Design an Application Layer Multicast

Post on 23-Feb-2016

38 views 0 download

Tags:

description

Design an Application Layer Multicast. CHUN-TING, WU Teacher: KAI-WEI, KE. Outline. Introduction MANET Examples Performance Matrics Conclusions. Overview. Traditional network architectures distinguish between two types of entities - PowerPoint PPT Presentation

transcript

Design an Application Layer Multicast

CHUN-TING, WUTeacher: KAI-WEI, KE

2

Outline Introduction MANET Examples Performance Matrics Conclusions

3

Overview Traditional network architectures

distinguish between two types of entities› end systems (hosts) and the network

(routers and switches). The key architectural question is: what

new features should be added to the IP layer?› Multicast and QoS

4

What is a multicast Delivery of information to a group Creating copies only when the links to

the multiple destinations

5

Why multicasting video-on-demand live media streaming video conferencing multiplayer games

Real time and large data flow

6

IP layer vs. Application layer

In deciding whether to implement multicast services at the IP layer or at end systems, there are two conflicting considerations that we need to reconcile.

7

IP layer vs. Application layer

First, IP Multicast requires routers to maintain per group state› violate the “stateless” architectural

principle, and introduce high complexity and serious scaling.

Second, IP Multicast calls for changes at the infrastructural level› slows down the pace of deployment.

8

Finally, IP Multicast is providing higher level features such as reliability, congestion control, flow control, and security has been shown to be more difficult than in the unicast case.

9

We consider a model in which multicast related features, such as group membership, multicast routing and packet duplication, are implemented at end systems, assuming only unicast IP service.

10

To be resolved An overlay approach to multicast,

however efficient, cannot perform as well as IP Multicast.

It is impossible to completely prevent some redundant traffic on physical links.

Communication between end systems involves increasing latency.

11

(b) naive unicast transmission. (c) the IP Multicast tree constructed by

DVMRP (d) an “intelligent” overlay tree

12

Conceptual comparisonIssues IP multicast Application

multicastefficiency in terms of delay/bandwidth

High Low — Medium

Complexity or Overhead

Low Medium — High

Ease of deployment Low Medium — High

OSI layer works Network layer Application layer

13

ALM properties GROUP MANAGEMENT

› Mesh First vs. Tree First› Source Specific Tree vs. Shared Tree› Refinement

ROUTING MECHANISM› Shortest Path› Minimum Spanning Tree› Clustering Structure

14

Mesh First vs. Tree First Tree-based

› Source node as the root, thus there is only one single path between every pair of sender and receiver.

› It’s very efficient since the routing information needs to be maintained is very little.

Mesh-based› More than one path between each sender and

receiver pair exists.› More robust but less efficient.

15

Source Specific Tree vs. Shared Tree

Source-tree-based› It construct a multicast tree among all the member nodes

for each source node.(usually this is a shortest path tree)› More efficient.› Too much routing information to maintain.

Shared-tree-based› It constructs only one multicast tree for a multicast group

including several source nodes. (usually this is a minimum spanning tree)

› Every source uses this tree to do multicast. › Less efficient.› It reduces the overhead greatly by maintaining less

routing information.

16

Refinement Constructed trees might be different

› Depending upon the order of joining requests

› Construct the tree in real time and have no a-priori knowledge of node arrivals

Local optimum to the global optimum and improves the system’s performance

17

Shortest Path A Shortest Path Tree constructs a

minimum cost path from a source node to all its receivers

A source-specific multicast tree or in graph theoretic terms a rooted tree

18

Minimum Spanning Tree Construct a low cost tree Used by a shared tree

19

Clustering Structure Construct a hierarchical cluster of

nodes with each cluster having a head Advantages

› Reduction in control overhead› Faster joining and leaving› A sub-optimal tree

20

MANET Form a temporary and dynamic

wireless network on a wireless channel without the fixed infrastructure

Self-organizing collection of Mobile Nodes

Low bandwidth, mobility and low power Due to the limited transmission range,

multiple hops may be needed

21

IP network and MANET router vs. forwarding node traverse internet vs. multi-hops routing fixed vs. topology changes frequently

22

Flooding-based Proactive (table-driven)

› continuously evaluate routes› maintain up-to-date routing information› periodically flood its location to other

nodes› maintains a location table

Reactive (on-demand)› routing creates routing only when desired› in searching of the destination

23

Quorum-based Explicit

› Location update is sent to a defined subset (update quorum)

› Location query is sent to a subset (query quorum)

Implicit› Location servers are chosen via a hashing

function

24

MAODV Multicast On-demand Distance Vector

routing protocol This protocol uses broadcast to find the

route in an on-demand way and constructs a shared routing tree.

