Transcript Chen-WPC10-slide

User Oriented Regional RegistrationBased Mobile Multicast Service
Management in Mobile IP Networks
Ing-Ray Chen and Ding-Chau Wang
1
PRESENTED BY
A. B. C.
Multicasting
2
According to Wikipedia:
“IP multicast is a technique for one-to-many communication
over an IP infrastructure in a network. It scales to a larger
receiver population by not requiring prior knowledge of
who or how many receivers there are. Multicast uses
network infrastructure efficiently by requiring the source to
send a packet only once, even if it needs to be delivered to a
large number of receivers.”
 Multicasting is not BROADCASTING though
 Group management is a key attribute of multicasting
Uses of Multicasting
3
 Video/Voice conferencing
 Database replication
 IP television
 Bulk Software updates to subscribers over the
network/intranet
How is Multicasting achieved?
4
 IGMP – Internet Group Management Protocol
 Multicast distribution trees
 Use of class D addresses 224.0.0.0 – 239.255.255.255
 Essentially Multicasting boils down to sending and
receiving from a common address.
Multicasting challenges within Mobile IP
5
Dynamic Topology
 No fixed underlying infrastructure makes the creation of Multicast
distribution tree difficult.
 The router will keep on asking the question:
“How do I send this message to the group member that is not within my
transmission range???”
Dynamic Group Membership
 Changing infrastructure compromises the “common” destination that
the publisher and subscribers rely upon for data dissemination
So what we need is URRMoM…a.k.a “Your Mom”
6
 It actually stands for user-oriented regional registration based mobile
multicast protocol
 Supratik will explain why “Your MoM” does things way better than
some other “MoM”s
 Fenye will explain SPNP performance model to quantify the claims in
the paper made about “Your MoM”
 Reghu will present the analytical and simulation results followed by the
conclusions
 Let the best “MoM” win!!!!
Some background on “MoM”
7
Two basic schemes were originally proposed by the IETF…
Remote Subscription
 Mobile Host always needs to subscribe to its multicast group when it
enters or changes a foreign networks.
 This essentially means that the current local router of the Mobile host
is made part of the Group subscription tree
•
•
Pros – Optimal trees are constructed for message delivery
Cons – Too many reconstructions of the delivery tree.
Bi-directional tunneling
 Multicast delivery tree stays the same since the Home Agent of the
mobile host is responsible for sending the message out to the Mobile
Host…
•
•
Pros – The Multicast Delivery tree stays the same even though source and destination
nodes hop across networks
Cons – Routing path is not optimal and Foreign agents may receive duplicate packets
“Your MoM”??
8
Is going to combine the advantages of Remote
Subscription with Bi-directional tunneling,
without the disadvantages…
Lets look at some proposed ways...
9
 Local registration to support Multicast services
 Usage of Multicast Agents(MA)
 mMom
- A hybrid approach of Bi-directional tunneling and Remote Subscription
- If Mobile Host is highly mobile BT is used otherwise RS is used
- The FA makes the determination whther MH is mobile or immobile
Cons – Does not factor in co-located care of address in MIP(???)
 RBMoM
- Uses MMA to tunnel packets to the FA serving the MH
- Current MMA information is stored in the MH’s HA
- If MH is out of MMA range, MMA handoff occurs
- Agent Table updates at FA to know which MMAs are around
- MH can look at Agent table to figure out the nearest MMA.
Cons – Excessive Communication Overhead
URRMOM
10
Design Goals of URRMoM:
 No overhead of maintaining Agent tables
 Minimize network traffic generated due to Multicast packet
delivery and Multicast tree maintenance
 Simplicity, Scalability and Efficiency
Key Players of URRMoM
11
 Mobile Multicast Agents (MMA)
•
•
•
Tunneling Multicast Packets to foreign agents
Maintain knowledge about the regional service area (number of subnets covered)
Unsubscribe from the Multicast tree
 Mobile Host (MH)
•
•
•
•
Maintains a counter to record the number of subnets crossed within the service area of
an MMA
Checks if the Foreign Agent of the subnet is part of the Multicast group
Checks if a counter has reached the regional area size
Subscribes to a new MMA
 Foreign Agent (FA)
•
•
•
Receives tunneled messages from an MMA
Forward the tunneled messages to the Mobile host
Act as an MMA for Mobile Hosts
URRMoM Algorithm
12
 MH moves across subnets
 For every move the MH counter increases by one
 If the MH encounters a FA within a subnet which is an MMA for other
Mobile Hosts, the MH will change its regional MMA and set the counter
to zero
 If the MH moves across the regional service area, the new FA becomes
the MH’s new MMA. The FA will subscribe to the Multicast tree in case
its not an MMA originally.
