Transcript WIRE07-slide

DMAP: integrated mobility
and service management
in mobile IPv6 systems
Authors: Ing-Ray Chen
Weiping He
Baoshan Gu
Presenters: Chia-Shen Lee
Xiaochen Ding
Outline






Introduction
Related Work
DMAP
Model
Numerical Results
Applicability and Conclusion
Introduction

MIPv6 - Mobile IPv6


HMIPv6 - Hierarchical Mobile IPv6


A version of mobile IP, it allows an IPv6 node to be mobile
and still maintain existing connections;
Proposed enhancement of MIPv6, it is designed to reduce
the amount of signaling required and to improve handoff
speed for mobile connections;
MAP – Mobility Anchor Point

Serving as a local entity to aid in mobile handoffs, it can be
located anywhere within a hierarchy of routers;
Introduction

HA - home agent


CoA - care of address


A temporary IP address for a mobile node that enables
message delivery when the device is connecting from
somewhere other than its home network;
Location handoff


A router on a mobile node’s home network that maintains
information about the device’s current location, as identified
in its CoA;
Mobile node moves across a subnet boundary;
Service handoff

Mobile node moves across a DMAP domain boundary;
Related Work

MIP-RR


MIP Regional Registration, uses a Gateway
Foreign Agent to provide a regional CoA, which
acts as a proxy for regional movement
management;
The design is for mobility management only
without considering service management-induced
network cost.
Related Work

Hierarchical MIPv6



In HMIPv6, a regional CoA (RCoA) is allocated to
a mobile node, in addition to a CoA, whenever the
mobile node enters a new MAP domain;
MAPs in HMIPv6 are statically configured and
shared by all mobile nodes in the system;
There is no mechanism provided to determine the
size of a MAP domain in HMIPv6 for all mobile
nodes that would minimize the network cost.
Related Work

IDMP


It introduces the concept of domain mobility with a
domain and a domain agent to keep track of CoA
of a mobile node as the mobile node roams within
a domain;
It can be combined with fast handoff mechanisms
utilizing multicasting to reduce handoff latency
and paging mechanisms to reduce the network
signaling cost for intra-domain movements.
DMAP


The essence of DMAP is the notion of
integrated mobility and service management,
which is achieved by determining an optimal
service area size;
The objective is to minimize the total network
signaling and communication overhead in
servicing the mobile node’s mobility and
service management operations;
DMAP

Inter-regional move


The mobile node makes the AR of the subnet as
the DMAP when it crosses a service area, and it
also determines the size of the new service area;
MN acquires a RCoA as well as a CoA from the
current subnet and registers the address pair to
the current DMAP in a binding request message;
DMAP

Inter-regional move


The MN also informs the HA and CNs of the new
RCoA address change in another binding
message so that the HA and CNs would know the
MN by its new RCoA address;
DMAP intercepts the packet destined for RCoA,
inspects the address pair stored in the internal
table, finds out MN’s CoA and forwards the packet
to the MN through tunneling;
DMAP

Intra-regional move

When the MN subsequently crosses a subnet but
is still located within the service area, it would
inform the MAP of the CoA address change
without informing the HA and CNs to reduce the
network signaling cost;
DMAP
DMAP



A MN’s service area can be modeled as
consisting of K IP subnets;
The MN appoints a new DMAP only when it
crosses a service area whose size is
determined based on the mobility and service
characteristics of the MN in the new service
area;
The service area size of the DMAP is not
necessarily uniform;
DMAP


A large service area size means that the
DMAP will not change often, while a small
service area size means that the DMAP will
be changed often so it will stay close to the
MN;
There is a trade-off between two cost factors
and an optimal service area exists;
DMAP

The service and mobility characteristics of a
MN are summarized by two parameters:



The resident time that the MN stays in a subnet,
represented by using the MN’s mobility rate σ;
The service traffic between the MN and server
applications, represented by using the data
packet rate λ;
The ratio of λ/ σ is called the service to
mobility ratio (SMR) of the MN;
Model

We devise a computational procedure to
determine the optimal service area size


The intent to find the optimal service area based
on the MN’s mobility and service behaviors
The computational procedure requires


Every AR must be capable of acting as a MAP
Each MN must be powerful enough to collect data
dynamically and perform simple statistical
analysis
Model

