1
DEMAPS: A Load-Transition-Based Mobility Management Scheme
for an Efficient Selection of MAP in Mobile IPv6 Networks
Tarik Taleb, Member, IEEE, Abbas Jamalipour, Fellow, IEEE,
Yoshiaki Nemoto, Senior Member, IEEE, and Nei Kato, Senior Member, IEEE
IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 58, NO. 2, FEBRUARY 2009
2
Outline • Introduction• Related Work• DEMAPS Scheme• Performance Evaluation• Discussion• Concluding Remarks
3
Introduction • Mobile IP (MIP) – an important protocol for accommodating the IP mobility.
• Hierarchical Mobile IPv6 (HMIPv6) – locally handle handovers by the usage of an entity called
mobility anchor point (MAP). – overcome the excessive delay and signaling involved when
handover– a more efficient way for mobility management in IP networks,
4
• HMIPv6 overview
5
Introduction (cont’d)• For large mobile network, HMIPv6 does not control traffic among multiple
MAPs – if the selected MAP is overloaded, extensive delays are experienced during the
routing process.
• the traditional distance-based MAP selection scheme– The MN select the MAP that is most distant, provided that its preference value is
not zero (RFC 5380)– Some MAPs may overly congested with packets, whereas others are underutilized.
• Dynamic Efficient MAP Selection (DEMAPS) – works similar to HMIPv6 when the network is not overloaded.– When the network gets heavy loads, the selection of MAPs is based on an
estimation of MAP load transition • using the exponential moving average method.
6
Outline • Introduction• Related Work• DEMAPS Scheme• Performance Evaluation• Discussion• Concluding Remarks
7
Related Work• To reduce handoff-signaling delays in macro mobility, the
central theme in the pioneering studies pertains to adopting hierarchical management strategies by using local agents– to localize the binding traffic
• Determining the optimal size of local networks is one of the most challenging tasks1. the average total location update and packet delivery cost2. mobility patterns, registration delays, and the CPU processing
overhead
8
• some local agents get congested with traffic, whereas other agents are not efficiently utilized. should be dynamic– ex. Dynamic domain list
• deliver packets to users via multiple levels of ARs. leads to long packet delivery delay and congestion of the
selected ARs with redundant traffic.
• One possible solution to this issue is to reduce the size of the subnet domains.
lead to frequent interdomain handoffs and, consequently, excessive BU cost.
9
Related Work (cont’d)• Referring to the mobility pattern– Velocity: if high, register to a MAP at higher level
• Fixed and Dynamicconsists of high accuracy in estimating the velocity, not always simple,
– moving range(area)• how the moving range of each MN can be defined, • how the scheme can be applied to MNs that keep changing their moving
areas• longer delivery delay
– the session-to-mobility ratio (SMR)• the ratio of the session arrival rate to the handover frequency. • Small SMR high MAP is selected
10
Outline • Introduction• Related Work• DEMAPS Scheme• Performance Evaluation• Discussion• Concluding Remarks
11
DEMAPS Scheme overview
Optimum MAP
vdynamic MAP discovery
No change on MN save battery life
If OMAP!=Previous MAP, then inter-domain handover
Router Sol.
Router Adv.
v
12
load of a MAP_to_AR link• each access point receives MAP option messages from high-layer MAPs
every Δ period of time• Define the load of the ith MAP, as shown in its kth downstream node,
at the nth time slot as Mi→k[n]
– Ci→k : the data processing speed of the ith MAP on the link to its kth downstream node
– pi→k[n] : the total number of data packets that are forwarded on the link as a mere router at the nth time slot
– qi→k[n] : the number of data packets that are destined to MNs registered with the ith MAP at the nth time slot
– W : a weight factor for reflecting the difference in pi→k[n] and qi→k[n]
13
predicting future transitions in a MAP’s loads
• exponential moving average (EMA)– a cut-and-dry approach for analyzing and predicting
performance– easy to implement and requires minimal computational load
• the EMA value of the ith MAP load with respect to its kth downstream node at the nth time slot:
r is set to 0.9
14
• Ei→k[n] < Mi→k[n]– the load of the ith MAP on the link to its kth downstream node
has more tendency to increase [i.e., the load increase (LI) tendency]
• Ei→k[n] > Mi→k[n]– the MAP load on the same link may likely decrease [i.e., the
load decrease (LD) tendency].
