Multicast Routing

• Unicast: one source to one destination

15-441 Computer Networking
• Multicast: one source to many destinations
Lecture 11 – Multicast
• Main goal: efficient data distribution

Lecture 11: 10-3-2006 2

Overview Multicast – Efficient Data Distribution

• IP Multicast Service Basics Src Src

• Host/Router Interaction

• MOSPF/DVMRP

• Overlay Multicast

Lecture 11: 10-3-2006 3 Lecture 11: 10-3-2006 4

1
Example Applications IP Multicast Architecture

• Broadcast audio/video Service model Hosts

• Push-based systems
• Software distribution Host-to-router protocol
• Web-cache updates (IGMP)
Routers
• Teleconferencing (audio, video, shared
whiteboard, text editor)
• Multi-player games Multicast routing protocols
• Server/service location (various)
• Other distributed applications

Lecture 11: 10-3-2006 5 Lecture 11: 10-3-2006 6

Logical Naming Multicast Router Responsibilities

• Learn of the existence of multicast groups

• Single name/address maps to logically related set (through advertisement)
of destinations • Identify links with group members
• Destination set = multicast group • Establish state to route packets
• Replicate packets on appropriate interfaces
• Key challenge: scalability • Routing entry:
• Single name/address independent of group growth or
changes Src, incoming interface List of outgoing interfaces

Lecture 11: 10-3-2006 7 Lecture 11: 10-3-2006 8

2
IP Multicast Service Model (rfc1112) IP Multicast Addresses

• Each group identified by a single IP address • Class D IP addresses

• Groups may be of any size • 224.0.0.0 – 239.255.255.255
• Members of groups may be located anywhere in the
1 110 Group ID
Internet
• Members of groups can join and leave at will • How to allocated these addresses?
• Senders need not be members • Well-known multicast addresses, assigned by IANA
• Group membership not known explicitly
• Transient multicast addresses, assigned and reclaimed
• Analogy: dynamically, e.g., by “sdr” program
• Each multicast address is like a radio frequency, on which anyone
can transmit, and to which anyone can tune-in.

Lecture 11: 10-3-2006 9 Lecture 11: 10-3-2006 10

IP Multicast API Multicast Scope Control – Small TTLs

• Sending – same as before • TTL expanding-ring search to reach or find a

• Receiving – two new operations nearby subset of a group
• Join-IP-Multicast-Group(group-address, interface)
• Leave-IP-Multicast-Group(group-address, interface) s
• Receive multicast packets for joined groups via normal 1
IP-Receive operation
• Implemented using socket options 2

Lecture 11: 10-3-2006 11 Lecture 11: 10-3-2006 12

3
Multicast Scope Control – Large TTLs Overview

• Administrative TTL Boundaries to keep multicast traffic • IP Multicast Service Basics

within an administrative domain, e.g., for privacy or
resource reasons
• Host/Router Interaction
The rest of the Internet

• MOSPF/DVMRP

TTL threshold set on • Overlay Multicast

interfaces to these links,
greater than the diameter
An administrative domain of the admin. domain

Lecture 11: 10-3-2006 13 Lecture 11: 10-3-2006 14

IP Multicast Architecture Internet Group Management Protocol

Service model • End system to router protocol is IGMP

Hosts
• Each host keeps track of which mcast groups are
Host-to-router protocol subscribed to
(IGMP) • Socket API informs IGMP process of all joins
Routers
• Objective is to keep router up-to-date with group
membership of entire LAN
Multicast routing protocols • Routers need not know who all the members are, only
(various) that members exist

Lecture 11: 10-3-2006 15 Lecture 11: 10-3-2006 16

4
 How IGMP Works How IGMP Works (cont.)

Routers: Q Routers: Q

Hosts: Hosts: G G G G

• On each link, one router is elected the “querier” • When a host’s timer for group G expires, it sends a Membership
• Querier periodically sends a Membership Query message to the Report to group G, with TTL = 1
all-systems group (224.0.0.1), with TTL = 1
• Other members of G hear the report and stop their timers
• On receipt, hosts start random timers (between 0 and 10
seconds) for each multicast group to which they belong • Routers hear all reports, and time out non-responding groups

