Transcript Mohanty-TMC06-slide

A Cross-Layer (Layer 2 + 3)
Handoff Management Protocol for
Next-Generation Wireless Systems
By
Shantidev Mohanty and Ian F. Akyildiz, Fellow, IEEE
Presentation By
Muhammed Syyid
NGWS
 Next Generation Wireless System
 Multiple kinds of wireless systems
deployed




UMTS (WAN)
802.11 (WLAN)
Bluetooth (PAN)
Satellite (Global)
 Unification of systems to provide optimal
data availability
NGWS
NGWS Design Goals
 Support for the “best” network selection
 Mechanism to ensure high-quality and
security
 Seamless inter-system mobility
 Scalable architecture (any # of wireless
systems)
 QoS provisioning
Mobility Management
 Location Management
 Track Location of users between
consecutive communications
 Handoff Management
 Keep connections active while moving
between base stations
Handoff in NGWS
Handoff in NGWS
 Horizontal Handoff
 Link Layer Handoff
 IntraSystem Handoff
 Vertical Handoff
 InterSystem Handoff
Current Status
 Link Layer Handoff
 Efficient algorithms available in literature
 InterSystems and IntraSystems
Handoff
 Signaling Delay
 Packet Loss
Goals for Seamless Handoff




Minimize Handoff Latency
Minimize Packet Loss
Limit Handoff Failure
Minimize False Handoff Initiation
Handoff Protocols By TCP/IP Layer
 Network Layer
 Mobile IP
 Transport Layer
 TCP-Migrate
 MSOCKS (Split proxy & TCP SPLICE)
 Modification of SCTP (Stream Control
Transmission Protocol)
 SIP
Mobile IP



Issues



Triangular Routing
High Global Signaling Load
High Handoff Latency


Handoff Requirement Detection
Registration at New Foreign Agent (NFA)

Triangular Routing
Reasons for handoff latency
Proposed Solutions in Literature


Route Optimization




Hierarchical Mobile IP (HMIP)
Cellular IP
IDMP
HAWAII
High Global Signaling Load / Registration at NFA
 Solution to Handoff Latency due to
Requirement Detection
 Use Link Layer Information
 Calculate probability of Handoff
 Factors affecting handoff signaling
delay




Traffic Load on the network
Wireless Link Quality
Distance between user and home network
User’s Speed
Analysis of Current Systems















RSS: Received Signal Strength
BS: Base Station
MT: Mobile Terminal
OBS: Old Base Station
NBS: New Base Station
FA: Foreign Agent
OFA: Old Foreign Agent
NFA: New Foreign Agent
Sth: The threshold value of RSS to initiate handover
Sath: The Adaptive threshold value of RSS to initiate handover
Smin: The minimum value of RSS required for successful
communication
a: Cell Size
d: Cell Boundary
v: Speed of MT’s movement
 : Handoff signaling delay
Movement during Handoff
Handoff Scenario
1) MT moves with speed v
2) RSS of the OBS decreases continuously
3) RSS drops below Sth at the point P marking
cell boundary d.
4) RSS < Sth triggers registration for NFA.
5) Pre-registration messages sent through OBS
to NFA (must be completed before signal
drops below Smin)
False Handoff
 At point p, it can move in any direction  with equal
probability
 F()=1/2 where - <  < 
 Handoff possible only when
 [(- 1, 1)]
Where
1=arctan(a/2)/(d)=arctan(a/2d)
 Probability of False Handoff Initiation is
 Using S=Vt (Distance=Velocity X Time) i.e.
 t=S/V or t=d/v
 The largest possible distance to cover while
travelling to NBS is (a/2)2+(d)2
 As velocity increases the time to cover
distance will decrease
 When the time to leave the cell falls below
the handoff signaling delay, handoff will fail
 Therefore Pf = 1
 Using S=Vt (Distance=Velocity X Time) i.e.
 t=S/V or t=d/v
 As velocity increases the time to cover distance will
decrease
 While the time to leave the cell is greater then
handoff signaling delay, handoff will succed
 When d/v >  the handoff will succeed
 Therefore Pf = 0
False Handoff Initiation
 As cell boundary d is increased, the probability
of false handoff initiation increases (keeping
cell size a constant)
 As cell size a is decreased the probability of
false handoff increases (keeping d constant)
 Cell sizes are currently trending towards
smaller size to cope with capacity and improve
data rates.
 Hence, value of d must be carefully selected.
Handoff Failure and Speed
 From the above, handoff failure depends on
speed (keeping a,d,Sth fixed).
 As speed increases the probability of failure
increases
 For intersystem handoffs the handoff
latency is higher making it more susceptible
to failure
 Increasing the value of d/Sth reduces
the probability of failure
Handoff Failure & Signaling Delay
 The higher the signaling delay the
greater the probability of failure (over
a constant d)
 The higher the value of d the lower
the probability of failure for a single
value for signaling delay
 Therefore to optimize and minimize
handoff failure , the distance (and
therefore Sth) must be adaptive to
signaling delay.
Analysis Summary
 For Fixed value of d(and Sth) handoff
failure probability increases as MT’s
speed increases
 For Fixed value of d(and Sth) handoff
failure probability increases as
handoff signaling delay increases
 Large values of d(and Sth) increase
the probability of false handoff
initiation
CHMP
 Derive information from link layer (2)
and network layer(3) to create
adaptive architecture
 Titled proposed solution as CrossLayer (Layer 2+3) Handoff
Management Protocol or CHMP
CHMP Modules
 Neighbor Discovery Unit
 Determines BS’s neighboring the MT’s current
BS
 Uses network discovery protocols

