Principal Lecturer, Middlesex University

fearlessquickMobile - Wireless

Dec 12, 2013 (3 years and 9 months ago)

76 views

SAT SEMINAR
SERIES

2013

Glenford Mapp

Principal Lecturer, Middlesex University

SAT SEMINAR
SERIES

2013


Brian Ondiege


Ferdinand Katsriku


David Silcott


Jonathan Loo


Haris
Pervaiz



Qiang

Ni

SAT SEMINAR
SERIES

2013


Motivation for the work


Handover Classification


Proactive Handover


Analysis of Urban/Suburban context


Results for Urban/Suburban context


Analysis of Motorway context


Results for Motorway context


Implications for future networking
infrastructure (VANETs, etc)


Future Plans

SAT SEMINAR
SERIES

2013

HARDWARE PLATFORM


(MOBILE NODE)



HARDWARE PLATFORM


(BASE STATION)



NETWORK ABSTRACTION


(MOBILE NODE)



NETWORK ABSTRACTION


(BASE STATION)

VERTICAL HANDOVER

POLICY MANAGEMENT

END SYSTEM TRANSPORT

QOS LAYER

APPLICATION ENVIRONMENTS

CONFIGURATION LAYER

NETWORK MANAGEMENT

CORE TRANSPORT

NETWORK QOS LAYER

SERVICE PLATFORM

CORE NETWORK

PERIPHERAL NETWORK

SAS

NTS

NAS

QBS

SECURITY LAYERS

SAT SEMINAR
SERIES

2013


Can’t explain everything about Y
-
Comm



It’s too big


Several institutions work on Y
-
Comm


Including Middlesex, Cambridge, USP and Lancaster
University


See Y
-
Comm

Research Webpage:


http://www.mdx.ac.uk/research/areas/softw
are/ycomm_research.aspx


This talk looks at handover issues


In particular we are trying to understand the
relationship between handover, the velocity of the
mobile node and mobile infrastructure

SAT SEMINAR
SERIES

2013


Hard
vs

Soft Handovers


Hard
-

break before make


Soft


make before break


Network
vs

Client Handovers


Network


network in control (current)


Client


future (Apple’s patent)


Upward
vs

Downward


Upward


smaller to bigger coverage


Downward


bigger to smaller

SAT SEMINAR
SERIES

2013

HANDOVER

IMPERATIVE

ALTERNATIVE

REACTIVE

PROACTIVE

KNOWLEDGE
-
BASED

MODEL
-
BASED

NETPREF

USERPREF

CONTEXT

SERVICES

UNANTICIPATED

ANTICIPATED

SAT SEMINAR
SERIES

2013


Benefits:


Allows us to minimize disruption due to packet loss
or service degradation during handover by
signalling to the higher layers that a handover is
about to happen


Interested in 2 main parameters


Time Before Vertical Handover (TBVH)


Network Dwell Time (NDT)


the time a mobile spends
in a given network due to mobility

SAT SEMINAR
SERIES

2013

REQ (Time , TBVH, NDT)

A

WIRELESS

NETWORK

TBVH

NDT

A

SAT SEMINAR
SERIES

2013


Proactive policies can themselves be divided
into 2 types


Proactive knowledge
-
based systems


Knowledge of which local wireless networks are
operating at a given location and their strengths at
that point


We also need a system to maintain the integrity,
accessibility and security of that data

SAT SEMINAR
SERIES

2013


Knowledge
-
based approach


Gather a database of the field strengths for
each network around a city


Need to maintain the database and also know
how the results might be affected by seasonal
effects

SAT SEMINAR
SERIES

2013

SAT SEMINAR
SERIES

2013


Uses a simple mathematical model


Defines a radius at which handover should
occur


Finds out how much time I have before I hit
that circle (TBVH), given my velocity and
direction


Used simulation (OPNET)


Can be used in the real world as well as in simulation

SAT SEMINAR
SERIES

2013

Threshold Circle
coverage
Real coverage
Exit coverage
Exit threshold
circle
Handover
threshold circle
SAT SEMINAR
SERIES