 When the MMA is no longer serving any MHs, it will unsubscribe itself
from the multicast tree
Fig. 1 a Counter reset due to new FA being a MMA. b Counter reset due to service area handoff
13
Regional Service Area
14
 The Mobile Host can make the determination of an optimal
service area that will minimize the network traffic
overhead. This optimal service area (R) will be compared
against the current value of the MH counter to make MMA
handoff decisions
 The Mobile Host can guage the optimal service area based
on the following parameters:
•
•
•
•
Mobility of Mobile Hosts
Number of Mobile Hosts
Size of the network
Topology of the network
Performance Model
15
 A multicast group with a single source;
 The source is a fixed host;
 The multicast group membership does not change dynamically
but mobile members may roam dynamically.
M group members
n by n mesh network
Performance Model
16
 MH’s residence time in a FA is exponentially distributed with
parameters μ. (mobility rate is μ)
 MH’s expected residence time in one FA is 1/μ. MH’s expected
residence time in n2-1 FAs is n2-1/μ. Thus, MH’s inter-arrival time to
any FA is n2-1/μ.
 The arrival rate of a single MH to any FA is λ:
Performance Model
17
 Arrival-departure process of M members with respect to a FA:
 Probability that a MMA does not contain any group member:
 The average number of members being resided under one FA:
Performance Model
18
 Every FA is capable of acting as a MMA.
 Every MH keeps a counter to record number of FAs it has crossed from
its current MMA.
 A MH sets current FA as its new MMA when:


(1) current FA is a MMA;
(2) counter value equals to threshold R. (R is per-MH based depending on it service
and mobility characteristics.)
 Each MMA on average covers R subnets,
group members. Thus,
there are roughly
MMAs in the system.
 The probability that a FA (a MH just enters) is a MMA, denoted by
PMMA:
Performance Model
19
 Performance model for describing a MH’s behavior
Performance Model
20
 We aim to find per-MH based optimal service area R, such that network
traffic cost per time unit is minimized.
Multicast group
management (tree
maintenance) cost,
will decrease as R
increases
Tunneling (multicast
packet delivery) cost,
will increase as R
increases
 The optimal R is per-MH based, depending on MH’s service-to-
mobility ratio.
Performance Model
21
 Multicast tree maintenance per time unit:
Tree subscription /
un-subscription rate
Tree un-subscription rate
Per-hop communication cost
Average number of hops
between MMA and source
Tree subscription rate
Performance Model
22
 Multicast packet delivery per time unit:
Multicast packets delivery rate
Hops from source to MMAs
Per-hop communication cost
Hops from MMAs to MHs
Cost vs. R (varying n)
23
 M=100, τ=0.025s, λp=10, β=15, μ = 0.00167
 Optimal service area size minimized
Cost vs. R (varying MHs)
24
 8x8 Mesh network
 As M increases, optimal R decreases – multicast packet delivery cost dominates
multicast tree maintenance cost
Effect of distance b/w source and MMA
25
• M = 100
 When β increases, the optimal range R increases
 Higher the disatnce, higher the maintenance cost – system prefers to have a
larger service area to reduce rate of tree subscription/un-subscription operations
Comparison – Maintenance Cost
26
URRMoM always produces the least amount of network traffic
compared with RS and RBMoM.
Comparison – Control Message Overhead
27
Comparison – Transmission Delay
28
• URRMoM performs comparably with RBMoM since both use the optimizing R
values
• Basic RS scheme performs the best in terms of packet delay - at the expense
of the much higher maintenance cost and control message overhead
SMPL Simulation – Cost vs. R for varying m
29
SMPL Simulation – Cost vs. R for varying n
30
Conclusions
31
 Combines distinct performance advantages of
remote subscription and bi-directional tunneling.
 URRMoM has simpler system requirements and
less computation complexity than RBMoM.