1. Katz, Stoica F04 EECS 122: Introduction to Computer Networks Multicast Computer Science Division Department of Electrical Engineering and Computer Sciences University of California, Berkeley Berkeley, CA 94720-1776

2. Katz, Stoica F04 2 Today’s Lecture: 16 Network (IP) Application Transport Link Physical 2 7, 8, 9 10,11 17, 18, 19 14, 15, 16 21, 22, 23 25 6

3. Katz, Stoica F04 3 Motivation Example: Internet Radio  www.digitallyimported.com (techno station) www.digitallyimported.com -Sends out 128Kb/s MP3 music streams -Peak usage ~9000 simultaneous streams only 5 unique streams (trance, hard trance, hard house, eurodance, classical) -Consumes ~1.1Gb/s bandwidth costs are large fraction of their expenditures (maybe 50%?) -If 1000 people are getting their groove on in Berkeley, 1000 unicast streams are sent from NYC to Berkeley

4. Katz, Stoica F04 4 This approach does not scale… Backbone ISP Broadcast Center

5. Katz, Stoica F04 5 Instead build trees Backbone ISP Broadcast Center Copy data at routers At most one copy of a data packet per link LANs implement link layer multicast by broadcasting Routers keep track of groups in real-time Routers compute trees and forward packets along them

6. Katz, Stoica F04 6 Multicast Routing Approaches  Kinds of Trees -Source Specific Trees -Shared Tree  Tree Computation Methods -Link state -Distance vector

7. Katz, Stoica F04 7 Source Specific Trees 6 7 8 5 4 3 1 2 12 10 13 11 Each source is the root of its own tree One tree per source Tree can consists of shortest paths to each receiver Members of the multicast treeSender

8. Katz, Stoica F04 8 Source Specific Trees 6 7 8 5 4 3 1 2 12 10 13 11 Very good performance but expensive to construct/maintain; routers need to manage a tree per source Each source is the root of its own tree One tree per source Tree can consists of shortest paths to each receiver

9. Katz, Stoica F04 9 Shared Tree 6 7 8 5 4 3 1 2 12 10 13 11 One tree used by all members in a group Easier to construct/maintain but hard to pick “good” trees for everyone!

10. Katz, Stoica F04 10 Shared Tree  Ideally, find a Steiner tree - the minimum-weighted tree connecting only the multicast members 6 7 8 5 4 3 1 2 12 10 13 11 12 2 3 21 2 2 11 15 12 2 1 2 2 7 2 2

11. Katz, Stoica F04 11 Shared Tree 6 7 8 5 4 3 1 2 12 10 13 11 12 2 3 21 2 2 11 15 12 2 1 2 2 7 2 2  Ideally, find a Steiner tree – minimum-weighted tree connecting only the multicast members  Finding Steiner Tree is NP hard  Heuristics are known

12. Katz, Stoica F04 12 Shared Tree 6 7 8 5 4 3 1 2 12 10 13 11 12 2 3 21 2 2 11 15 12 2 1 2 2 7 2 2  Alternatively, find a minimum-spanning tree – minimum-weighted tree connecting all nodes in the network  Finding a minimum spanning tree is much easier

13. Katz, Stoica F04 13 Shared Tree 6 7 8 5 4 3 1 2 12 10 13 11 12 2 3 21 2 2 11 15 12 2 1 2 2 7 2 2  Alternatively, find a minimum-spanning tree – minimum-weighted tree connecting all nodes in the network  Finding a minimum spanning tree is much easier. How?

14. Katz, Stoica F04 14 Shared Tree  Alternatively, find a minimum-spanning tree – minimum-weighted tree connecting all nodes in the network  Finding a minimum spanning tree is easier. How?  Prune back to get multicast tree 6 7 8 5 4 3 1 2 12 10 13 11 12 2 3 21 2 2 11 15 12 2 1 2 2 7 2 2

15. Katz, Stoica F04 15 Multicast Service Model  Receivers join a multicast group which is identified by a multicast address (e.g. G)  Sender(s) send data to address G  Network routes data to each of the receivers  Note: multicast vs. broadcast -Broadcast: packets are delivered to all end-hosts in the network -Multicast: packets are delivered only to end-hosts that are in (have joined) the multicast group R 0 joins G R 1 joins G R n-1 joins G S R0R0 R1R1...... [G, data] R n-1 Net

16. Katz, Stoica F04 16 Multicast Service Model (cont’d)  Membership access control -open group: anyone can join -closed group: restrictions on joining  Sender access control -anyone can send to group -anyone in group can send to group -restrictions one which host can send to group

17. Katz, Stoica F04 17 Multicast and Layering  Multicast can be implemented at different layers -data link layer e.g. Ethernet multicast -network layer e.g. IP multicast -application layer e.g. End system multicast  Which layer is best?