25

26

MAODV Example

L

Group Join ProcessBroadcast - RREQOnly GM RespondsMulticast ActivationBroadcast Group Hello

Group member Multicast Tree member

Ordinary node

Potential Group memberMulticast link

Communication link

27

MAODV Example

L

Leaving a Multicast Group

Non leaf NodeMust remain as a Tree

member

Leaf NodeCan remove itself

from MTAgain Leaf Node Remove himself from

MT

Group member Multicast Tree member

Ordinary node

Potential Group memberMulticast link

Communication link

28

Local Connectivity Management Node must periodically hear from active

neighbors to know they are still within range

Every time hear broadcast, update lifetime If no broadcast with hello_interval,

broadcast Hello packet Failure to hear from a neighbor for

› (1 + allowed_hello_loss ) * hello_lifetime indicates loss of link

29

Unicast Route Discovery Source broadcasts

› Route Request(RREQ)› <J_flag, R_flag, Bcast_ID,

Src_Addr, Src_Seq#, Dst_Addr, Dst_Seq#, HopCnt>

Node can reply to RREQ if› – It is the destination› – It has a “fresh enough” route to

the destination Nodes create reverse route entry Record Src IP Addr / Broadcast ID

to prevent multiple processing

Source

Destination

RREQ

30

Unicast Route Discovery Source broadcasts

› Route Request(RREQ) Node can reply to RREQ if

› – It is the destination› – It has a “fresh enough”

route to the destination Nodes create reverse

route entry Record Src IP Addr /

Broadcast ID to prevent multiple processing

Source

DestinationRREP

31

Unicast Route Discovery Source broadcasts

› Route Request(RREQ) Node can reply to RREQ if

› – It is the destination› – It has a “fresh enough”

route to the destination Nodes create reverse

route entry Record Src IP Addr /

Broadcast ID to prevent multiple processing

Source

Destination

32

Forward Path Setup Nodes along path

create forward route to dest

Source begins sending data when receives first RREP

Source

Destination

Data

33

Design properties GROUP MANAGEMENT

› Tree First› Shared Tree› No refinement

ROUTING MECHANISM› Minimum Spanning Tree

MANET ROUTING› Reactive› Flat

34

Performance Matrics Latency

› end-to-end delay Bandwidth

› throughput at the receiver Stress

› number of identical copies of a packet carried by a physical link

35

Performance Matrics Resource Usage

› 57 / 30 / 32 Protocol Overhead

› total bytes of non-data traffic

36

Conclusions Although, MANET multicast is an ALM,

using wired mech. cannot perform well. Actually, there is no a “one-for-all” scheme

that works well with different scenarios. Highly dynamic environment, nodes move

arbitrarily, thus network topology changes frequently and unpredictably

Moreover, bandwidth and battery power are limited.

Make multicast extremely challenging

37

References Tu, W. & Jia, W., “An End Host Multicast Protocol for Peer-to-Peer

Networks,” LCN '05: Proceedings of the The IEEE Conference on Local Computer Networks 30th Anniversary IEEE Computer Society, 2005, pp. 392-399

Hosseini, M., Ahmed, D., Shirmohammadi, S. & Georganas, N., “A Survey of Application-Layer Multicast Protocols,” Communications Surveys & Tutorials, IEEE, 2007, Vol. 9(3), pp. 58-74

Junhai, L., Danxia, Y., Liu, X. & Mingyu, F., “A survey of multicast routing protocols for mobile Ad-Hoc networks,” Communications Surveys Tutorials, IEEE, 2009, Vol. 11(1), pp. 78 -91

Perkins, C. & Royer, E., “Ad-hoc on-demand distance vector routing,” Mobile Computing Systems and Applications, 1999. Proceedings. WMCSA '99. Second IEEE Workshop on, 1999, pp. 90 -100