Speed Estimation Unit
 Uses VEPSD (Velocity Estimation using the Power Spectral
Density of RSS) to estimate speed.
 The doppler frequency is used to determine speed v


V=(c/fc)fm
where c= speed of light in free space
fc = carrier frequency of RSS
fm = maximum doppler frequency
Handoff Signaling Delay Estimation Unit
 Estimates delays associated with intra/intersystem handoffs
 Handoff Trigger Unit
 Collects previously collected and
calculated information to determine
the appropriate time to initiate
handoff
 Handoff Execution Unit
 Triggers the Actual handoff at the
appropriate time calculated by the
Handoff Trigger Unit
Operation
Neighborhood Discovery
 Determine neighbors using the
neighbor discovery unit.
 If OBS and NBS have common FA
link-layer handoff occurs (CHMP is not
used)
 IF OBS and NBS have different FA
(intrasystem) or belong to different
systems (intersystem) CHMP is used.
Handoff Signaling Delay Estimation








Unknown which BS the MT will move to
Using the neighborhood discovery step, compile list of
possible BS/FA’s.
Send an invalid Authentication Extension message to the
GFA (for intrasystem) or HA (intersystem).
GFA/HA respond with an HMIP Registration Reply indicating
registration failure.
The round trip response time is used to estimate the
handoff signaling delay.
Uses existing HMIP protocol without any extra
implementation
Causes extra signaling overhead but solution still improves
performance significantly.
Alternative delay estimation algorithms available in
literature if signaling overhead is not tolerable.
Handoff Anticipation
 When the RSS continuously
decreases, a handoff is anticipated
 Existing movement prediction
techniques used to estimate the next
BS
 Retrieve estimated signaling delay
from the Handoff Delay Estimation
Unit.
Handoff Initiation
 Estimate optimal moment to initiate
handoff
 Estimate Sath using speed and
handoff signaling delay estimates.
 Trigger handoff when RSS < Sath
 Sath for intrasystems is referenced as
Sath1
 Sath for intersystems is referenced as
Sath2
 Pr(x) Received power at point x
 Pr depends on various factors, including
frequency, antenna heights, antenna
gains etc
 d0 is known as reference distance
 Typical values for d0 are
 1 km for macrocells
 100 m for outdoor microcells
 1m for indoor picocells
  is the path loss exponent
 Depends on cell size and local terrain
characteristics
 Typical values range between 3-4 for macro and
2-8 for microcelluar environments
  is a random variable representing
variation in Pr(x) due to shadowing
 Typical value is 8dB
 Pr(x)=Pr(d0)(d0/x) + 
 Sath=10log10[Pr(a-d)]
Handoff Execution
 HMIP registration started when the handoff trigger is
received.
 After registration the MT is switched to the NBS
 Simultaneous Binding preserved for a limited time by
binding CoA of both OFA and NFA to the GFA for
intrasystem and HA for intersystem, this avoids the
ping-pong effect.
 Packets are forwarded to both CoA’s
 If the MT returns to the old BS there is no need to
carry out HMIP handoff again.
 If the MT does not return to the old BS, it deregisters
from the old BS
CHMP Location