18. Katz, Stoica F04 18 Multicast Implementation Issues  How are multicast packets addressed?  How is join implemented?  How is send implemented?  How much state is kept and who keeps it?

19. Katz, Stoica F04 19 Data Link Layer Multicast  Recall: end-hosts in the same local area network (LAN) can hear from each other at the data link layer (e.g., Ethernet)  Reserve some data link layer addresses for multicast  Join group at multicast address G -Network interface card (NIC) normally only listens for packets sent to unicast address A and broadcast address B -To join group G, NIC also listens for packets sent to multicast address G (NIC limits number of groups joined) -Implemented in hardware, thus efficient  Send to group G -Packet is flooded on all LAN segments, like broadcast -Can waste bandwidth, but LANs should not be very large  Only host NICs keep state about who has joined  scalable to large number of receivers, groups

20. Katz, Stoica F04 20 Problems with Data Link Layer Multicast  Single data link technology  Single LAN -limited to small number of hosts -limited to low diameter latency -essentially all the limitations of LANs compared to internetworks

21. Katz, Stoica F04 21 Network Layer (IP) Multicast  Overcomes limitations of data link layer multicast  Performs inter-network multicast routing -relies on data link layer multicast for intra-network routing  Portion of IP address space defined as multicast addresses -2 28 addresses for entire Internet  Open group membership  Anyone can send to group -flexible, but leads to problems

22. Katz, Stoica F04 22 IP Multicast Routing  Intra-domain -Distance-vector multicast -Link-state multicast  Inter-domain -Protocol Independent Multicast -Single Source Multicast

23. Katz, Stoica F04 23 Distance Vector Multicast Routing Protocol (DVRMP)  An elegant extension to DV routing  Use shortest path DV routes to determine if link is on the source-rooted spanning tree  Three steps in developing DVRMP -Reverse Path Flooding -Reverse Path Broadcasting -Truncated Reverse Path Broadcasting

24. Katz, Stoica F04 24 Reverse Path Flooding (RPF)  Extension to DV unicast routing  Packet forwarding -If incoming link is shortest path to source -Send on all links except incoming -Packets always take shortest path assuming delay is symmetric  Issues -Some links (LANs) may receive multiple copies -Every link receives each multicast packet, even if no interested hosts s:2 s s s:1 s:3 s:2 s:3 r r

25. Katz, Stoica F04 25 Example  Flooding can cause a given packet to be sent multiple times over the same link  Solution: Reverse Path Broadcasting x x y y z z S S a b duplicate packet

26. Katz, Stoica F04 26 Reverse Path Broadcasting (RPB)  Chose parent of each link along reverse shortest path to source  Only parent forward to a link (child link)  Identify Child Links 1.Routing updates identify parent 2.Since distances are known, each router can easily figure out if it's the parent for a given link 3.In case of tie, lower address wins x x y y z z S S a b 5 6 child link of x for S forward only to child link

27. Katz, Stoica F04 27 Don’t Really Want to Flood!  This is still a broadcast algorithm – the traffic goes everywhere  Need to “Prune” the tree when there are subtrees with no group members  Solution: Truncated Reverse Path Broadcasting

28. Katz, Stoica F04 28 Truncated Reverse Path Broadcasting (TRPB)  Extend DV/RPB to eliminate unneeded forwarding  Identify leaves -Routers announce that a link is their next link to source S -Parent router can determine that it is not a leaf  Explicit group joining on LAN -Members periodically (with random offset) multicast report locally -Hear an report, then suppress own  Packet forwarding -If not a leaf router or have members -Out all links except incoming r1 r2 S S NL LL

29. Katz, Stoica F04 29 Pruning Details  Prune (Source,Group) at leaf if no members -Send Non-Membership Report (NMR) up tree  If all children of router R send NRM, prune (S,G) -Propagate prune for (S,G) to parent R  On timeout: -Prune dropped -Flow is reinstated -Down stream routers re-prune  Note: a soft-state approach

30. Katz, Stoica F04 30 Pruning Details  How to pick prune timers? -Too long  large join time -Too short  high control overhead  What do you do when a member of a group (re)joins? -Issue prune-cancellation message (grafts)

31. Katz, Stoica F04 31 Distance Vector Multicast Scaling  State requirements: -O(Sources  Groups) active state  How to get better scaling? -Hierarchical Multicast -Core-based Trees

