MANET MulticastingBroadcasting

1
MANET Multicasting/Broadcasting

• Computer Science and Engineering Department
• University of Minnesota
• Wireless Networking Seminar

• Hung Quang Ngo
• hngo_at_cs.umn.edu
• June 21, 2001

2
Outline

• What is Multicasting/broadcasting ?
• IP Routing Revisited
• IP Multicasting
• MANET Multicasting
• ODMRP
• Research Issues
• Summary

3
What Is Broadcasting/multicasting?

• Broadcasting one to all
• Multicasting one to many (a predefined group)
• Applications
• Corporate messages to employees
• Live stock quotes
• Video/Audio conferencing,
• In wireless environments
• Collaborative battle field
• Search and rescue
• Disaster recovery
• Wired broad/multi casting does not apply to the
  MANET environment

4
Multicast/Broadcast Characteristics

• Real time
• One to many
• Intrinsically lots of overhead
• No matter what we do, we can only reduce the
  overhead to a certain level

5
Multicast Components

• Addressing and hardware support
• Group Management Join Leave
• Algorithm
• Protocol

6
Addressing

• 1110 28 bit group ID
• Mapping lower 23 bits into the lower 23 bits of
  an Ethernet card ? 32 groups mapped to the same
  MAC address
• Messages sent to a group forwarded down to MAC
  layer with the corresponding MAC address
• MAC layers at hosts look at MAC address to pick
  it up
• IP layer removes packets that the applications
  are not interested in (may be 31 of them)

7
Group Management IGMP

• Locally, hosts send report to inform local
  routers that it wants to join a group
• Routers send group queries periodically
• If no response after several queries, prune the
  group (v1)
• Election for group querier if gt 1 router
• Hosts can send leave for efficient pruning (v2)
• Hosts can specify the set of sources that its
  interested in (v3)

8
IP Routing Revisited Algorithms

• Distance Vector
• Routers exchange global information locally
• Features simple to implement, do not scale well
• Actual protocol RIP (RFC 1058) still most
  widely used
• Link State
• Router exchange local information globally
• Features more complex, more info exchanged,
  scale well
• Actual protocol OSPFv2 (RFC 1583) the future
  is here
• Path Vector
• A variation of distance vector (exchanged vectors
  with path information)
• Features very complex, used for inter-AS routing
• Actual protocols BGPv4 (RFC 1771) for IPv4, IDRP
  (ISO 10747) for IPv6

9
IP Routing Revisited Protocols

• Autonomous Systems (AS)
• Group of connected routers exchanging info via a
  common protocol
• Managed by a single organization
• Interior Routing Protocols (IRP)
• Used for intra-AS routing
• RIP is an IRP (?, distance-vector), OSPF
  (link-state)
• Exterior Routing Protocols (ERP)
• Used for inter-AS routing
• ERP (obsolete), BGP, IDRP (path-vector)

10
IP-Multicasting Basic Ideas

• Use class D addresses (1110.) as group addresses
• Hosts could join and leave a group dynamically
• IGMP (Internet Group Management Protocol) for
  hosts to inform routers of membership status
• Senders might not belong to a group, routers
  collaboratively deliver messages to groups
• Need hardware support (NIC multicast-enabled)
• Typically all routers from sender to receivers
  are multicast-enabled
• Actually tunneling through non-multicast parts
• Similar to IP routing need algorithms and
  protocols

11
Tunneling (courtesy of IPMI)
12
IP-Multicasting Algorithms

• Flooding way too inefficient
• Spanning Trees simple but do not scale
• Source-Based Tree
• Reverse Path Broadcasting (RPB) relatively
  efficient and simple, does not take group
  membership into account need an entry for each
  (source, group) pair ? memory problem
• Truncated RPB (TRPB) truncate the leaves if no
  group member is in a leaf router
• Reverse Path Multicasting (RPM) propagate the
  truncating information upward widely used
• Core Based Tree (CBT) one entry for each group,
  messages forwarded toward the core ? bottleneck

13
IP-Multicasting Protocols

• Two ways to classify protocols
• Is it unicast-protocol dependent?
• Unicast-dependent Distance Vector Multicast
  Routing Protocol (DVMRP), Multicast Open Shortest
  Path First (MOSPF), Core-Based Tree (CBT)
• Unicast-independent Protocol Independent
  Multicast (PIM) Dense Mode (DM) and Sparse Mode
  (SM)
• At which mode is it efficient?
• Dense-Mode DVMRP, MOSPF, PIM-DM
• Sparse-Mode CBT, PIM-SM

14
Dense-Mode Multicasting

• Group members are densely distributed
• Bandwidth isnt a problem
• Flooding used to propagate information
• Data-Driven approach to build the trees
• Protocols
• DVMRP
• MOSPF
• PIM-DM

