Impact Analysis of Cheating in Application Level Multicast

1
Impact Analysis of Cheating in Application Level
Multicast

• s1090176 Masayuki Higuchi

2
Purpose

• Application level multicast is efficient
  communication technology, and it is used in
  peer-to-peer service, but they have problems.
• To evaluate performance of Application-Level
  Multicast.

3
IP multicast

• Consist of multicast routers, and the router
  control packets, but all router must be multicast
  router, so infrastructure building is difficult.
• It is unreliable because of the best-effort
  communication by IP layer.

packet
4
Application Level multicast

• A technique whereby hosts or end-nodes are
  organized into an overlay distribution tree
  without requiring any specific support from the
  network.

overlay
network
physical
network
5
HBM protocol
RP

• A Host Based Multicast protocol
• Control of a single host, the rendezvous point
  (RP) or controller.
• Periodically, each group member measures its
  distance to others.
• The RP is then responsible for the overlay
  topology calculation and its dissemination among
  the group members.

new member
6
P2P streaming service

• Sharecast
• Without high performance server.Users nodes
  relay streaming data
• How to connect.
• Send IP address, MAC address, and channel ID,
  and version to administrator.
• Response access point.

7
(No Transcript)
8
Reason of Cheating
delay
7sec
9
How to cheat

• Always reports a distance of 5 to the source.
• Adds 10 seconds to the RTT distances it measured
  to any other group member.
• Also, delays by 10 seconds any measurement
  probes if receives from any other group member.

root
5
10
cheat-node
10
Performance measures for Application level
multicast

• The link stress
• Defined as the number of redundant copies of a
  data packet on the network link.
• The relative delay penalty (RDP)
• Defined as the ratio TD/UD.
• TD is the tree delay.
• UD is the unicast delay.

TD
UD
11
Impact of cheating in worst case
3
3
cheat-node
2
2
2
2
3
cheating
2
2
3
The link stress 5 The maximum RDP 2/2
The link stress 1 The maximum RDP 14/2
12
Impact of cheating in general
3
3
cheat-node
cheating
5
The link stress 6 The maximum RDP 11/7
The link stress 3 The maximum RDP 9/7
13
Countermeasures against Cheating

• Users receives the data from two or more nodes.
  As a result the influence of the switch of the
  connection can be decreased.
• To detect the cheating, transmit ping to the
  router that connect one hop short of the
  cheat-node, and compare with RTT that cheat-node
  returned.

cheat-node
real value
cheated value
14
Conclusion and Future Works

• The countermeasures is effective measures in
  theory. As a future work, through simulations,
  confirm the effectiveness of them.