• MAP 將 load index 與 tendency 放在 RA 的 MAP option中送出
15
deciding the most appropriate MAPs• ARs decide the most appropriate MAPs for future visiting
mobile users
Stage 1: When the network is not overloaded– MAPs with loads < β = 80% are sorted– The farthest MAP among the sorted MAPs is selected first
Stage 2: the loads of all MAPs > β– MAPs with LD at higher hierarchy are preferably selected
• create large MAP domain for MNs so that their future handoffs can locally be handled
– If all MAPs have LI, select the higher hierarchy MAP with the minimum traffic load
16
Simulation Setup• Use QualNet 4.0
for simplicity
Over each inter-MAP link, the rate of the background traffic is randomly chosen from within 40% to 70% of the link capacity
17
• two mobility models – the random waypoint model– the group mobility model
• four groups are simulated, with each consisting of 12 or 13 MNs– the MN speed is set to [0, 2] m/s, and the pause time is set to null
• Compare DEMAPS to – HMIPv6
• The farthest MAP (preference!=0) is selected– HMIPv6-UP
• velocity-based MAP selection scheme in [24]• if the velocity > 0.5 m/s, the MN registers with an upper MAP
[24] E. Natalizio, A. Scicchitano, and S. Marano, “Mobility anchor point selection based on user mobility in HMIPv6 integrated with fast handover mechanism,” in Proc. IEEE WCNC, New Orleans, LA, Mar. 2005, pp. 1434–1439.
18
Simulation Results• Load transitions
HMIPv6 HMIPv6-UP DEMAPS
rand
om w
aypo
int
grou
p m
obili
ty
19
• In HMIPv6, when MNs roam from one AR to another AR, each MN always selects the same MAP that was previously used– The same in HMIPv6-UP when the MN’s mobility pattern does not change.
• DEMAPS increase the throughput and reduce the end-to-end delay– mostly attributed to the selection of the most appropriate MAP– Higher throughput MAPs with LD tendencies are preferably selected
grou
p m
obili
ty
20
aggregate performance
• both UDP data packets and signaling packets are plotted• the overall bandwidth consumption in DEMAPS is almost the same as
that of HMIPv6 and HMIPv6-UP– the additional cost due to signaling packets is minimal
• DEMAPS achieves better distribution of traffic load among the MAPs. – Almost no packet drop
• mostly due to the transmission of packets over congested links upon handoff
grou
p m
obili
ty
21
binding updates
• HMIPv6 and HMIPv6-UP reduce the frequency of BU messages to HAs– however, many data packets drop at congested links
• the average BU latency for some MNs in the case of HMIPv6 and HMIPv6-UP is higher than in the case of DEMAPS– due to the selection of heavily loaded MAPs high queuing delays
22
the effect of changing Δ
• Φ captures the efficiency of traffic distribution over the network– αi is the number of packets that were processed by the ith link – N is the number of inter-MAP links
• (Δ = 1 s) represents a good tradeoff between an efficient distribution of data traffic and a reduced frequency of MAP option packets
23
• small values of Δ consists of the guarantee of high prediction accuracy of the EMA method
Δ=1 s Δ=10 s
24
the effect of changing r
• the system guarantees an efficient traffic distribution for all the values of r
• reaches its optimum when r takes large values (=1)
25
Discussion • DEMAPS
– neither generates any new signaling packets, – modifies the HMIPv6 protocol itself,– nor require any modifications at the mobile terminals
• In DEMAPS, the frequency of re-registering to HA is high• Solution:
– MNs should be given freedom in choosing the MAPs to which they register– If the load of the old MAP is not at a critical point, the MN can keep registering with it
• Selecting the MAP may require some energy at the MN– This operation is, however, performed only upon handoff
• Solution:– implement MAP decision-making mechanism at ARs– the MN could notify the AR of the old MAP via the RS message
26
Discussion (cont’d)• The working of DEMAPS can further be enhanced by
anticipating the occurrence of MN handoffs.
• DEMAPS can easily be applied with minor modifications to the networks with mobile routers– as used for seamless Internet access in public transportation
and in wireless metropolitan networks
27
Concluding Remarks• DEMAPS significantly improves the performance of HMIPv6
in large mobile networks– a dynamic efficient technique for selecting the most appropriate
MAP for registration– based on an estimation of MAP load transition using the EMA
method– easy to implement– the additional cost that signaling packets requires is proven to be
minimal• Extensive simulation results have demonstrated that DEMAPS
improving the average communication delay, reducing the number of losses, and making better utilization of the network resources.
28
comments• DEMAPS is simple but effective• Can be applied to Local Mobility Anchor selection in
Proxy Mobile IPv6• Knowing the MAP list of new AR before handover can
improve handover performance– In the case of PMIPv6, 也可選擇由 the new MAG or the new
LMA 來 initiate handover-performance improving 的各項operation