1 reliable multicast protocols for manet reporter : 吳政鴻 date : 2005/5/17

1

Reliable Multicast Protocols for MANET

Reporter : 吳政鴻Date : 2005/5/17

2

Outline

• An Overview of MANET• An Overview of Reliable Multicast Protocol• Automatic Retransmission reQuest Based

Reliable Multicast Protocols (ARQ-Based)– Reliable Multicast Algorithm (RMA)– Reliable Adaptive Light Weight Multicast

Transport protocol (RALM)– Reliable, Adaptive, Congestion-Controlled Adhoc

Multicast Transport Protocol (ReAct)

3

• Gossip-Based Reliable Multicast Protocols– Anonymous Gossip (AG)– Route Driven Gossip (RDG)

• Forward Error Correction (FEC) Based Reliable Multicast Protocols– Reliable Multicast Data Distribution Protocol

(RMDP)

• Comparisons• Conclusions

4

An Overview of MANET

• Definition

• Characteristics

• Limitations

5

Definition

• Mobile ad hoc network(MANET), or simply ad hoc network, comprises nodes that freely and dynamically self-organize into arbitrary and temporary network topology without any infrastructure support. (Chlamtac,Conti,andLiu,2003)

• Nodes are communication devices comprise of laptop computer, PDA, mobile phone and etc. Nodes formanet work to communicate with each other.

• Networking infrastructure refers to the facility of which the sole purpose is to carry the data generated by each node to the respective destination node

6

Characteristics

• Infrastructure-less or with minimum infrastructure support

• Self-organizing and self-managing

• Multi-hop

7

Infrastructure-less or with minimum infrastructure support

• A Pure ad hoc network does not have, or simply does not rely on infrastructure support (for routing, network management, and etc.)

• A hybrid ad hoc network consists of both client nodes and infrastructure nodes,i.e. nodes whose function is merely transporting traffic for the client nodes

• Hybrid networks are more common

8

Self-organizing and self-managing

• Since network infrastructure is not available, the nodes must organize and maintain the network by themselves

9

Multi-hop

• Since each node can route traffic for the others, multi-hopping is possible.

• Single hop ad hoc network does not form a large scale network

10

Limitations

• Ad hoc network is not a one size fit all measure.

Its current limitations are listed below:– Killer application has not turned up– Acceptance by users is unclear– Delay caused by multi-hopping – Quality of service is difficult to be guaranteed – Prone to security threat

11

An Overview of Reliable Multicast Protocol

• What is Multicast Communication?

• What is Reliable Multicast?

• What is Reliable Multicast use in MANET?

• Three categories according to the recovery mechanisms being used

12

What is Multicast Communication?

• Group communication mechanism

– Provides one-to-many and many-to-many communication

• Efficient dissemination of messages

– Network-based duplication (when needed)

– Multicast retransmissions

– Bandwidth savings

– Parallel delivery at multiple locations

13

IP Multicast Communication

14

Example IP Multicast Use (Access Grid )

15

What is Reliable Multicast?

• Properties similar to TCP• Application-level program• Uses IP Multicast as the underlying communication

mechanism• Reliable and ordered delivery of messages within a

group• Tracks group membership• IETF Reliable Multicast Transport Working Group is

defining standardized building blocks

16

Example Reliable Multicast Use (RemoteInstrument Access)

17

What is Reliable Multicast use in MANET ?

• Reliable multicast becomes a very challenging research problem due to high packet loss rate pertained to MANET

• Reliable multicast solutions proposed for wired network can not be directly ported for MANET– link breakages– route changes– concentrated retransmissions– concentrated retransmissions and heavy overhead

18

Three categories according to the recovery mechanisms being used

• Automatic Retransmission Request (ARQ)-based– Lost packets are retransmitted by the sources until

they are recovered at all the receivers• gossip-based

– multicast packets are repeatedly transmitted for a few times by a few of the multicast members in a peer-to-peer fashion

• Forward Error Correction (FEC)-based– embed redundant data (e.g., erasure code) in each

packet before transmitting.

