Infrastructure Design for IPTV Services

IPTV AsiaNovember 8-9, 2006

Grand Copthorne Waterfront Hotel, Singapore

Sue MoonJoint Work with

Meeyoung Cha (KAIST)W. Art Chaovalitwongse (Rutgers/DIMACS)

Gagan Choudhury, Zihui Ge, Aman Shaikh, Jenniver Yates (AT&T)

2

Push behind IPTV

TV service over IP Replacement of TV distribution networks Core service of “Triple Play” (voice, data, video) and “Quadr

uple Play” (+wireless/mobile)

Evolution Path Controversy over distinction between broadcasting and co

mmunication Bundled vs blended services As seen here so far!

3

Technical Challenges of IPTV

Distribution network WAN, MAN, and access technologies

Resilient design required

QoS guaranteeSame level of quality as today’s TV offers

Platform Standardizations: AV coding, EPG/ESG (eletronic program

ming/service guide), device mgmt, ... Middleware, settop box DRM (digital rights mgmt)

Today’s conditional access system not enough

4

Talk Outline

Service Architecture Overview

Comparison of Design Choices [Cha06-1] Path Protection Routing in WDM Mesh

Networks [Cha06-2] Efficient and Scalable Algorithms [Cha06-3]

5

SHO

Regional Network

Regional Network

Video Hub Office (VHO)

2 SHOs and 40 VHOs across the US

customers

Regional Network

Regional Network

Backbone Distribution Network

Super Hub Offices (SHO)

VHO

VHO

Broadcast TVVoD

Regional Network

Regional Network

How can we provide reliable IPTV servicesover the backbone network?

Service Architecture of IPTV

6

IPTV Traffic

Type Broadcast TV: realtime, 1-3Gb/s Popular VoD: non-realtime download to VHOs Niche (esoteric) VoD: realtime, 0-3 Gb/s per VHO

Characteristics Uni-directional and high-bandwidth High traffic variability expected for VoD Multicast for broadcast TV / unicast for VoD

Comparison of Design Choices

8

Design Space

Technology: layer 1 optical vs. layer 3 IP/MPLS Service layer topology: hub-and-spoke vs. meshed

(ring-based) Access connections: dual-homed vs. ring

Dual-homed Ring

Backbone Backbone

VHO

9

Design Space

Reliability Goal: resilient to single SHO/router/link failures Mechanisms: Fast-failover + routing protocols

working pathSrc

Dst

Failure

switching

Optical layer SONET protection

Src Dst

working path

protection path

IP layer fast-reroute (FRR)

Failure

10

IP designs

Optical design

Potential IPTV Designs

New dedicated IP backbone for IPTV Integrating with existing IP backbone Dedicated overlay over existing IP backbone Directly inter-connect IP routers (no backbone) Integrating with existing optical backbone

11

SHOSHO

BackboneVHO VHO

Support IPTV as multicast application (VoD as unicast) VHO receives single stream from the nearest SHO

Single network to manage Backbone links are shared (careful QoS) Various access connections, fast-failover schemes

Alt #1: Integrate With Existing IP Backbone

12

Backbone

SHO SHO

VHO VHO

Inter-connect common backbone routers with dedicated links

Backbone links are dedicated for IPTV (no QoS) Overhead for managing overlay Various access connections, fast-failover schemes

Alt #2: Dedicated Overlay of Existing IP Backbone

13

Connect geographically close VHOs into regional rings

Inter-connect rings with long haul links

Security is higher than using IP backbone No access part Fast-failover

Meshed topology (carry “through” traffic)

Alt #3: Flat IP (No Backbone)

Long haul links

SHO

SHO

VHO

VHO

14

Alt #4: Integrating with Existing Optical Backbone

Multicast capabilities at optical nodes (new technology) SHOs establish multicast trees, VHO receiving single best stream

Fast-failover is not yet supported in optical multicasting

SHOSHO

L1 network

VHO

15

Review: Design Choices

Technology

Service layer topology

Fast-failover

Link capacity

IP or optical

SONET links, fast-reroute, or physically diverse paths

Dedicated or shared

Hub-and-spoke or highly meshed

AccessDual-homed or ring

16

Design Instances

Design Layer Link-Capacity Access Type

Fast-Failover

Int-IP-HSInt-IP-HS-FRRInt-IP-RingInt-IP-Ring-FRR

IP......

Shared......

Dual-homed

..Ring

..

SONET linksFast re-routeSONET linksFast re-route

Ded-IP-HSDed-IP-HS-FRRDed-IP-RingDed-IP-Ring-FRR

IP......

Dedicated......

Dual-homed

..Ring

..

SONET links

Fast re-route

SONET links

Fast re-route

P2P-DWDM P2P-DWDM-FRR

Optical..

Dedicated..

None..

