Transcript itng09-slide

DMAP-FR: Integrated Mobility
and Service Management with
Failure Recovery Support for
Mobile IPv6 Systems
Greg Bilodeau
Mike Reed
What is DMAP-FR?
• An extension of Dynamic Mobility
Anchor Points (DMAP)
• DMAP is an extension of Hierarchical
Mobile IPv6 (HMIPv6)
• HMIPv6 is an extension of Mobile
IPv6
Mobile IPv6
• Mobility in IPv6 Networks.
• MIPv6 is expected to have wide
deployment in the future for all-IP
mobile systems.
• More mobile apps will access
multimedia and data services over
IP
Mobile IPv6 - Advantages over
MIPv4
• Specialized "foreign agent" routers
not needed.
• Support for route optimization
fundamental part of protocol.
• Packets sent to mobile node (MN)
sent using IPv6 routing header
rather than IP encapsulation
• Dynamic home agent (HA) discovery
mechanism returns single reply to
the mobile node.
Mobile IPv6 - Problems
• Does not solve local or hierarchical
forms of mobility management.
• Effective mobility and service
management schemes to reduce
network traffic needed.
• Fault tolerance for service
continuity despite network
router failures.
Hierarchical MIPv6
• Allows local mobility handling.
• Designed to reduce the amount of
signalling traffic between the MN
and home agent (HA) and
correspondent nodes (CNs).
• Utilizes local home agents called
mobile anchor points (MAPs).
Hierarchical MIPv6 - MAPs
Advantages of HMIPv6 and MAPs
• Unlike FA in IPv4, MAPs not required
on every subnet.
• Limit the amount of IPv6 signalling
traffic outside the local domain
• Allow MNs to hide their location from
CNs.
• MN may chose which MAP (or MAPs) to
associate with.
Disadvantages of HMIPv6 and
MAPs
• Static domain in terms of number of
subnets covered.
• Single point of failure.
• While mobility is addressed, service
and performance management is
not considered.
Dynamic Mobile Anchor Points
(DMAP)
• Integrated mobility and service
management.
• MN not only determines which MAP
to bind to, it determines which
access router (AR) acts as a MAP.
• MAP binding based on both mobility
and service requirements of the
specific MN.
Dynamic Mobile Anchor Points
(DMAP)
Location handoff: MN moves across
subnet boundary
Service handoff: MN moves across
DMAP domain boundary
MAP domain size: number of subnets
in region covered by the MAP
Dynamic Mobile Anchor Points
(DMAP)
Dynamic Mobile Anchor Points
(DMAP) - The Tradeoff
Choosing a MAP "further" from the MN
decreases the number of service
handoffs, but increases the triangular
routing overhead and location handoffs
Choosing a MAP "closer" to the MN
reduces the intra-subnet routing, but
increases the frequency of service
handoffs
Dynamic Mobile Anchor Points
(DMAP) - Finding Optimal MAP
MN must be capable of collecting required
statistical information.
Goal is the minimization of "communication
cost" per time unit.
Dynamic Mobile Anchor Points
(DMAP) - Performance Evaluation
Dynamic Mobile Anchor Points
(DMAP) - Performance Evaluation
Calculating MN2DMAP:
F(Mark(Xs)+1) returns the number of hops
between the current subnet and the DMAP
separated byMark(Xs)+1 subnets.
The argument of the F(x) function is added by 1 to
satisfy the initial condition that Mark(Xs) = 0 in
which the DMAP has just moved into a new service
area, so at the first subnet crossing event, the
distance between the DMAP and the subnet is one
subnet apart
Dynamic Mobile Anchor Points
(DMAP) - Performance Evaluation
Calculating NewDMAP:
As MN must inform HA and all N client
nodes of new RCoA
Dynamic Mobile Anchor Points
(DMAP) - Performance Evaluation
Calculating average communication
overhead
Includes delays between CN and
DMAP, DMAP to AR of current subnet,
and wireless link between AR and MN
Dynamic Mobile Anchor Points
(DMAP) - Performance Evaluation
Calculating average location
change overhead
Dynamic Mobile Anchor Points
(DMAP) - Performance Evaluation
Total communication cost per time
unit:
Dynamic Mobile Anchor Points
(DMAP) - Performance Evaluation
DMAP-FR
• Dynamic Mobility Anchor Points –
Fault Recovery
• The addition of fault tolerance to
DMAP.
Failure Recovery Design
• Assume two things:
o For each Access Router there is an overlapping
coverage from other Access Routers since the
failure of an AR will disconnect all Mobile Nodes
attached to it
o in the case that a router(not a Mobility Anchor Point)
fails or a link goes down, it can be handled by the
recovery mechanism of the routing protocol.
Failure Recovery Design Cont'd
• Based on dynamically selecting an Access Router as
the Mobile Anchor Point of a Mobile Node.
