Multicast Routing

1
Multicast Routing Protocols

• Acute Communication Corp.
• Edward Jin-Ru Chen
• jzchen_at_acutecomm.com.tw

2
Content

• Introduction
• Algorithms
• Flooding, Spanning Tree, Reverse Path Broadcast,
  Reverse Path Multicast and Core Based Tree
• Routing Protocols
• DVMRP, MOSPF, PIM-SM, PIM-DM and CBT
• Testing Issue

3
Why Multicast Routing?

• Reduce the bandwidth waste

D
D
S
S
D
D
Broadcast
Multicast
4
Sender-base VS Receiver-base

• Whether the sender or the receiver establishes
  the connections necessary for forming a multicast
  group
• Sender-Based
• Suitable for small group
• Hard to extend to large or dynamic groups
• Receiver-Based
• Source no need to maintain receiver lists
• Easier to make group dynamic

5
Sender-based multicasting
Conformation
Source
Receivers
Join requests
6
Receiver-based multicasting
Data
Data
Data
Receivers
Source
Join requests
7
Multicast Backbone

• MBONE is an interconnected set of subnetworks and
  routers that support the delivery of IP multicast
  traffic
• Start from 40 subnets in four different countries
  in 1992
• Connect each island through virtual
  point-to-point links called tunnels

8
IP Multicast Address

• Class D address
• Address assignment
• 224.0.0.0 is reserved and cannot be assigned to
  any group
• 224.0.0.1 to 224.0.0.255 is reserved for the use
  of routing protocols and other low-level topology
  discovery or maintenance protocol
• 224.0.1.0 to 239.255.255.255 are assigned to
  various multicast applications
• 239.0.0.0 to 239.255.255.255 are to be reserved
  for site-local administratively scoped
  applications

9
Multicast Path Decision

• Source to router
• Transmit to direct connected router
• Router to router
• Base on multicast routing protocol
• Such as DVMRP, MOSPF ...
• Router to destinations
• Base on local membership register protocol
• Such as IGMP (Version1, 2, 3)

10
Multicast Routing Concept

• Class of routing protocols
• Dense - lots of receivers nearby
• Sparse - few receivers spread out
• Delivery Trees can be
• Source- rooted (one per sending host) or,
• Shared one delivery tree that all senders use
• Establishment of a delivery tree triggered when
• Multicast data sent out from the sender or
• Receivers joining a multicast group

11
Source-Rooted Trees
12
Source-Rooted Trees (II)

• Source- rooted delivery tree built for each
  (Source, Group Address) multicast pair
• also called a shortest- path tree (SPT)
• Routers must maintain O( S x G) state
• Better performance and optimal path from sender
  to receivers via shortest- path trees (SPT) but
  routers have to maintain more information

13
Shared Trees
14
Shared Trees (II)

• Single delivery tree (per group) is built rooted
  at Core
• Routers must maintain O( G) state
• Routers must maintain less information but Core
  performance, locality and control could be issues

15
Dense Mode Protocol

• Group membership is dense
• Data- driven from Source
• SPT is not built until multicast data begins to
  flow from sender
• packets are forwarded to all subnets until
  routers prune interfaces/ links
• Use Flood and Prune
• Suitable for campus LANs and corporate intranets

16
Sparse Mode Protocol

• Group membership is sparse
• Flood and Prune wastes bandwidth and resources
• Better to use shared tree approach
• Data is not sent from the Core/ RP router to a
  downstream router unless that router joins the
  shared tree (by sending an explicit JOIN towards
  the CORE/ RP)
• Suitable for large corporate intranets and the
  Internet

17
Switch VS Protocol

• Cooperation between protocol engine and
  forwarding engine
• Protocol engine
• Used to be real-time OS working over processor
• Forwarding engine
• ASIC
• FPGA
• ASIC with embedded micro-processor

18
Switch VS Protocol (II)

• Protocol engine decide the forwarding ports and
  tell the forwarding engine
• Forwarding engine just follow the known
  information to forwarding the multicast packet
• Based on exactly match to do address search

19
Multicast Forwarding Algorithm

• Flooding
• Spanning Trees
• Reverse Path Broadcasting (RPB)
• Truncated Reverse Path Broadcasting (TRPB)
• Reverse Path Multicasting (RPM)
• Core-Based Trees (CBT)

20
Flooding Algorithm

• Simplest technique
• Flood the received packet to interfaces other
  than the receiving interface
• Avoid the forwarding of looped packet

21
Spanning Tree Algorithm

• What is the spanning tree?
• A subset of the Internet topology that forms a
  loop free tree
• Forward the received packet to other interfaces
  belong to the spanning tree
• Filter out the received packet when received from
  interfaces not belong to the spanning tree
• No loop free mechanism required
• Powerful and relatively easy, but centralize
  packet flow and path is not the most efficient

22
Reverse Path Broadcasting

• Build a group-specific spanning tree for each
  potential source
• Construct source-rooted delivery trees
• Operation
• For each packet arrives on a link that the router
  considers to be the shortest patch back to the
  source, forward the packet on all interfaces
  except the incoming interface