Implemented at MT referred to as Mobile Assisted network
controlled Hand Off (MAHO)



MT implemented components



Speed Estimation
RSS Measurement
Handoff Signaling Delay Estimation


Handoff Trigger Unit
Handoff Execution Unit
Network implemented components
Implemented at Network referred to as Network Assisted mobile
controlled Hand Off (NAHO)


Network implemented components



Speed Estimation
RSS Measurement
Handoff Signaling Delay Estimation


Handoff Trigger Unit
Handoff Execution Unit
MT implemented components
Types of Handoffs
 Intersystem
 Macro-Inter: Between a macro-cellular system
and another macro-cellular system
(Inter_MA_HO)
 Micro-Inter: Between a microcellular system and
another microcellular system (Inter_MI_HO)
 Macro-Micro-Inter: Between a macro-cellular
system and a micro-cellular system
(Inter_MAMI_HO)
 Micro-Macro-Inter: Between a micro-cellular
system and a macro-cellular system
(Inter_MIMA_HO)
Usually Microcellular systems are
overlapped by macrocellular systems.
Therefore for Inter_MAMI_HO there is no
handoff failure
 Intrasystem
 Macro-Intra: Between two cells of a
macro-cellular system (Intra_MA_HO)
 Micro-Intra: Between two cells of a
microcellular system (Intra_MI_HO)
Performance Evaluation
 Relationship between Sath and Speed
 Sath increases as speed increases
 That is for a high speed MT handoff should
be initiated early
 Sath increases as  increases
 When  is large the handoff must start earlier
to allow time for registration/handoff to
complete
 In order to compensate for Shadow fading
and errors in estimation, Sath was
increased by 10 percent
 Relationship between Handoff Failure
Probability and Speed
 When MT’s speed is known, there is a 70-80
percent reduction in Handoff Failure Probability
with CHMP
 With CHMP in use probability of failure becomes
independent of speed
 Comparing the figures for fixed RSS thresholds,
failure probabilities are different for intra and
intersystem handoffs
 This further enhances the case for adaptive
thresholds
 Relationship between Handoff Failure
Probability of CHMP and Handoff Signaling
Delay
 There is a 70-80 percent reduction in Handoff
Failure Probability with CHMP compared to fixed
RSS schemes.
 With CHMP in use probability of failure becomes
independent of 
 Probability of failure is limited to desired values
irrespective of speed and variation of handoff
signaling delay
 Relationship between False Handoff
Initiation Probability of CHMP and Speed
 Fixed value of RSS Threshold Sth is calculated
such that a user with highest speed is
guaranteed the desired value of handoff failure
probability.
 Comparatively the adaptive CHMP reduces the
false handoff initiation probability by 5-15
percent
 CHMP initiates handoff while preventing an early
handoff (minimizing false handoff initiation) and
late handoff (minimize probability of failure)
Conclusions
 When a fixed value of RSS threshold (Sth) is
used handoff failure probability increases with
an increase in speed or handoff signaling delays
 The adaptive CHMP Protocol estimates speed
and handoff signaling delay of possible handoffs
creating a dynamic RSS threshold (Sath)
 CHMP significantly enhances the performance of
both intra and intersystem handoffs
 CHMP reduces the cost associated with false
handoff initiation