SONET links

Fast re-route

Opt-Switched Optical Time-divisioned Dual-homed

Disjoint paths

Alt #1

Alt #2

Alt #3

Alt #4

17

Cost Analysis: Capital Expense vs Traffic Loads

Cost comparison across traffic demands

0.0

5.0

10.0

15.0

20.0

M1+

U0

M2+

U0

M3+

U0

M1+

U1

M2+

U2

M3+

U3

M1+

U0

M2+

U0

M3+

U0

M1+

U1

M2+

U2

M3+

U3

Rel

ativ

e co

st

access

backbone

Int-IP-HS-FRR Opt-Switched

Ma+Ub: multicast a Gb/s + unicast b Gb/s

Increase in VoD loads has significant impact on the overall cost. → Having highly accurate VoD load forecasts is important!

MulticastMulticast

Unicast+

Multicast

Unicast+Multicast

18

Capital Expense Across Designs (Broadcast TV)

Multicast 3Gbps + Unicast 0Gbps

0.0

1.0

2.0

3.0

4.0

5.0

6.0

Ded

-IP

-HS

Ded

-IP

-HS

-FR

R

Ded

-IP

-Rin

g

Ded

-IP

-Rin

g-F

RR

Int-

IP-H

S

Int-

IP-H

S-F

RR

Int-

IP-R

ing

Int-

IP-R

ing-

FR

R

P2P

-DW

DM

P2P

-DW

DM

-FR

R

Opt

-Sw

itche

d

Rel

ativ

e co

st

accessbackbone

1. Optical designs are more economical than IP-based ones.2. Cost is dominated by access part (except for flat IP designs).3. For IP designs, FRR is economical then using SONET links.

19

Access Structure vs Traffic LoadsMulticast 3Gbps + Unicast 0Gbps

0.0

1.0

2.0

3.0

4.0

5.0

6.0

Ded

-IP

-HS

Ded

-IP

-HS

-FR

R

Ded

-IP

-Rin

g

Ded

-IP

-Rin

g-F

RR

Int-

IP-H

S

Int-

IP-H

S-F

RR

Int-

IP-R

ing

Int-

IP-R

ing-

FR

R

P2P

-DW

DM

P2P

-DW

DM

-FR

R

Opt

-Sw

itche

d

Rel

ativ

e co

st

access

backbone

Multicast 3Gbps + Unicast 3Gbps

0.0

10.0

20.0

30.0

40.0

Ded

-IP

-HS

Ded

-IP

-HS

-FR

R

Ded

-IP

-Rin

g

Ded

-IP

-Rin

g-F

RR

Int-

IP-H

S

Int-

IP-H

S-F

RR

Int-

IP-R

ing

Int-

IP-R

ing-

FR

R

P2P

-DW

DM

P2P

-DW

DM

-FR

R

Opt

-Sw

itche

d

Rel

ativ

e co

st

access

backbone

Ring access Dual-homed accessmulticast only multicast + VoD

multicast only multicast + VoD Ring access is more economical when only multicast traffic is considered. Dual-homed is better for VoD (no through traffic).

Flat IP design becomes expensive when VoD considered.

Dual-homed Ring

20

Summary

Explore potential IPTV designs in backbone network Comparison across different architectural

alternatives (use realistic capital cost model)

Design instances generated based on real topologies

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

Path Protection Routingin

WDM Mesh Networks

22

Motivation

Optical design known most economical [cha06-01] Fast fail-over not yet available in optical multicast

Provisioning approach in optical backbone [SRLG]- Design multicast trees (from SHOs to VHOs) in a failure-resilient and cost-effective manner

23

Layered architecture

Link failure in one layer → multiple failures in the upper layerTwo disjoint links may belong to a common SRLG

What is SRLG (Shared Risk Link Group)?

24

Examples of SRLGs

two sources

multiple destinations

riskspath conduit

bridge, tunnel

25

Service Requirements of IPTVIPTV Backbone Design Goals

Fault Tolerance Customers expect “always-on” service Resiliency against SRLG failures

Use redundant SRLG diverse paths from SHOs to VHOs

Low Cost To be competitive in the market Each link associated with port / transport cost

Find minimum cost multicast trees

26

SHOSHO

VHO

VHO

VHOVHO

Backbone

VHO

Path Protection Routing Problem

How to create two multicast trees such that (1) provisioning cost is minimized and (2) VHOs have physically disjoint paths to SHOs?

Path Protection Routing Problem

27

Link-Diverse vs SRLG-Diverse

d1 s2

s1 d2

d3

d1 s2

s1 d2

d3

(a) Link-diverse routing, cost=8 (b) SRLG-diverse routing, cost=9

risk1

risk2

risk1

risk2

unused Multicast path by s1 Multicast path by s2

28

An SRLG-Diverse Solution: Active Path First

1. Construct a minimum spanning tree from one source2. Remove all SRLG links of the first tree3. Build the second minimum spanning tree with remaining links

d1 s2

s1 d2

d3

d1 s2

s1 d2

d3

First tree from s1 Second tree from s2 (reduced graph)

(a) Active Path First routing, cost=10

risk1

risk2

29

Trap Situation of APF

d1 s2

s1 d2

d3

d1 s2

s1 d2

d3

First tree from s2 Fail to find second tree from s1

(b) Active Path First routing, trap situation

risk1

risk2

30

Our Provisioning Approach

Include SRLG-diverse constraints and solve the problem thru Integer Programming (IP)

Compare against APF (Active Path First) heuristic Less resilient source-diverse design Less resilient link-diverse design

31

Integer Programming Formulation

Minimize total cost

SRLGdiversity

Flowconservation

32

Applying Our IP Formulation

Dataset2 SHO and 40 VHO locations in the US

IP formulation amenable to realistic topologies!