23
RPB (II)

• Reverse Path Broadcasting - Forwarding Algorithm

24
Enhanced RPB

• Enhanced to reduce unnecessary packet duplication
  based on whether the neighboring router is on the
  shortest path back to the source
• Intuitive on link-state protocol
• Use Poison Reverse on distance-vector protocol

A
B
C
D
F
E
25
Truncated Reverse Path Broadcasting

• To overcome the limitations of Reverse Path
  Broadcasting
• Use IGMP to determine the group memberships on
  each leaf
• The spanning delivery tree is truncated by the
  router if a leaf does not have group members

26
TRPB (II)

• Examples

27
Reverse Path Multicasting

• RPM creates a delivery tree that spans only
• Subnetworks with group members
• Routers and subnetworks along the shortest path
  to subnetworks with group members

28
RPM (II)

• Operation of RPM

29
RPM (III)

• Receive packet for (source, group), the first one
  is forwarded as TRPB
• Router at the edge of the network is called leaf
  router
• If the group member is on the leaf network
• Forward packet based on IGMP information
• If no subnetwork has group member
• Transmit a prune message on its parent link
• Prune messages are sent only one hop back
• When all leaf are pruned, parent node sends prune
  message

30
RPM (IV)

• The membership and network topology are
  dynamically changed
• Remove the pruned branch periodically
• Limitations
• Periodically forward multicast packet to all
  routers
• States of each source and group pair must be
  maintained

31
Core-Based Tree (CBT)

• Construct a single delivery tree that is shared
  by all members of a group
• Different group can have different tree

32
CBT (II)

• Construct a CBT backbone
• Join the group must send join message to core
  tree
• Potential group member only need to know the
  address of the groups core router
• Unicast join request
• By passed router record the passed interface to
  construct the branch of core tree

33
CBT (III)

• Benefits
• Only need one router to record state information
  for each group
• No periodic flooding and pruning requirement
• Limitation
• Traffic concentration and bottlenecks near core
  routers
• May create suboptimal routes
• Need to manage the selection and replacement of
  cores

34
Distance Vector Multicast Routing Protocol (DVMRP)

• Distance-vector routing protocol
• Origin defined in RFC1075
• Derived from RIP
• Use Truncated Reverse Path Broadcasting algorithm
• Difference from RIP
• RIP calculates the next hop to a destination
• DVMRP calculate the previous hop back to a source
• Version 3.8 and vender implementations employ
  Reverse Path Multicasting

35
DVMRP Tunneling

• Supports tunneling of multicast packets through
  non- multicast routers
• IP over IP encapsulation
• Enables IP Multicast islands to communicate
• Contributes to increased network utilization
• DVMRP tunnels are manually configured between
  end-points

36
DVMRP Operation
37
DVMRP Scope Control

• Use TTL value to limit the scope of transmission

38
DVMRP Operation

• First datagram for any (source, group) pair is
  forwarded to entire internetwork
• The prune messages result in the removal of
  branches from the tree
• After a period of time, the pruned branch grow
  back and the next datagram flood again
• If a pruned branch has newly added group member,
  the graft message is send

39
DVMRP Routing Table
40
DVMRP Forwarding Table
41
Multicast Extensions to OSPF (MOSPF)

• Multicast Extensions to OSPF
• not actually a separate multicast routing
  protocol
• Leverages already existing OSPF topology database
  to compute source-rooted shortest path delivery
  tree
• Defines Group Membership LSA that is added to
  topology database and propagated throughout OSPF
  AS

42
MOSPF (II)

• Generates a source- rooted delivery tree that
  branches out to all members-only subnetworks
• calculates as soon as first multicast packet hits
  the router
• because MOSPF routers maintain the same topology
  database in an area, they will calculate the same
  delivery tree
• RFC1584

43
MOSPF Characteristics

• Not supported on all routers
• MOSPF can interoperate with OSPF to support
  unicast but MOSPF does not support tunnels
• Does not support equal cost paths

44
MOSPF Characteristics(II)

• Advantages
• SPT is calculated "in memory" thus initial
  multicast packets are not flooded
• leverages existing OSPF function
• Disadvantages
• may require router CPU to calculate each SPT as
  number of senders increases
• Group Membership LSAs are propagated throughout
  network

45
MOSPF Inter-Area routing
46
Inter-Area Routing (II)

• Area border routers are configured as wild- card
  multicast receivers (WCMR) and inter- area
  multicast forwarders (IAMF)
• WCMR receives all multicast packets and forwards
  to the next area - must not be pruned from the
  source- rooted SPT
• IAMF forwards group membership from non- backbone
  areas to the backbone
• Inter- area delivery trees built using
  information supplied by inter- area multicast
  forwarders and summary- links LSAs

47
Protocol Independent Multicast (PIM)

• Supports both a Dense Mode and Sparse Mode
  function
• Dense mode to operate under both group members
  are relatively densely packed and bandwidth is
  plentiful
• Sparse mode to operate under group members are
  widely dispersed across the Internet
• Independent of any underlying unicast routing
  protocol
• Require a unicast routing protocol to provide
  routing table information and to adapt to
  topology changes