19

• Automatic Retransmission reQuest Based Reliable Multicast Protocols (ARQ-Based)– Reliable Multicast Algorithm (RMA)– Reliable Adaptive Light Weight Multicast

Transport protocol (RALM)– Reliable, Adaptive, Congestion-Controlled

Adhoc Multicast Transport Protocol (ReAct)

20

Reliable Multicast Algorithm (RMA)

• Assumption

• Protocol description

• Advantage versus Disadvantage

21

Assumption

• RMA is a multicast protocol supporting reliable transmission via acknowledgement from receivers and retransmissions from the sources

• RMA assumes that the sources have the full knowledge of group membership via JOIN or ACK messages

22

Protocol description

• RMA works in two phases: – multicast – retransmission

• Two types of multicast messages to the group member– MKNOWN

• unicast

– MUNKNOWN• broadcast

23

• Source waits for MACKs for a period of time after the messages being sent out

• If the source is not able to collect all the MACKs from all the group members, the source enters the retransmission phase and sends a MUNKNOWN message with a flag in RETRANSMIT field

• Receiver could broadcast MACK to the source (BMACK), if a return path is not valid

24

Hello Message

25

MKNOWN Message

26

MUKNOWN Message

27

MACK Message

28

BMACK Message

29

Advantage versus Disadvantage

• Advantage– The sender guarantees retransmissions of lost

packets– less message forwarding and less bandwidth

usage

• Disadvantage– all the receivers must send ACKs back to the

sender == >> Feedback implosion

30

Reliable Adaptive Light Weight Multicast Transport protocol (RALM)

• Assumption



31

Assumption

• RALM assumes that the group membership is known to the sources

• This enables the sources to maintain a Receiver List

32


• Source selects a node from the receiver list as a feedback receiver in a round-robin fashion and notifies it together with the data packets

• The feedback receiver is responsible for replying ACK or NACK to the source until it collects all data packets

33

• If Source receives a NACK, it enters the retransmission phase by slowing down the transmission rate first and retransmits the lost packets to the group until ACK to the lost packets are received and the current feedback receiver successfully obtains all the packets

34

• This single-node feedback approach is effective when packet losses are due to congestion at a bottleneck link

35

A Rec

B Rec

C Rec

A Fee

B Rec

C Rec

B Rec

C RecA

B

C

Receiver List

A

D

B Fee

C Rec

D Rec

38


• Advantage– RALM also reduces control overhead by

requiring one receiver at a time– Effectively reduces the burden at the sender in

receiving and processing the feedbacks and reduces congestion around the sender == >> Solve Feedback implosion

• Disadvantage– RALM works well for static MANET

39

Reliable, Adaptive, Congestion-Controlled Adhoc Multicast Transport Protocol (ReAct)

• Assumption



40

Assumption

• source-oriented component works the same as RALM

41


• ReAct adds a new recovery mechanism “local recovery” to RALM

• Local recovery occurs right after the receiver detects a lost packet

• the receiver requests one of the upstream group members (recovery node) starting from the closest one

42

• The recovery node responses with the expect packets if it has them or it rejects the request

• Upon receiving the rejection, the receiver will retry recovery by choosing a farther away upstream node as a recovery node

• Only after several failures of the local requests, the receiver sends a NACK to the source for retransmission

43

C

A

B Packet Loss

44


• Advantage– Local recover gets missing packets faster

than source-oriented retransmission, reduces the burden/congestion at the source, and alleviates potential feedback implosion problem

• Disadvantage– when local recovery frequently fails and

source recovery is triggered all the time

45

• Gossip-Based Reliable Multicast Protocols– Anonymous Gossip (AG)– Route Driven Gossip (RDG)

46

Anonymous Gossip (AG)

• Assumption



47

Assumption

• Implements gossip-based recovery on top of a multicast operation

• Gossip-messages only contain sequence numbers for missing packets

• Routing information of MAODV at receiver side is adopted for sending gossips

48


• AG works in two phases: – multicast

• Source sends multicast packets in best-effort

– Recovery• runs at background for recovering lost packets

• a group member periodically transmits a gossip request message about missing and successfully received packets to a pseudo-randomly selected neighbor node

49

• If receiver receiving the gossip request– non-group-member neighbor simply forwards

the packet to one of its neighbors– Group-member neighbor will accept and reply

the gossip message with a certain probability

• This procedure ends until a node replies the gossip message or the lifetime of the message expires.

50

• Works in background with multicast protocol.

• Members “gossip” randomly with other members to recover lost messages.

• Probabilistic reliability guarantees.

Multicast Protocol

Classical Gossip+ Probabilistic

Reliability=

51

Classical Gossip

S

D

52

Anonymous Gossip

S

D

53

Informed Gossip

S

D

54


• Advantage– AG is a reliable multicast protocol that does not

require membership information.– To reduce the network traffic, gossip requests are

sent to nearer members with higher probability than to farther members.

