document

Case Study: Resilient Backbone
Design for IPTV Services
Meeyoung Cha, Gagan Choudhry, Jennifer Yates, Aman
Shaikh and Sue Moon
Presented by Yuanbin Shen
March 25, 2009
1/32
Introduction

Nation-wide TV broadcast



Satellite-based
Terrestrial-based (typically over IP networks → IPTV)
IPTV architectural design




Integrate IPTV services with existing IP backbone
Construct a dedicated overlay network on top of IP
Construct a direct interconnected flat IP network
Integrate with an existing switched optical network
What is the best architecture for supporting IPTV?
2/32
Overview of IPTV Architecture
3/32
IPTV Traffic

Type




Broadcast TV: realtime
VoD download: non-realtime download to VHOs
Realtime VoD: realtime
Characteristics

Uni-directional and high-bandwidth
VoD traffic: highly variable

Multicast for broadcast TV / unicast for VoD

4/32
Design Options

Technology:


layer1 (optical) v.s. layer3 (IP/MPLS)
Topology:

hub-and-spoke v.s. meshed
5/32
Design Options (cont’d)
Access connections
Failure
working path
Failure recovery
Src
Dst
working path
Failure
Src
Dst
protection path
IP layer fast-reroute (FRR)
switching
Optical layer SONET protection
6/32
Model 1: Integrate With Existing IP Backbone




Backbone links are shared and access links are dedicated
Rapid deployment: using existing infrastructure
High resource utilization: share bandwidth between
applications
Drawback: IPTV quality easily impacted by Internet traffic
7/32
Model 2: Dedicated Overlay



Use common backbone routers to construct dedicated IPTV overlay
Easy for performance management: links are dedicated
Overhead to construct the overlay
8/32
Model 3: Flat IP (No backbone)
SHO
SHO
VHO
Long super links




Services routers (SR) directly connected using point-to-point links
over dense wavelength division multiplexors (DWDMs)
Connect geographically close VHOs into regional rings
Inter-connect rings with long super links
No existing infrastructure used
9/32
Model 4: Integrate with switched optical
network
SHO
SHO
L1 network
VHO




Multicast capabilities at optical nodes (new technology)
SHOs establish multicast trees, VHO receiving single best stream
Failure recovery: rapid switch between different paths
How to find physically-diverse paths from SHOs to each VHO?
→ NP-hard → use IP-based approach to create trees
10/32
Design Instances
Design
Int-IP-HS model.1
Int-IP-HS-FRR
Int-IP-Ring
Int-IP-Ring-FRR
Ded-IP-HS model.2
Ded-IP-HS-FRR
Ded-IP-Ring
Ded-IP-Ring-FRR
Layer
Link-Capacity
Access Type
Fast-failover
IP
..
..
..
Shared
..
..
..
Dual-homed
..
Ring
..
SONET links
Fast re-route
SONET links
Fast re-route
IP
..
..
..
Dedicated
..
..
..
Dual-homed
..
Ring
..
SONET links
model.3
P2P-DWDM
IP
P2P-DWDM-FRR
..
model.4
Opt-Switched
Optical
Fast re-route
SONET links
Fast re-route
Dedicated
..
None
..
SONET links
Fast re-route
Time-divisioned
Dual-homed
Disjoint paths
11/32
Evaluation
- Cost (capital) comparison of multicast and unicast


Multicast is much more economical than unicast
Optical network is more economical than IP network
12/32
Evaluation
- Cost (capital) comparison across design instances




Optical networks are more economical than IP networks
Total cost is dominated by access cost (except for IP flat design)
Ring access is good of multicast; dual-homed access is good for
unicast(VoD)
For backbone cost, the flat IP model is the most expensive
13/32
Conclusion





Explore potential IPTV designs in backbone network
Comparison across different design architectures
Significant benefits of using multicast for broadcast
TV
Optical design more economical than IP designs
Ring access attractive for broadcast TV; dual-homed
access attractive for VoD
14/32
When is P2P Technology
Beneficial for IPTV Services?
Yin-Farn Chen, Yennun Huang, Rittwik Jana, Hongbo
Jiang, Michael Rabinovich, Bin Wei and Zhen Xiao
Presented by Yuanbin Shen
March 25, 2009
15/32
Introduction

Problems in providing IPTV:


high deployment and maintenance cost
Server bandwidth limits

One solution → using P2P technology
Does P2P technology always works well for IPTV?
When is it beneficial?

