1. 1 CSE 401N:Computer Network LECTURE-14 MULTICAST ROUTING

2. 2 Chapter 4 roadmap 4.1 Introduction and Network Service Models 4.2 Routing Principles 4.3 Hierarchical Routing 4.4 The Internet (IP) Protocol 4.5 Routing in the Internet 4.6 What’s Inside a Router? 4.7 IPv6 4.8 Multicast Routing 4.9 Mobility

3. 3 Multicast: one sender to many receivers r Multicast: act of sending a datagram to multiple receivers with single “transmit” operation r Question: how to achieve multicast Multicast via unicast r source sends N unicast datagrams, one addressed to each of N receivers multicast receiver (red) not a multicast receiver (grey) routers forward unicast datagrams

4. 4 Multicast: one sender to many receivers r Question: how to achieve multicast Application-layer multicast r receivers involved in multicast copy and forward unicast datagrams among themselves

5. 5 Multicast: one sender to many receivers r Question: how to achieve multicast Network multicast r Router actively participate in multicast, making copies of packets as needed and forwarding towards multicast receivers r Only a single copy of a datagram will ever traverse a link Multicast routers (red) duplicate and forward multicast datagrams

6. 6 Multicast Groups How to identify the receivers of a multicast datagram? r Let datagram carry the IP addresses of all the recipients m Only workable with a small number of recipients m Sender needs to know the addresses of all receivers m Not a good solution! r Internet solution: use multicast group m A multicast group identifies a set of receivers and has a single identifier m Packet sent to a multicast group are received by all members in the group

7. 7 Internet Multicast Service Model r Multicast group in the Internet has its own Class D multicast address m IP addresses 224.0.0.0 to 239.255.255.255 reserved for multicast (IP addresses begin with 1110) 128.119.40.186 128.59.16.12 128.34.108.63 128.34.108.60 multicast group 226.17.30.197

8. 8 Internet Multicast Service Model  Host group semantics: oanyone can join/leave a multicast group oanyone can send to a multicast group ono network-layer identification of hosts that have joined a multicast group r Questions to answer: m How are hosts join/leave a group? A: Internet Group Management Protocol (IGMP) m How do the routers deliver a multicast datagram to all group members? A: multicast routing protocols

9. 9 IP multicasting: two-step process r local: host informs local router of desire to join a multicast group: IGMP r wide area: local router interacts with other routers to receive multicast datagrams m many multicast routing protocols (e.g., DVMRP, MOSPF, PIM) IGMP wide-area multicast routing

10. 10 Internet Group Management Protocol (IGMP) r RFC 2236 r Operate between a host and its directly attached router r Provide the means for: m A host to inform its attached router that it wants to join/leave a multicast group m A router to detect if a LAN has any members for a particular multicast group

11. 11 IGMP Messages Three message types: membership query, membership report, leave group r General membership query  Sent by a router to all nodes on the LAN m Query multicast groups joined by attached hosts m Multicast group address field set to 0 r Specific membership query m Sent by a router to all nodes on the LAN m Query if a specific multicast group is joined by attached hosts m Multicast group address field set to the multicast group address being queried Type Max. Resp. Time Checksum Multicast group address

12. 12 IGMP Messages (cont’d) r Membership report message m Sent by host to all nodes on the LAN in response to a membership query message or when first joins a multicast group m Contains the address of a group the host has joined r A router only cares about whether one or more hosts belong to a given group  want to hear from only one of the hosts that belongs to each group m Each membership query message includes a “maximum response time” field m A host waits a random time between 0 and maximum response time before sending a membership report message m Feedback suppression: if the host observes a membership report message from some other host for the multicast group, discard its own pending membership report message

13. 13 IGMP Messages (cont’d) r Leave group message: m sent by host to report leaving a group m leave group message is optional Router periodically sends a membership query message A router infers that no hosts are joined to a group when no host responds to the query message Soft state protocol: state is removed via a timeout event if it is not explicitly refreshed r IGMP messages are encapsulated within an IP datagram r There is no control over who sends to the group and who joins the group

14. Multicast Routing: Problem Statement r Goal: Given a multicast group, find a tree connecting routers having local multicast group members m shared tree: a single tree used by all group members m source-based tree: different tree for each sender Shared tree Source-based trees

15. Approaches for building mcast trees r group-shared tree: group uses one tree m minimal cost (Steiner) tree m center-based tree r source-based tree: one tree per source m shortest path tree m reverse path forwarding …we first look at basic approaches, then specific protocols adopting these approaches

16. Group-Shared Tree: Steiner Tree r All multicast packets are routed along the same tree, regardless of the sender r Steiner Tree: minimum cost tree connecting all routers with attached group members m problem is NP-complete m good heuristics exist r not used in practice: m information about all links needed m rerun whenever link costs change m not able to use unicast routing info Router having attached group members Router having no attached group members 4 1 1 1 2 22 3

17. Group-Shared Tree: Center-based tree r single delivery tree shared by all r one router identified as center of tree r to join: m router with attached group members sends unicast join message addressed to the center m join message “processed” by intermediate routers and forwarded towards center m join message either hits existing tree branch for this center, or arrives at center m path taken by join message becomes new branch of tree for this router R1 R2 R3 R4 R5 R6 R7 2 3 1 router with attached group member router with no attached group member path order in which join messages generated 1 center