We aim to minimize the communication cost




The signaling overhead for mobility management
for informing the DMAP of the CoA changes
Informing the HA and CNs of the RCoA changes
The communication overhead for service
management for delivering data packets between
the MN and CNs
Our SPN model is shown later
Model
Symbol
Meaning
λ
Data packet rate between the MN and CNs
σ
SMR
Mobility rate at which the MN moves across
subnet boundaries
Service to mobility ratio (λ/σ)
N
Number of server engaged by the MN
F(K)
A general function relating the number of
subnets K to the number of hops
Model
Symbol
Meaning
K
Number of subnets in one service area
τ
α
1-hop communication delay per packet in
wired networks
Average distance between HA and MAP
β
Average distance between CN and MAP
γ
Cost ratio between wireless vs. wired
network
Stochastic Petri Net
K
(Guard:Mark(Xs)<K-1)
tmp
Pi=1
A
K
Moves
MN2DMAP
Move
Xs
Pj=1
NewDMAP
B
(Guard:Mark(Xs)=K)
(Guard:Mark(Xs)=K-1)
A token represents a subnet crossing event by the MN
Stochastic Petri Net-Places
K
A temporary place
holds tokens from
transition A
Mark(Xs) holds
the number of
subnets
crossed in a
service area
Mark(Moves)=1
means that the (Guard:Mark(Xs)<K-1)
MN just moves
aross a subnet
tmp
Pi=1
A
K
Moves
MN2DMAP
Move
Pj=1
Xs
NewDMAP
B
(Guard:Mark(Xs)=K-1)
(Guard:Mark(Xs)=K)
Stochastic Petri Net-Transitions
A guard for transition A that is enabled if a move will not cross a service area
A timed
K
A timed transition for the
transition for the
A timed
MN to inform the DMAP of
MN to inform
transition for the
the CoA change
the HA and CNs
MN to move
(Guard:Mark(Xs)<K-1)
of the RCoA
across subnet
change
areas
tmp
Pi=1
A
K
Moves
Xs
MN2DMAP
Move
Pj=1
NewDMAP
B
(Guard:Mark(Xs)=K)
(Guard:Mark(Xs)=K-1)
A guard for transition B that is enabled if a move will cross a service area
Transition Rate

MN2DMAP
Wireless one-hop
communication
delay per packet
1
  F (Mark (Xs )  1) 
The number of hops
between the current
subnet and the DMAP
seperated by
Mark(Xs)+1 subnets
Transition Rate

NewDMAP

The communication cost includes that for the MN to
inform the HA and CNs of the new RCoA change
1
  N
The average
distance in
hops
between the
MN and the
HA via wired
network
The average
distance in
hops
between the
MN and N
CNs via wired
network
Cost of Service Management


Pi: The steady-state probability that the
system is found to contain i tokens in place
Xs such that Mark(Xs)=i
Ci,service: The communication overhead for the
network to service a data packet when MN is
in the i-th subnet in the service area
C
service
K
K
i 0
i 0
  ( Pi  Ci ,service )       ( Pi  F (i ) )
A delay in the
wireless link
form the AR to
the MN
A delay between
the DMAP and a
CN in the fixed
network
A delay from DMAP to the
AR of the MN’s current
subnet in the fixed network
Cost of Location Management

Ci,location: The network signaling overhead to
service a location handoff operation given the
MN is in the i-th subnet in the service area

If i < K


Only a minimum signaling cost will incurred for the MN
to inform the DMAP of the CoA address change
If i = K


The location handoff also triggers a service handoff
A service handoff will incur higher communication
signaling cost to inform the HA and N CNs of the
RCoA address change
Cost of Location Management
K
C
location
  ( Pi  Ci ,location)
i 0
K 1
 PK (    N )   {Pi (  F (i ) )}
i 0
A location handoff and
a service handoff
A minimum signaling cost
for the MN to inform the
DMAP of the CoA address
change
Cost of DMAP

Summarizing above, the total communication
cost per time unit for the Mobile IP network
operating under our DMAP scheme to service
operations associated with mobility and
service management of the MN is calculated
as:
C
DMAP
 Cservice    Clocation 
Service
management
cost
Mobility
management
cost
Numerical Results

We calculate CDMAP as a function of K and
determine the optimal K


K represents the optimal “service area” size
The size will minimize the network cost given


A set of parameter values charactering the MN’s
mobility and service behaviors
We present results to show that


There exists an optimal service area under DMAP
Demonstrate the benefit of DMAP over basic
MIPv6 and HMIPv6
Numerical Results

MIPv6
MIPv6
Cservice
   
A delay in the
wireless link
from the AR
to the MN
MIPv6
Clocation
     N
A communication
delay from the
CN to the AR of
the current
subnet
A delay in the
wireless link from
the MN to the AR of
the subnet that it
just enters into
CMIPv6  C
MIPv6
service
 C
MIPv6
location
A delay
from
that AR
to the
HA

A delay
from that
AR to the
CNs
Numerical Results

HMIPv6


The placement of MAPs is predetermined
Each MAP covers a fixed number of subnets


A MN crosses a subnet within a MAP


KH = 4
It only informs the MAP of its CoA
A MN crosses a MAP



Changes the MAP
Obtain a new RCoA
Informs the HA and CNs of the new RCoA
Numerical Results
Comparing DMAP with basic MIPv6 and HMIPv6 head-to-head from the
perspective of Kopt
Numerical Results
Cost difference between basic MIPv6, HMIPv6, and DMAP
Numerical Results
Effect of α and β on CHMIPv6 − CDMAP
Numerical Results
Effect of F(k) on CHMIPv6 − CDMAP
Applicability and Conclusion



We proposed a novel DMAP scheme for integrated
mobility and service management
To apply the analysis results in the paper, one can
execute the computational procedure at static time
to determine optimal Kopt over a possible range of
parameter values
In the future, we plan to consider the implementation
issue by building a testbed system to validate the
analytical results as well as testing the sensitivity of
the results with respect to other time distributions
other than the exponential distribution used in the
analysis
Q&A