Lecture 11: 10-3-2006 17 Lecture 11: 10-3-2006 18

How IGMP Works (cont.) Overview

• Note that, in normal case, only one report • IP Multicast Service Basics
message per group present is sent in response
to a query • Host/Router Interaction
• Power of randomization + suppression

• Query interval is typically 60-90 seconds • MOSPF/DVMRP

• When a host first joins a group, it sends one or • Overlay Multicast

two immediate reports, instead of waiting for a
query

Lecture 11: 10-3-2006 19 Lecture 11: 10-3-2006 20

5
IP Multicast Architecture Routing Techniques

• Basic objective – build distribution tree for multicast

Service model packets
Hosts
• Flood and prune
Host-to-router protocol • Begin by flooding traffic to entire network
(IGMP) • Prune branches with no receivers
Routers • Examples: DVMRP, PIM-DM
• Unwanted state where there are no receivers
• Link-state multicast protocols
• Routers advertise groups for which they have receivers to entire
Multicast routing protocols network
(various) • Compute trees on demand
• Example: MOSPF
• Unwanted state where there are no senders

Lecture 11: 10-3-2006 21 Lecture 11: 10-3-2006 22

Multicast OSPF (MOSPF) Example

• Add-on to OSPF (Open Shortest-Path First,

a link-state, intra-domain routing protocol) Source 1
Z
• Multicast-capable routers flag link state routing
advertisements
W
• Link-state packets include multicast group
addresses to which local members have joined Q

• Routing algorithm augmented to compute

shortest-path distribution tree from a source to any T
set of destinations Receiver 1

Receiver 2

Lecture 11: 10-3-2006 23 Lecture 11: 10-3-2006 24

6
Link Failure/Topology Change Membership Change

Source 1 Source 1
Z Z
Receiver 3

W W

Q Q

T T

Receiver 1 Receiver 1

Receiver 2 Receiver 2

Lecture 11: 10-3-2006 25 Lecture 11: 10-3-2006 26

Impact on Route Computation Distance-Vector Multicast Routing

• Can’t pre-compute multicast trees for all possible • DVMRP consists of two major components:
sources • A conventional distance-vector routing protocol (like
RIP)
• Compute on demand when first packet from a • A protocol for determining how to forward multicast
source S to a group G arrives packets, based on the routing table
• New link-state advertisement • DVMRP router forwards a packet if
• May lead to addition or deletion of outgoing interfaces if • The packet arrived from the link used to reach the
it contains different group addresses source of the packet (reverse path forwarding check –
RPF)
• May lead to re-computation of entire tree if links are
changed • If downstream links have not pruned the tree

Lecture 11: 10-3-2006 27 Lecture 11: 10-3-2006 28

7
Example Topology Broadcast with Truncation

G G G G

S S

G G
Lecture 11: 10-3-2006 29 Lecture 11: 10-3-2006 30

Prune Graft

G G G G
G
Report (g)

Prune (s,g) Graft (s,g)

S Prune (s,g) S Graft (s,g)

G G
Lecture 11: 10-3-2006 31 Lecture 11: 10-3-2006 32

8
Steady State Overview

G G • IP Multicast Service Basics

G

• Host/Router Interaction

• MOSPF/DVMRP
S

• Overlay Multicast

G
Lecture 11: 10-3-2006 33 Lecture 11: 10-3-2006 34

Failure of IP Multicast Supporting Multicast on the Internet

• Not widely deployed even after 15 years!

• Use carefully – e.g., on LAN or campus, rarely over
WAN Application
? At which layer should
• Various failings
• Scalability of routing protocols multicast be implemented?
• Hard to manage IP ?
• Hard to implement TCP equivalent
• Hard to get applications to use IP Multicast without
existing wide deployment
Network
• Hard to get router vendors to support functionality and
hard to get ISPs to configure routers to enable
Internet architecture
Lecture 11: 10-3-2006 35 Lecture 11: 10-3-2006 36

9
IP Multicast End System Multicast

MIT1
MIT MIT
Berkeley Berkeley
MIT2
UCSD UCSD

CMU1
CMU CMU
routers
end systems CMU2
multicast flow Berkeley MIT1
Overlay Tree
• Highly efficient MIT2
UCSD
• Good delay
CMU1 CMU2
Lecture 11: 10-3-2006 37 Lecture 11: 10-3-2006 38