2013


Exit Time (ET) is defined as how much time a
mobile node can be in a given network before
it must begin handing over to another
network


ET is primary dependent on NDT which is in turn
dependent on the velocity


T
EH



the time taken to handover to the next
network

SAT SEMINAR
SERIES

2013


If ET is less than or equal to zero, then the
handover to the first network should not take
place as no work will be done because the
interface will be forced to immediately begin
handing over to the next network


This work looks at the effect of this
observation on heterogeneous environments


Need to avoid useless handovers

SAT SEMINAR
SERIES

2013

X

NETWORK A

NETWORK B

SAT SEMINAR
SERIES

2013


This was part of David
Cottingham’s

PhD
work. Handover is dependent on 4 delays:


T
d
is the detection time


time to discover that you
are on a new network


T
c

is the configuration time


time to get and
configure your network interface with a new IP
address called the Care
-
of
-
Address (COA)


T
r

is called the registration time


time taken to
register the new COA with the Home Agent and
Corresponding Nodes


T
a
is called the adaptation time


the time it takes
for the higher layers, such as TCP, to make use of
the bandwidth of the new network


SAT SEMINAR
SERIES

2013

SAT SEMINAR
SERIES

2013


For Reactive Handover we need to add all 4
delays


Because the device is reacting to information from
its interfaces, it is not planning ahead


For proactive handover, we may avoid the
need to add all 4 delays


Because of TBVH, we can signal to the upper layers
that handover will occur after a certain time, so they
could take evasive action, especially at the
transport level


SAT SEMINAR
SERIES

2013


If we assume the use of low
-
level triggers
and IPv6 auto
-
configuration techniques


T
d

and T
c

are effectively zero


So for reactive and proactive handovers we
have:



SAT SEMINAR
SERIES

2013

SAT SEMINAR
SERIES

2013


NDT in a wireless network is given by the
reciprocal of the mobility leave rate. In the
literature, the mobility leave rate is given by:





SAT SEMINAR
SERIES

2013


Assuming circular coverage, we use
propagation models to tell us the handover
radius for different networks.




SAT SEMINAR
SERIES

2013


This is highly dependent on the
transportation model observed by a
population


Must be realistic to get good results


Two main contexts


Urban/Suburban context


Motorway context

SAT SEMINAR
SERIES

2013


Urban/Suburban context


Mobile users are everywhere, both pedestrians,
people in cars (not the driver, of course!)


Cars observe a maximum velocity or speed limit


Cars and people can mingle; traffic lights, people
crossing the road, etc.


Motorway context


No pedestrians, mobile users are in cars


Motorways follow well
-
defined roads



We can work out the exact distance between two
points on a motorway using GPS


Much higher speed limit compared to the
urban/suburban case

SAT SEMINAR
SERIES

2013


Since pedestrians and cars are mingling and
there is a speed limit, V
MAX
, it is reasonable to
set the expected velocity to V
MAX
/2



You cannot know every mobile user’s exact NDT so
you will have to use a probability distribution



So if we plug this into our formula for NDT we
get:


SAT SEMINAR
SERIES

2013


So we found out the expected rate of NDT for
different values of V
MAX


Used an exponential distribution, reasonable
in the urban context


Decided to use simulation to generate results


HANDSIM is a simulation developed by myself
and Eser Gemikonakli to study handover



The team extended it to look at different
velocities and different types of handovers

SAT SEMINAR
SERIES

2013


So the simulation generated handover
requests for different users via a Poisson
distribution


At a given maximum velocity, it generated
handover requests with a given NDT using the
expected value of NDT and the distribution


We then subtracted the handover time for the type
of handover being considered from NDT to get the
Exit Time. If the Exit Time was less than or equal to
zero, that handover request was rejected


We plotted the % rejected handover requests
against the maximum velocity

SAT SEMINAR
SERIES

2013

SAT SEMINAR
SERIES

2013

SAT SEMINAR
SERIES

2013

SAT SEMINAR
SERIES

2013

SAT SEMINAR
SERIES

2013


WLAN handovers do not do well


Much smaller handover radius


Also the time to handover is fairly long compared to
3G, i.e., 4 seconds for WLANs and 1 second for 3G