• It can recover from two kinds of failures:
o The current Access Router can become the Mobile
Node's DMAP if the DMAP fails
o Access Router failure/recovery can be treated as
disconnection/reconnection. The failure of DMAP can
be detected by not receiving the announcement
message by timeout.
Failure Recovery Design Cont'd
Failure Recovery Design Cont'd
• There are Three Cases for Failure Recovery:
o Failure of MN's DMAP which is not current AR.
o Failure of MN's DMAP which is current AR.
o Failure of MN's current AR.
Failure Modes - Failure of MN's DMAP
which is not current AR
• Suppose that the MN is currently under AR2 and the
current DMAP is AR1 (based on Figure 1).
• In this case, the Current AR becomes the MN's DMAP.
AR2 will inform the HA and CN's that it is now the
DMAP.
Failure Modes - Failure of MN's DMAP
which is current AR
• The MN is under AR1 which is the current DMAP and it
fails. In this case, since the wireless coverage area of
the current AR is overlapping, the MN could be under
radio range of several other subnets.
• The MN will register with a new AR near by which will
become the new MN's DMAP. AR2 will inform the HA
and CN's of the Regional Care of Address change.
Failure Modes - Failure of MN's current
AR
• The MN is under AR3 when AR3 fails and the DMAP is
on AR1.
• In this case the MN locates another AR, i.e. AR1, or
AR2, to replace AR3. The MN will register with the new
AR through a binding message.
Performance Analysis
Performance Analysis cont'd
• Using Stochastic Petri Nets because of:
o their ability to deal with general time distribution for events
o their concise representation of the underlying state machine to deal
with a large number of states
o their expressiveness to reason about a MN's behavior as it migrates
from one state to another in response to events occurring in the
system.
Performance Analysis cont'd
• The number of tokens accumulated in place Xs, that is,
Mark(Xs), represents the number of subnets crossed by
the MN since the MN entered a new service area.
• We allow it to accumulate to K (the subnet size we're
trying to test), at which point we perform a service
handoff.
Petri Net Explanation
• The Mobility rate at which location handoffs occur
is s which is the transition rate assigned to Moving.
• When a Mobile Node moves across a subnet area, a
token is put in place Moves
Petri Net Explanation cont'd
• After moving into a subnet, the Mobile Node obtains a new Care Of
Address, and informs the DMAP of the Care Of Address change.
• This is modeled by enabling and firing transition MN2DMAP while
disabling transition Moving.
• After MN2DMAP is fired, a token in place MOVES flows to place Xs,
representing that a location handoff has been completed and the DMAP
has been informed of the Care of Address change of the Mobile Node.
Petri Net Explanation cont'd
•
•
•
•
If the Number of tokens in place Xs has accumulated to K then it means that
the Mobile Node has just moved into a new service area and a service handoff
ensues.
This is modeled by assigning an enabling function that will enable transition
MovingDMAP after K tokens have been accumulated in place Xs.
After transition MovingDMAP is fired, all K tokens are consumed and place Xs
contains no tokens, representing the action that the DMAP has just moved into
a new service area.
The rate at which transition MovingDMAP fires depends on the cost of
informing the Home Agent and Corresponding Nodes of the DMAP Care of
Address change.
Petri Net Explanation cont'd
• The DMAP alternates between "work" and "Failure" states.
Initially the DMAP is in the work state.
• After some time has elapsed, the DMAP goes to the failure
state.
• This is modeled by transition failing.
• Note that if the DMAP is already in place Failure, transition
failing cannot fire.
Petri Net Explanation cont'd
• While the DMAP is in failure mode, after time has elapsed
representing the recovery time, the DMAP goes to the work
state.
• The is modeled by the transition recovering.
Petri Net Explanation cont'd
• For case 1 the DMAP fails but the current AR is alive, as illustrated in
Figure 1.
• In this case, the current Access Router will become DMAP, the new
DMAP will inform the Home Agent and Corresponding Nodes of the
Regional Care of Address change.
• This is modeled by firing transition recovering with a transition rate
reflecting the cost.
• Firing this transition will flush all the tokens in place Xs as if a service
handoff had happened. This is modeled by a variable input arc from
place Xs to transition recovering.
Petri Net Explanation cont'd
• For Case2, the DMAP fails and the current Access Router
happens to be the DMAP, as illustrated in Figure 1 where the
MN's current AR and DMAP is AR1 and AR1 fails.
• In this case, the MN will register with a new AR near by
• The new AR will become the Mobile Node's DMAP who will
then inform the Home Agent and Corresponding nodes of the
new Regional Care of Address.This event is also modeled by
firing transition recovering.
Petri Net Explanation cont'd
•
•
•
•
•
For Case 3, the current AR fails but the DMAP is alive, as illustrated in Figure
1.
In this case, the Mobile Node will register with another Access Router nearby.
The new Access Router then only needs to inform the DMAP of the Care Of
Address change.
This event can also be modeled by transition recovering.
