Chapter 8: Deployment & Data

Access MAC Scheduling

Protocols on VANETs
Optimal Placement of Gateways in
Vehicular Networks
Yuguang Fang, Xiaoxia Huang, Pan Li, and
Ph
Phone Li
Lin
Department of Electrical and Computer Engineering, University
of Florida
D
Department
t t off Computer
C t Science
S i andd Information
I f ti Entineering,
E ti i
National Taiwan University

IEEE Transactions on Vehicular

Technology, TVT 2007
Introduction

Mobile users in vehicles require to access the Internet.

How to place gateways to link mobile nodes to the Internet ?
Goals

Optimally placing gateways.

Minimize the average number of hops from APs to gateways.

Minimize the total power consumption.

Maximize the average capacity of each AP.

1-D Vehicular Networks
Assume that several APs have been deployed.
Which AP should play the role of gateway?

INTERNET
 One Gateway
y

1 2 x n-1 n

1
H ( x) [( x 1) ( x 2) ... 1 0 1 2 ... (n x)]
n
Average number of hops

1 n 2 1
2
n 1
H ( x) x
n 2 4
To minimize H(x), we have

n 1 n 1
x orx
2 2
One Gateway

1 2 3 4 5 6 7 8 9

n 1
n9 x 5
2

1 2 3 4 5 6 7 8

n 1
n 8 x 4
2
Multiple
p Gateways
y
How to select m gateways from n APs?
n

x1

x2

xm

1 n

n1 n2 nm

1 ni 1
2
ni2 1
H i ( xi ) xi
ni 2 4

A
Average number
b off hhops off each
h group.
 Multiple Gateways
n
Let C be the number of APs in a group
m

Case1: C is an odd integer:

C 1
xi (i 1)C
2
Example: n=15, m=3, C=5
C

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

5 1
x1
2
 Multiple Gateways
n
Let C be the number of APs in a group
m

Case2: C is an even integer:

C
xi (i 1)C
2
Example: n=16, m=4, C=4
C

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

4
x1
2
 Multiple Gateways
n
Let C be the number of APs in a group
m

Case3: C is not an integer but exists k(1km) such that C=(n-k)/m is an odd integer :

C 1
(i 1)C , if 1 i m k 1
xi 2
xm k 1 [i (m k 1)](C 1) , if m k 2 i m

Example: n=18, m=5, C=3

C C C+1 C+1 C+1

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

C C C+1 C+1
 Multiple Gateways
n
Let C be the number of APs in a group
m

Case3: C is not an integer but exists k(1km) such that C=(n-k)/m is an even integer :

C
(i 1)C , if 1 i m k
xi 2
xm k [i (m k )](C 1) , if m k 1 i m

Example: n=22, m=5, C=4

C C C C+1 C+1

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22

C C C+1 C+1
Energy Balance

GatewayAP
Energy Balance

T. Rappaport, Wireless Communications: Principles and Practice, 2nd ed.

Prentice-Hall PTR, 2002.

Pt transmitted power
Pr received power
Gt antenna gains for the transmitter
Pt Gt Gr
2
Pr Gr antenna gains for the receiver
(4 ) 2 Ld
L( L 1) the system loss factor
the wavelength
g
the path loss exponent
 Energy Balance
Given n APs and the gateway has been known,
how to determine the distance of neighboring APs for achieving energy balance?

Pt Gt Gr 2 (4 ) 2 L
Pr P P d Td
(4 ) 2 Ld Gt Gr 2
t r

1 2 x-1 x n-x n-1 n

P (d1 d 2 ... d x1 )T (d 2 ... d x1 )T ... d x1T d x T ... (d x d x1 ... d n1 )T

GatewayAP
y AP !
One Gateway

P (d1 d 2 ... d x1 )T (d 2 ... d x1 )T ... d x1T d x T ... (d x d x1 ... d n1 )T

P {[d1 2d 2 ... ( x 1)d x1 ] [d n1 2d n 2 ... (n x)d x ]}T

x 1 n x
P id i jd n j T
i 1 j 1
One Gateway

x 1 n x

i n j
id jd

1

x 1 n x n 1
id i jd n j
i 1 j 1

n 1 i 1 j 1

1
x 1 n x n 1
P (n 1) id i jd n j T
i 1 j 1

The equality holds when id

jd r
n j for all
i [1, x 1] and j [1, n x]
x 1 n x i
P id i jd n j T
i 1 j 1
One Gateway