15
Sparse-Mode

• Group members scattered all over
• Does not mean that there are few members, just
  that they are widely dispersed
• Bandwidth is limited
• Hence, must be more selective on setting up and
  maintaining the multicast trees
• Receiver-Driven approach to build the trees
• Protocols
• CBT
• PIM-SM

16
Source-Based Tree
17
Core-Based Tree
18
DVMRP (RFC 1075)

• RIP dependent (Distance Vector)
• Use RPM algorithm
• Additionally, graft back a branch when new
  members are added, for faster group joining
• Still periodically flood packets (build the whole
  tree)
• Does not scale well
• Used widely in MBONE

19
MOSPF (RFC 1583)

• OSPF dependent (link state)
• Periodically flood membership information and
  link-state packets
• Routers build source-based trees based on these
  packets
• Does not scale well

20
PIM-DM

• Mostly the same as DVMRP
• Unicast independent, but still need the routing
  table
• More redundant information

21
PIM-SM

• Same like Core-Based Tree
• Core is now called a Rendezvous-Point

22
MANET Multicasting The Big Pic

• IP-Multicasting protocols cannot be applied
• Network topology is dynamic
• Limited resources (bandwidth, computational
  power, storage capacity, battery power, )
• A lot of variations of IP-Mcast were proposed
• Publications on MANET Multicasting
• lt 1998 4 papers
• 1999 12 papers, 2 internet-draft
• 2000, 2001 11 papers, several PhD theses
• About 20 different protocols were proposed
• Result MPM (Multicast Protocol Mess)

23
MANET Multicasting Performance Parameters
Metrics

• Difficult to say, since no actual multicast
  backbone is available. Many criteria are
  arbitrary in this sense. (Took me some time on
  this ?)
• Many are arbitrarily simulated, some use
  GloMoSim, ns-2

24
MANET Multicasting Protocols
25
Trade-offs When Design Protocols

• Flooding and others
• Might be good for specialized applications
• Excessive overhead, but simple and scalable
• Tree based
• Vulnerable when mobility is high
• Efficient when mobility is low (same as wired
  networks)
• Less overhead for tree maintenance
• Mesh based
• Scale well to changing topology
• More overhead to maintain forwarding group
• Higher throughput
• Tree vs. Mesh overhead vs. throughput
  scalability

26
A Typical MANET Mcast Protocol

• Group membership maintenance mechanism
• How to join
• How to leave
• Configuration creation mechanism
• When a source has something to send, create a
  multicast tree (or mesh)
• Configuration maintenance mechanism
• Dealing with broken link
• Pruning unused branches
• Loop handling mechanism
• How to deal with potential loops when hosts move
• Error handling mechanism (best effort or reliable)

27
ODMRP (Lee, Su Gerla UCLA)

• Mesh creation/maintenance periodic on-demand
• Source periodically (scoped) floods JOIN_DATA
  packet to refresh membership and route
  information
• Upon receiving a non-duplicate JOIN_DATA, a node
  inserts or updates its routing table the entry to
  the source
• If TTL gt 0 (scoping), it rebroadcast JOIN_DATA
• If the node is in the group, it creates a
  broadcasts JOIN_TABLE contains (source, next-hop)
  pairs
• If a node gets JOIN_TABLE and next-hop is itself,
  it flags itself as a forwarding node and send a
  JOINT_TABLE of its own back to the source
• The forwarding group, a mesh, is created in the
  process

28
The Mesh
29
ODMRP (cont.)

• Multicasting
• The source can send data via selected routes and
  the forwarding group
• Source sends JOIN_DATA after every INTERVAL
• Forwarding nodes keep functioning while FG_FLAG
  is not expired
• If GPS is utilized, it can be used to predict
  INTERVAL to reduce overhead
• Leaving a group soft-state approach
• A receiver leaves by stop sending JOIN_TABLE back
• A sender leaves by stop sending JOIN_DATA

30
ODMRP pros and cons

• Advantages
• Simple
• Effective, robust to mobility
• Scalable to large number of nodes
• Can be used as a unicast protocol
• Disadvantages
• Non optimal routes
• Effectiveness deteriorates quickly as group size
  increases (flooding info overhead)

31
Research Issues

• A mesh-based protocol with lower control overhead
  than ODMRP
• Evaluate ODMRP others under different metrics
  multimedia traffic, delay, power consumption
• Might be good to have something like PIM-DM and
  PIM-SM (ODMRP-LM ODMRP-HM ?)
• Incorporate other criteria into ODMRP-like
  routing (power consumption and all that)
• Detail different approaches, trade-offs and
  phases of a multicast protocols for MANETs

32
Summary

• Overviewed
• IP multicasting
• MANET multicasting
• Indicated research issues
• Classified 25 papers into 4 categories
• Junk
• ½ Junk
• OK
• Good