– AG operates independent of topology changes

• Disadvantage– AG can not guarantee the missing packets will be

answered eventually

55

Route Driven Gossip (RDG)

• Assumption



56

Assumption

• RDG does not use full multicast membership information, but partial knowledge

• RDG builds on top of a MANET unicast routing protocol DSR

57


• each existing member will only reply to the solicitation with a probability, resulting in a partial membership view at the joining node

• A gossip message generated at each session contains both new data packets and packet IDs of missing packets

• The gossip message is sent to F (fan out parameter) other group members randomly picked up from its partial member view

58

• each new data packet will be gossiped for a few number of times (quiescence threshold) to ensure its spreading

59Data packets, digests of missing packets, view

1

30

2

Data Structures

Identifier

Group identifier

View

• active

• passive

• remove

Data buffer

• new

• old

JOIN

RECEIVEGREQUEST

GOSSIP

RECEIVEGOSSIP

LEAVE

5

45

0

0

5

1

5

0

1

0

5

1

1

0

fanout

Fquiescence threshold

τqPushPull

60

I n f e c t e d S u s c e p t i b l e N o n m e m b e r n o d e s

R o u n d 0

R o u n d 2

R o u n d 1

R o u n d 3

1 5

93

1 2

2 06

1 8

1 31 61 1

5

7 1

1 4

41 0

8

1 7

2

1 9

1 5

9

3

1 2

2 06

1 8

1 31 61 1

5

7 1

1 4

41 0

8

1 7

2

1 9

1 5

9

3

1 2

2 06

1 8

1 31 61 1

5

7 1

1 4

41 0

8

1 7

2

1 9

1 5

9

3

1 2

2 06

1 8

1 31 61 1

5

7 1

1 4

41 0

8

1 7

2

1 9

C r a s h e d M e m b e r

15 9, 10

15 5, 39 10, 1310 2, 8

All members receive the message.

9 3, 55 1, 1910 1, 1313 10, 152 1, 8

61


• Advantage– RDG eliminates burdens at sources for handling

retransmission; instead, every group member participates in loss recovery

– The performance of the protocol can be turned through the parameters fan out and quiescence threshold

• Disadvantage– RDG can not guarantee reliable delivery of all the

packets

62

• Forward Error Correction (FEC) Based Reliable Multicast Protocols– Reliable Multicast Data Distribution

Protocol (RMDP)

63

Reliable Multicast Data Distribution Protocol (RMDP)

• Assumption



64

Assumption

• FEC transmits redundant data with the original data transmission

• the k packets will be encoded in to n

(n > k) packets. The n packets include redundant information

65


• RMDP is a hybrid FEC+ARQ protocol for reliable distribution of bulk data receivers

• After it receives k different packets, it decodes for the original source data

• The source adjusts its sending pointer to the packet where the largest number of packets is requested by different receivers

67


• Advantage– FEC technique helps RMDP to tolerant packet losses and to

recover from losses with less feedback packets to the sources==>> Solve Feedback implosion

– when errors or packet losses happen at the receiver, original data can be reconstructed using the ones received

• Disadvantage– RMDP incurs long packet latency because a receiver has to

wait for the reception of k packets before it can decode and delivery them to applications

– Using the redundant data to increase packet length

68

Comparisons

69

Conclusions

• classification based on the recovery mechanisms

• The analyses and comparisons will help in choosing a suitable reliable multicast protocol for specific network conditions

70

REFERENCES

• Beini Ouyang and Xiaoyan Hong, Yunjung Yi, A Comparison of Reliable Multicast Protocols for Mobile Ad Hoc Networks, IEEE, 2005.

• Thiagaraja Gopalsamy, Mukesh Singhal, D. Panda and P. Sadayappan, A Reliable Multicast Algorithm for Mobile Ad hoc Networks, IEEE, 2002.

• Ken Tang, Katia Obraczka, Sung-Ju Lee, Mario Gerla, Reliable Adaptive Lightweight Multicast Protocol, IEEE, 2003.

• Ken Tang, Katia Obraczka, Sung-Ju Lee, Mario Gerla, A Reliable, Congestion-Controlled Multicast Transport Protocol in Multimedia Multi-hop Network, IEEE, 2002.

71

• Ranveer Chandra, Venugopalan Ramasubramanian, Kenneth P. Birman, Anonymous Gossip: Improving Multicast Reliability in Mobile Ad-Hoc Networks, IEEE, 2001.

• Jun Luo, Patrick Th. Eugster, Jean-Pierre Hubaux, Route Driven Gossip: Probabilistic Reliable Multicast in Ad Hoc Networks, IEEE, 2003.

• Luigi Rizzo, Lorenzo Vicisano, RMDP: an FEC-based Reliable Multicast protocol for wireless, CiteSeer, 1998.

72

• END

1 reliable multicast protocols for manet reporter : 吳政鴻 date : 2005/5/17

Documents