Network models




Cloud model: overestimate P2P benefits
Physical model: more practical
Provide three incentive models to encourage P2P
sharing in IPTV under a physical model
16/32
Cloud Model


Simple for modeling
Does not consider the constraints of the underlining
service infrastructure
17/32
Physical Model
B2S
B1N
B1S
18/32
P2P Sharing within a Community
B2S
B1N
B1S
Bottleneck
Not beneficial
19/32
P2P Sharing within a Community
B2S
B1N
Bottleneck
B1S
Beneficial
20/32
P2P Sharing across Communities
B2S
B1N or B1S
Bottleneck
Not beneficial
21/32
Simulation Setup
B2S: 10 Gbps
Content server
(1000 programs, 120 mins, 6 Mbps)
22/32
Simulation Setup
20 communities
Content server
(1000 programs, 120 mins, 6 Mbps)
B1S
B2S: 10 Gbps
B1N: 0.622 Gbps
23/32
Results: cloud model v.s. physical model -1
Links across communities
are heavily utilized.
Limited by B1N
Total # of peers:
20*community size
24/32
Results: cloud model v.s. physical model -2
Don’t consider the bandwidth in the cloud
Traffic across communities increases
Limited by B2S
Total # of peers: 10000
Community size: 500
25/32
Results: cloud model v.s. physical model -3
Serves all active viewers
Limited by B1N
Limited by B1N, traffic across
communities reduces the bandwidth
Total # of peers: 10000
Community size: 500
26/32
Cost-Benefic Analysis

Maximum Profit for Conventional IPTV


Pnop2p = rN – Enop2p
P2P Incentive Models

Built-in Model:




Pb = rN – Enop2p – tN
r: fee paid by a viewer
N: number of viewers
tN: P2P installation expense
27/32
Cost-Benefic Analysis

Flat-reward Model:




Pf = rN – Enop2p – twN – dwN
w: percent of viewers sign up for P2P
d: reward per P2P user
Usage-based Model





Ps = rN – Enop2p – tN – qbuTN
u: average video rate
T: program length
q: credit per bit
b: percent of viewers download data from peers
28/32
Profit Per Unit Time
29/32
Simulation Results (Using MediaGrid Algorithm)
When system is sufficiently utilized
More peers → more benefits from P2P
Large differences among incentive models
Build-in model is the best under this setup
When system is under utilized
non-P2P may be better than P2P
30/32
Conclusion

Studied when P2P is beneficial for IPTV

Cloud model may overstate P2P benefits →
use physical model
Different incentive strategies lead to different
profits → choose a proper one for specific
application.

31/32
References




M. Cha, G. Choudhury, J. Yates, A. Shaikh, and S. Moon, “Case Study:
Resilient Backbone Design for IPTV Services”, In Proc. of International
Workshop on Internet Protocol TV Services over World Wide Web, May
2006
M. Cha, G. Choudhury, J. Yates, A. Shaikh, and S. Moon, Slides:
“http://an.kaist.ac.kr/~mycha/docs/mycha_www_iptv06.ppt”
Y. Chen, Y. Huang, R. Jana, H. Jiang, M. Rabinovich, B. Wei, and Z. Xiao,
“When is P2P Technology Beneficial for IPTV Services,” ACM NOSSDAV,
June 2007.
Meng-Ting Lu, Slides: “When is P2P Technology Beneficial for IPTV
Services,” http://nslab.ee.ntu.edu.tw/OESeminar/slides/When is P2P
Technology Beneficial for IPTV Services.ppt”
32/32