32. Katz, Stoica F04 32 Core Based Trees (CBT)  Pick a “rendevouz point” for the group called the core. -Shared tree  Unicast packet to core and bounce it back to multicast group  Tree construction is receiver-based -Joins can be tunneled if required -Only nodes on One tree per group tree involved  Reduce routing table state from O(S x G) to O(G)

33. Katz, Stoica F04 33 Example  Group members: M1, M2, M3  M1 sends data root M1 M2 M3 control (join) messages data

34. Katz, Stoica F04 34 Disadvantages  Sub-optimal delay  Single point of failure -Core goes out and everything lost until error recovery elects a new core  Small, local groups with non-local core -Need good core selection -Optimal choice (computing topological center) is NP hard

35. Katz, Stoica F04 35 Problems with Network Layer Multicast (NLM)  Scales poorly with number of groups -A router must maintain state for every group that traverses it -Many groups traverse core routers  Supporting higher level functionality is difficult -NLM: best-effort multi-point delivery service -Reliability and congestion control for NLM complicated  Deployment is difficult and slow -ISP’s reluctant to turn on NLM

36. Katz, Stoica F04 36 NLM Reliability  Assume reliability through retransmission  Sender can not keep state about each receiver -E.g., what receivers have received -Number of receivers unknown and possibly very large  Sender can not retransmit every lost packet -Even if only one receiver misses packet, sender must retransmit, lowering throughput  N(ACK) implosion -Described next

37. Katz, Stoica F04 37 (N)ACK Implosion  (Positive) acknowledgements -Ack every n received packets -What happens for multicast?  Negative acknowledgements -Only ack when data is lost -Assume packet 2 is lost S S R1 R2 R3 123

38. Katz, Stoica F04 38 NACK Implosion  When a packet is lost all receivers in the sub-tree originated at the link where the packet is lost send NACKs S S R1 R2 R3 3 3 3 2?

39. Katz, Stoica F04 39 Barriers to Multicast  Hard to change IP -Multicast means change to IP -Details of multicast were very hard to get right  Not always consistent with ISP economic model -Charging done at edge, but single packet from edge can explode into millions of packets within network  Troublesome security model -Anyone can send to a group -Denial-of-service attacks on known groups

40. Katz, Stoica F04 40 Application Layer Multicast (ALM)  Let the hosts do all the “special” work -Only require unicast from infrastructure  Basic idea: -Hosts do the copying of packets -Set up tree between hosts  Example: Narada [Yang-hua et al, 2000] -Small group sizes <= hundreds of nodes -Typical application: chat

41. Katz, Stoica F04 41 Narada: End System Multicast Stanford CMU Stan1 Stan2 Berk2 “Overlay” Tree Gatech Berk1 Berkeley Gatech Stan1 Stan2 CMU Berk1 Berk2

42. Katz, Stoica F04 42 Algorithmic Challenge  Choosing replication/forwarding points among hosts -how do the hosts know about each other -and know which hosts should forward to other hosts

43. Katz, Stoica F04 43 Advantages of ALM  No need for changes to IP or routers  No need for ISP cooperation  End hosts can prevent other hosts from sending  Easy to implement reliability -use hop-by-hop retransmissions

44. Katz, Stoica F04 44 Performance Concerns  Stretch -ratio of latency in the overlay to latency in the underlying network  Stress -number of duplicate packets sent over the same physical link

45. Katz, Stoica F04 45 Performance Concerns Duplicate Packets: Bandwidth Wastage CMU Stan1 Stan2 Berk2 Gatech Berk1 Delay from CMU to Berk1 increases Stanford Berkeley Gatech Stan1 Stan2 CMU Berk1 Berk2

46. Katz, Stoica F04 46 Single Sender Multicast  Many problems with IP multicast disappear if each group is associated with a single source  Hosts joining multicast group can send join messages to source -this sets up delivery tree -no worry about “root” being in wrong place  This solves several problems: -better security and charging model -simple algorithm

47. Katz, Stoica F04 47 Example  Group members: M1, M2, M3 source M1 M2 M3 control (join) messages data

48. Katz, Stoica F04 48 What’s Wrong with SSM?  Multiple sources? -Can set up group per source, or... -Source can serve as relay for other senders  Algorithm? -Trivial  So, why isn’t SSM the answer? -Multicast no longer serves as “rendezvous” -Ok for “broadcast” apps, not good for “meeting” apps

49. Katz, Stoica F04 49 What Do You Need to Know?  DVRMP  CBT  SSM  How they compare