3G handovers held their own


Fairly large radius


Handover times are fast for 3G compared to WLAN


Proactive handovers did improve the results


Needs further research

SAT SEMINAR
SERIES

2013


Because mobile users are in cars and we
know how to calculate the distance between
two points, this means that we can use a
different approach



We define the Network Dwell Distance (NDD)
as the distance travelled along a motorway
that is in coverage of a given network



NDT = NDD/E(
vel
)

SAT SEMINAR
SERIES

2013


There are two instances:



The straight road: in this context we expect that the
car will travel at or close to the maximum velocity



The other context is when there is a junction and
the car has to slow down to negotiate the junction
so the average velocity will fall and so NDT will
increase





SAT SEMINAR
SERIES

2013

F

C

E

H

NET B

NET A

SAT SEMINAR
SERIES

2013

F

B

Z

C

E

G

H

K

Y

R2

T

w

u

NET B

NET A

v

SAT SEMINAR
SERIES

2013

NET A

F

C

T

NET

B

S

E

H

SAT SEMINAR
SERIES

2013

NET A

F

Z

Y

C

E

S

T

H

NET

B

R

B

G

u

v

w

T

D

SAT SEMINAR
SERIES

2013



At T
-
junctions, cross
-
roads or roundabouts
we normally stop, so we have a expected
velocity of V
MAX
/2



For other junctions we take the cosine of the
angle of the two roads at the junction:


SAT SEMINAR
SERIES

2013



A

B

C

S

T

NET A

NET B

NET C

Scenario

Three WLANs in a single UMTS cell

SAT SEMINAR
SERIES

2013

Analysis

SAT SEMINAR
SERIES

2013

SAT SEMINAR
SERIES

2013

SAT SEMINAR
SERIES

2013

SAT SEMINAR
SERIES

2013

SAT SEMINAR
SERIES

2013

SAT SEMINAR
SERIES

2013

SAT SEMINAR
SERIES

2013


Straight paths have lower NDD and mobile
users travel at close to maximum speed so
these sections tend to have lower exit times


Junction S had the greater exit time because
it had the greater NDD as well as a lower
average velocity


Junction T did not have as much Exit Time as
Junction S because Junction T had a shorter
NDD and faster average velocity

SAT SEMINAR
SERIES

2013


Make the radius of your communication cell
larger



WLAN handover radius is too small


Intelligent Transport Systems


VANET work, Roadside Units (RSU)


Handover Radius is 1 Kilometre


Modified form of 802.11a, higher transmission power



Jonathan Loo and others are doing some research on
VANETs here at Middlesex


So we wanted to see how well this setup
would respond to our methods.


SAT SEMINAR
SERIES

2013

SAT SEMINAR
SERIES

2013

SAT SEMINAR
SERIES

2013


Towards small cells


Allows greater bandwidth


But our work shows that there is an issue with
small cells and mobility


Another way to deal with this is to look at
providing joint coverage along a road or
highway

SAT SEMINAR
SERIES

2013

NET A

NET B

URBAN ROAD

P

Q

SAT SEMINAR
SERIES

2013

NET A

NET B

URBAN ROAD

NET
C

P

P

Q

Q

SAT SEMINAR
SERIES

2013


Intersect distance for no loss of
communication:


PQ >= V
MAX
* T
EH


If we want to support a row of
intersecting cells along a straight road
then:


2R > 2(V
MAX

* T
EH
)


R > (V
MAX

* T
EH
)






SAT SEMINAR
SERIES

2013


Results for other networks (LTE, etc)


What is the effect of velocity on these networks



Handover times and how we could improve
them


Especially in 4G systems


Why is the handover time in WLANs so long!



Smaller cell configuration, user mobility and
networking infrastructure.


Issues of interference,
QoS
, etc.



SAT SEMINAR
SERIES

2013