48
PIM (II)

• Design objectives
• provide scalable inter-domain multicast routing
• efficient sparse group support
• shared- tree or SPT support
• routing protocol independent
• dense mode support
• robustness - uses soft- state
• interoperability with other multicast routing
  protocols

49
PIM - Dense Mode

• Designed for Dense multicast regions
• Uses Reverse Path Multicasting
• Default operation is to send
• implicit join (everybody is a member) with
  explicit prunes
• Simple to implement

50
PIM - DM (II)

• Differs from DVMRP in the following
• forwards out all interfaces until explicit
  pruning or truncating occurs DVMRP uses parent-
  child relationship, uses split- horizon to
  recognize downstream routers
• independent of specific unicast routing
  technology but rely on unicast routing protocol
  to adapt to topology changes DVMRP uses distance
  vectors
• does not support tunnels DVMRP does

51
PIM - Sparse Mode

• Designed for Sparse multicast regions
• RFC2362
• Basic concept is that multicast data is blocked
  unless a downstream router explicitly asks for it
• controls the amount of traffic that traverses
  network
• Similar to the Core-Based Tree (CBT) approach
• Employs the concept of a rendezvous point (RP)
  where receivers meet sources
• Only one RP for each multicast group

52
PIM - SP Rendezvous Point

• Rendezvous Point (RP)
• RP-list contains a primary RP and a small ordered
  set of alternative RPs
• single active RP per group but backups can exist
• Routers need to maintain RP-list before the
  arrival of data packets
• senders must register their existence with RP
• receivers must join RP- rooted tree

53
PIM - SP RP (II)

• Shared RP tree is initially used to distribute
  data but PIM routers with members may switch over
  to Source-Rooted Shortest Path Tree
• for better network performance and utilization

54
PIM - SM Operation

• PIM Routers locate RP via Bootstrap Mechanism
• downstream PIM routers need to know address of RP
  so that they can send PIM JOIN messages
• PIM Routers with local members Join group-
  specific multicast tree rooted at RP
• PIM Routers with local senders encapsulate data
  and send it (unicast) to RP
• PIM Routers with local members may initiate and
  switch-over to source-rooted shortest-path tree

55
PIM Register

• PIM Routers with local senders encapsulate
  multicast packets in PIM- Register messages and
  forward to RP
• RP de-encapsulates messages and forwards out
  multicast delivery tree
• RP may option to send PIM JOIN and PIM
  REGISTER-STOP messages to source PIM router. This
  enables source PIM router to send native
  multicast packets to RP

56
PIM Register (II)
57
PIM Join - SPT

• Based on a threshold, PIM routers with local
  members send PIM Join messages towards the source
• Source-rooted shortest path tree is built
• PIM routers with local members continue to
  receive packets from both the source and the RP

58
PIM Join - SPT (II)
59
PIM Shared Tree to SPT Switch

• After PIM routers receive multicast packets from
  both the SPT and RP delivery trees, PIM Prune
  messages are sent upstream towards the RP to
  prune the RP-tree. Now multicast data from the
  Source will only flow over the source- rooted SPT
  towards group members.

60
PIM Shared Tree to SPT Switch (II)
61
Core Base Tree

• RFC 2189
• CBT protocol builds and maintains a shared
  multicast distribution tree that spans only those
  networks and links leading to interested
  receivers.
• Core router is a "meeting point" between a sender
  and group receivers
• An on-tree router, which is part of a CBT
  distribution tree, maintains active state for the
  group

62
CBT Operation
Join_Request
Join_ACK
S
Data Flow
Core
R
R
63
CBT Receiver Operation

• Use IGMP to register membership of a group
• CBT router generates JOIN_REQUEST message to next
  hub to the core router
• Acknowledge message (JOIN_ACK) is sent from
• the core router
• another on-the-path router which itself has
  already joined the tree
• The host is on-tree after receiving JOIN_ACK

64
CBT Sender Operation

• Non-member sender data is encapsulated (IP-in-IP)
  by the first-hop router, and is unicast to the
  group's core router

65
Protocol Verification

• Packet Format
• Lower protocol parameter setting
• Entry field validity
• Timer
• Preciseness of each timer
• Algorithm
• Using entered packet to generate virtual
  environment to trigger algorithm calculation and
  performs respected packet forwarding

66
Protocol Verification

• Input process
• Check the processing result of different input
  packets
• Output process
• Check the processing result when router generate
  packets

67
Protocol Verification

• Cooperation between unicast routing protocol
• Such as MOSPF, PIM
• The information exchange must be correct
• Interoperability
• Using routers produced by different venders,
  connect them and check the working result
• Using different interface setting and multicast
  member register to verify the interoperability

68
Verification Example

• DVMRP needs to send out routing information every
  60 sec
• Test procedure
• Send in routing information
• Listen for 60 sec
• The DUT must send out routing information
  contains the same information

69
Benchmark

• RFC2432
• Benchmark items
• Multicast Speedup Index (MSI)
• Multicast Latency (ML)
• Group Join Delay (GJD)
• Group Leave Delay (GLD)