33

Cost Comparison Across Designs

ILP design more economical than heuristic.Cost for increased reliability affordable.

Most reliable Most Reliablecost

Reduced reliability Reduced reliability

34

Summary

First work on supporting IPTV on optical mesh network with SRLG constraints

Compact Integer Programming formulation Minimum design cost SRLG-diversity shown affordable

Efficient and Scalable Algorithms

for Large Network Topologies

36

Motivation

Improve path quality Set maximum latency Limit # of intermediate nodes and links

Solving an ILP exact algorithm not scalable

Net3

37

New Heuristic Approach

Divide-and-Conquer technique for large network topologies: Partition the problem into smaller ones Solve each small problem Integrate the solutions “well”

38

Proposed Heuristics

Greedy Local (GL) Divide into subgraphs with two sources and a destination Solve for each graph, and consolidate solutions

Improved Greedy Local (IGL) Do GL and find the minimum cost graph Fix the shorter of the two paths and solve the rest

Adaptive Search Use any routing algorithm to find initial tree Find SRLG-diverse paths; for those w/o such, run baseline ILP.

Modified Active Path First Build one MST first; then for each destination, check if a SRLG-divers

e path exists. If yes, then fix the path; otherwise, run baseline ILP.

39

Greedy Local (GL)

SHOSHO

VHO

VHO

VHO

Step1: For each VHO, find redundant SRLG diverse paths by ILP

Step2: Consolidate solutions

SRLGdiverse

SRLGdiverse

SRLGdiverseConsolidate!

40

Improved Greedy Local (IGL)

SHOSHO

VHO

VHO

VHO

Step1: Run GL Step2: For each VHO, fix the shorter path Step3: Find missing paths all together using ILP

Leave onlyshorter paths

Solution from GLFind missing paths

41

Adaptive Search (AS)

SHOSHO

VHO

VHO

VHOSRLG-diverse?

Yes!Then, fix as solution.

SRLG-diverse?No!

Then, replace with SRLG diverse paths.

Step1: Use any initial routing scheme to find paths Step2: For each VHO, make sure paths are SRLG-diverse

Initial routing paths

42

Modified Active Path First (MAPF)

Step1: Find minimum spanning tree from one source Step2: For each VHO, make sure SRLG counterpart part path

exists Step3: Find the missing paths all together using ILP

SHOSHO

VHO

VHO

VHODoes SRLG-diversecounterpart path exist?

Yes!Then, fix as solution.

Does SRLG-diverse counterpart path exist?No!

Then, replace with SRLG diverse paths.

Not possible!

SRLGdiverse SRLG

diverse

Minimumspanning

tree

Find missing paths w/ ILP

43

Capital Expense Comparison

Net5 (800sec) Net6 (2sec)

44

CAPEX Scalability Analysis

Net5

45

Computation Time Analysis

Net5

46

Summary

Additional quality improvements of SRLG-diverse paths latency limits # of intermediate nodes and links per-path upper bound of SRLGs

Efficient and scalable solutions for realistic network topologies

47

Implications for Other Networks

Cross-layer optimization Optical + IP layer info combined

Topological constraints Mesh vs star WAN vs MAN

Cost constraints OXC port vs router port

48

IPTV Service Monitoring [Kerpez]

Elements of IPTV Service Assurance Subscriber management

Billing, subscriptions, AAA, DRM

Video headendConverged services, VoD, Broadcast

Transport networkIP/MPLS, Ethernet, DSLAM/OLT, Gateways

49

References

[Cha06-1] Cha et al., “Case study: resilient backbone design for IPTV services,” IPTV Workshop (WWW 2006), Edinburgh, May, 2006.

[Cha06-2] Cha et al., “Path protection routing with SRLG constraints to support IPTV in WDM mesh networks,” 9th IEEE Global Internet Symposium, Barcelona, April, 2006.

[Cha06-3] Cha et al., “Efficient and scalable provisioning solutions for always-on multicast streaming services,” (in submission).

[SRLG] Sebos et al., “Auto-discovery of shared risk link groups,” IEEE OFC, March 2001.[APF] Xu et al., “On the complexity of and algorithms for finding the shortest path with a disjoint

counterpart,” IEEE/ACM ToN, 14(1):147-158, 2006.[Kerpez] K. Kerpez et al., “IPTV Service Assurance,” IEEE Communications, September, 206