Potential Benefits Over IP Multicast Concerns with End System Multicast

• Quick deployment • Self-organize recipients into multicast delivery overlay

tree
• All multicast state in end systems • Must be closely matched to real network topology to be efficient
• Computation at forwarding points simplifies • Performance concerns compared to IP Multicast
• Increase in delay
support for higher level functionality • Bandwidth waste (packet duplication)
MIT1
• Penalty can be kept small in practice
MIT
Berkeley Berkeley MIT1 MIT1
Berkeley
MIT2
UCSD UCSD MIT2 UCSD MIT2
CMU1
CMU1 CMU1

CMU IP Multicast CMU2 End System Multicast

CMU2
CMU2
Lecture 11: 10-3-2006 39 Lecture 11: 10-3-2006 40

10
Important Concepts Next Lecture: Wide Area Routing

• Multicast provides support for efficient data

delivery to multiple recipients • How to make routing scale to the size of the
• Requirements for IP Multicast routing Internet
• Keeping track of interested parties • How to accommodate business relationships in
• Building distribution tree routing
• Broadcast/suppression technique
• Difficult to deploy new IP-layer functionality
• End system-based techniques can provide similar
efficiency
• Easier to deploy

Lecture 11: 10-3-2006 41 Lecture 11: 10-3-2006 42

Routing Techniques

• Core based protocols

• Specify “meeting place” aka core
EXTRA SLIDES • Sources send initial packets to core
• Receivers join group at core
The rest of the slides are FYI • Requires mapping between multicast group address
and “meeting place”
• Examples: CBT, PIM-SM

Lecture 11: 10-3-2006 44

11
Shared vs. Source-based Trees Source-based Trees

• Source-based trees Router

S Source
• Separate shortest path tree for each sender
R Receiver
• DVMRP, MOSPF, PIM-DM, PIM-SM R
• Shared trees R

• Single tree shared by all members

• Data flows on same tree regardless of sender R
S
• CBT, PIM-SM
S

Lecture 11: 10-3-2006 45 Lecture 11: 10-3-2006 46

Shared Tree Shared vs. Source-Based Trees

Router • Source-based trees

S Source • Shortest path trees – low delay, better load distribution
R Receiver • More state at routers (per-source state)
R • Efficient for in dense-area multicast
R • Shared trees
• Higher delay (bounded by factor of 2), traffic concentration
RP • Choice of core affects efficiency
R • Per-group state at routers
S
• Efficient for sparse-area multicast
S
• Which is better? Æ extra state in routers is bad!

Lecture 11: 10-3-2006 47 Lecture 11: 10-3-2006 48

12
Multicast Backbone (MBone) MBone Tunnels

• An overlay network of IP multicast-capable routers

• A method for sending multicast packets through multicast-
ignorant routers
R R
• IP multicast packet is encapsulated in a unicast packet
addressed to far end of tunnel:

R H IP header, IP header, Transport header

dest = unicast dest = multicast and data…

R Host/router • Tunnel acts like a virtual point-to-point link

H R H R MBone router • Each end of tunnel is manually configured with unicast
Physical link address of the other end
Tunnel
Part of MBone

Lecture 11: 10-3-2006 49 Lecture 11: 10-3-2006 50

Link-Layer Transmission/Reception Using Link-Layer Multicast Addresses

• Transmission • Ethernet and other LANs using 802 addresses:

• IP multicast packet is transmitted as a link-layer multicast, on those
links that support multicast
• Link-layer destination address is determined by an algorithm IP multicast address
specific to the type of link 1110 28 bits
• Reception
• Necessary steps are taken to receive desired multicasts on a Group bit
particular link, such as modifying address reception filters on LAN
0000000100000000010111100 23 bits
interfaces
• Multicast routers must be able to receive all IP multicasts on a link, LAN multicast address
without knowing in advance which groups will be used • No mapping needed for point-to-point links

Lecture 11: 10-3-2006 51 Lecture 11: 10-3-2006 52

13