18. 18 Multicast Routing Using a Source-Based Tree r Construct a multicast routing tree for each source in the multicast group r Router maintains state for each (group, sender) pair r Shortest path tree: union of the shortest paths from source to each receiver m Obtained by Dijkstra’s algorithm Router having attached group members Router having no attached group members A B D E C F

19. 19 Steiner Tree v.s Shortest Path Tree 4 1 1 1 2 22 3 4 1 1 1 2 22 3 A A (source) Steiner tree: minimize the sum of the link costs in the tree Shortest path tree: minimize the cost from source to each receiver

20. 20 Compute Shortest Path Tree r Use link state algorithm: each router needs to know the state of each link in the network r Can we compute the tree without link state information? A: use reverse path forwarding algorithm

21. 21 Reverse Path Forwarding (RPF) r Forward packet from source S to all interfaces (except the incoming one) iff the packet arrived on the link that is on its shortest path to S m A router need only know the next hop on its shortest path to the sender r result is a source-specific reverse shortest path tree A B D E C F Packet that will be forwarded Packet not forwarded beyond receiving router G

22. 22 RPF: Pruning r forwarding tree contains subtrees with no mcast group members m no need to forward datagrams down subtree m “prune” msgs sent upstream by router with no downstream group members Router tells parent in the tree to stop forwarding Prune message propagate upstream A B D E C F G Prune message

23. 23 Rejoining If a router sends a prune message upstream, what should happen if it later needs to join the group? r Send a graft message to upstream router r Or, allow pruned branches to time-out and be added again to the tree

24. 24 Issues in Wide-Area Multicast r Hosts may join and leave a multicast group dynamically  need to dynamically update routing tree r State information for a multicast connection must be established and maintained in routers m Group-shared tree: per group state m Source-based tree: per (group, source) state

25. Internet Multicasting Routing: DVMRP r DVMRP: distance vector multicast routing protocol, RFC1075 r flood and prune: source-based tree, reverse path forwarding m each router uses a distance vector algorithm to compute the next hop on its shortest path to each source m initial datagram to mcast group flooded everywhere via RPF m routers not wanting multicast datagram: send upstream prune messages m routers store state for each (source, group) pair

26. DVMRP: continued… r soft state: DVMRP router periodically “forgets” branches are pruned: m multicast data again flows down unpruned branch m router: reprune or else continue to receive data r routers can quickly regraft to tree using graft message r odds and ends m commonly implemented in commercial routers m MBONE (Multicast Backbone) routing done using DVMRP

27. Tunneling r some Internet routers are not multicast-capable r Internet MBONE: a virtual network of multicast-capable routers on top of a physical network containing a mix of unicast and multicast routers  mcast datagram encapsulated inside unicast datagram  unicast IP datagram sent thru “tunnel” via regular IP unicast to receiving mcast router  receiving mcast router extracts mcast datagram physical topology logical topology A B C A B C

28. PIM: Protocol Independent Multicast r not dependent on any particular underlying unicast routing protocol (works with all) r two different multicast distribution scenarios: Dense mode:  group members densely located, many routers need to be involved in routing multicast datagrams Sparse mode:  # routers with attached group members small wrt total # routers

29. Consequences of Sparse-Dense Dichotomy Dense mode r group membership by routers assumed until routers explicitly prune r data-driven construction of mcast tree (e.g., RPF) Sparse mode r routers not involved in multicast distribution until explicitly join a group r receiver- driven construction of mcast tree (e.g., center-based)

30. PIM - Dense Mode  flood-and-prune reverse path forwarding  similar to DVMRP, but can interoperate with any underlying unicast routing protocol

31. PIM - Sparse Mode r center-based approach r router sends JOIN msg to rendezvous point (RP) to join the tree m intermediate routers update state and forward JOIN message toward RP  PRUNE message is sent when a router leaves a group R1 R2 R3 R4 R5 R6 R7 join rendezvous point

32. PIM - Sparse Mode r sender unicast data to RP, which multicasts down RP- rooted tree r RP can send STOP message to source if no routers are jointed to the tree m “no one is listening!” r after joining RP, router can switch to source-specific tree R1 R2 R3 R4 R5 R6 R7 join all data multicast from rendezvous point rendezvous point

33. 33 Switch from group-shared tree to a source-specific tree r In group-shared tree, E sends to RP, then RP sends to A r when data rate from E exceeds a threshold m A sends join message directly to E m A does not send prune message towards RP so it can continue to receive from other sources r increased performance: less traffic concentration, shorter paths source receiver RP Before switch After switch ABC E D

34. 34 MOSPF: Multicast Open Shortest Path First r Multicast extension to OSPF r Routers flood group membership information with Link State Advertisements r Each router independently computes source- specific, shortest-path trees for each multicast group

35. 35 Evaluating a Multicast Routing Protocol r Scalability: the amount of state required in the routers m Group-shared tree: per group state m Source-based tree: per group, per source state r Excess traffic received? m Yes: DVMRP, PIM Dense Mode m No: MOSPF, PIM Sparse Mode r Traffic Concentration m Group-shared tree tends to concentrate traffic on a smaller number of links m Source-based trees tend to distribute multicast traffic more evenly r Optimality of forwarding paths m Minimum cost multicast tree not used in practice m Shortest path trees and center based trees are used

36. 36 THANX.