Note that the rate to transition recovering depends on the system state which
will be characterized later.
Characterizing rate of the Recovering
transition
• When transition Recovering fires, the Mobile Node will contact
the DMAP. If the DMAP fails and the current Access Router is
the MAP, the Mobile Node will register with a new Access
router near by.
• The new access router will become the DMAP and inform the
Home Agent and Corresponding Nodes of the RCoA change.
• If the DMAP fails while the current AR is still alive, the current
AR will become the DMAP.
• In either case the current Access Router chosen becomes the
new DMAP and the cost involved is to inform the Home Agent
and Corresponding nodes.
Characterizing rate cont'd
• Since the new DMAP is F(Mark(Xs)) + g hops away from the
failed DMAP, the cost can be parameterized as
{ N [ b + F(Mark(Xs)) ] + [ a + F(Mark(Xs)) ] + g } t
• The rate transition Recovering is the reciprocal of this quantity.
Overall communication costs
• A Mobile Node and its DMAP determine the service area
dynamically to minimize the overall network signaling costs for
mobility management, service management and fault tolerance
related operations incurred by the Mobile Node. There are
three costs considered:
o
o
o
The service cost
The mobility cost
The failure recovery cost
Overall communication cost
• CTotal = Cservice * l + Cmobility * s + Crecovery * df
CTotal = overall cost incurred per time unit
Cservice = average communication cost to service a
data packet.
o Cmobility = average communication cost to service a
location handoff, including one that can trigger a
service handoff.
o
o
o Crecovery
= the communication overhead for
the network to recover from DMAP or AR
failures.
l = Data packet rate between the Mobile Node and
Corresponding node.
o df = DMAP failure rate
o
Average communication cost to perform
failure recovery
• Ci,recovery =
gt + F(Mark(i) + 1) t
 if the current AR fails while the DMAP is still alive
o gt + at + F(Mark(i) + 1) t + N(bt + F(Mark(i) + 1) t)
 if the DMAP fails
o
Cost versus K
• DMAP-FR has an optimal service area size Kopt to
minimize the overall network traffic cost, when given a
set of parameter values characterizing the mobility and
service behaviors of the Mobile Node and failure
behaviors of Access Routers in the Mobile IP
networks.
Kopt versus df
• Kopt increases as df increases.
• The reason is that as the failure rate increases, the Mobile
Node's DMAP likes to stay at a large service area to reduce
the location handoff cost such that a location handoff will most
likely only involve informing the DMAP of the location change
without incurring a service handoff to migrate the DMAP.
Cost difference between HMIPv6 and
DMAP-FR
• The cost difference between HMIPv6 and DMAP-FR as a function of
Service-to-Mobility Ratio.
• We observe that the cost difference between HMIPv6 and DMAP-FR
degenerates,then sharply rises as SMR continues to increase.
• We conclude that DMAP-FR performs better than HMIPv6, especially
when SMR is high.
Conclusion
• DMAP-FR - efficient mobility and service management
with failure recovery supporting Mobile IPv6
environments.
• Devised a computational procedure to compute the
optimal service area size that would minimize the
overall network traffic cost.
• Compare our scheme with HMIPv6
• Performance gain due to a proper selection of the best
service area dynamically.
References
•
•
•
•
•
•
•
•
W. He and I.R. Chen, "DMAP-FR: Integrated Mobility and Service Management with Failure
Recovery Support for Mobile IPv6 Systems," 6th IEEE International Conference on
Information Technology: New Generation, Las Vegas, April 2009.
I. R. Chen, W. He, and B. Gu. DMAP: A scalable and efficient integrated mobility and service
management scheme for Mobile IPv6 systems. Wireless Personal Communications,
43(2):711–723, 2007.
R. Ghosh and G. Varghese. Fault-tolerant Mobile IP. Technical Report WUCS-98-11,
Washington University, 1998.
D. Johnson, C. Perkins, and J. Arkko. Mobility Support in
IPv6. http://www.ietf.org/rfc/rfc3775.txt, IETF, Work in Progress, 2004.
H. Omar, T. Saadawi, and M. Lee. Supporting reduced location management overhead and
fault tolerance in Mobile-IP systems. In IEEE International Symposium on Computers and
Communications, pages 347 – 353, 1999.
H. Soliman, C. Castelluccia, K. El-Malki, and L. Bellier. Hierarchical Mobile IPv6
mobility management. http://www.ietf.org/rfc/rfc4140.txt, IETF, RFC, 2005.
T. You, S. Pack, and Y. Choi. Robust hierarchical Mobile IPv6 (RH-MIPv6): an enhancement
for survivability and fault-tolerance in Mobile IP systems. In IEEE 58th Vehicular Technology
Conference, pages 2014–2018, 2003.
X. Zhang, J. Castellanos, and A. Campbell. P–MIP: Paging extensions for Mobile IP. ACM
Mobile Networks and Applications, 7(2):127–141, March 2002.
Questions?