x 1 n x
Since d d
i 1
i
j 1
j L

The minimum value of P can be achieved when

i 1 i i x ( n i )
x 1 1 n 1 1

1
L , for i 1
1
d i i d1 , for i [ 2, x 1]
1
( n i ) d1 , for i [ x, n 1]

GatewayAP
y AP !
Multiple Gateways

1 x1
xi
xm n

S1 t1 Si ti Sm tm
P {[d1 2d 2 ... ( x1 1)d x1 1 ] [d t1 1 2d t1 2 ... (t1 x1 )d x1 ]}T
{[d t1 1 2d t1 2 ... ( x2 1)d x2 1 ] [d t2 1 2d t2 2 ... (t 2 x2 )d x2 ]}T
...
{[d tm1 1 2d tm1 2 ... ( xm 1)d xm 1 ] [d n1 2d n 2 ... (n xm )d xm ]}T

m xk 1 t k xk

P id tk 1 i jd tk j T

k 1 i 1 j 1
Multiple Gateways

1
xk 1 tk xk
m n 1
P (n 1) id tk 1 i jd tk j T

k 1 i 1

j 1

The equality holds when id tk 1 i jd trl j

m xk 1 t k xk
1] j [1, t x ]
P idfor
t k 1 i jd
all i [1 , xk t j ,T
k k k , and k, l [1, m]
k 1 i 1 j 1
Multiple Gateways

m xk 1 t k xk

Since d t k 1 i d t k j L

k 1 i 1 j 1

The minimum value of P can be achieved when

k 1 i 1 i j xk (t k j )
m x k 1 1 t k xk 1

1
L , for i t k 1
1
d i i d1 , for i [t k 2, xk 1 1]
1
(t k 1 i ) d1 , for i [ xk 1 , t k 1 1]

2-D Vehicular Networks

INTERNET
 One Gateway
g ( x, y ) denotes the location of the ggatewayy
( x1 , y1 ) ( xi , yi ) denotes the location of AP i
r ( n ) denotes the common transmission range for all n APs

( x x1 ) 2 ( y y1 ) 2

The number of hops ( x x )2 ( y y )2

from APi to gateway i i

r (n
( )
( x, y )

( x x3 ) 2 ( y y3 ) 2
To minimize H(g), we have

n

Minimize f ( x, y ) ( x xi ) 2 ( y yi ) 2
( x x2 ) 2 ( y y 2 ) 2

( x3 , y3 ) i 1

( x2 , y 2 )
One Gateway
1. Initialization:
1 n
( x1 , y1 ) x xi
n i 1
1 n Not the ideal location
y i 1 yi
n
2. Iteration: calculate Scos and Ssin, where
n
( xi x)
Ideal location Scos
i 1 ( x xi ) 2 ( y yi ) 2
n
( yi y )
Initialization location Ssin
i 1 ( x xi ) 2 ( y yi ) 2
( x, y ) 3. IF |Scos|Threshold and |Ssin|Threshold
4. Go to 9.
5. Else
6. x=x+step*Scos and y=y+step*Ssin
7. Go to 2.
( x2 , y 2 ) ( x3 , y 3 ) 8. END IF
9. Finish
One Gateways
Multiple Gateways

INTERNET
Discussion-Power

P Pt H (G )
Pt is the transmission power, which is the same for all APs,
G is the gateways in the network
H(G) is the average number of hops from an AP to a gateway

When H(G) is minimized,

minimized the average transmission power P is also minimized
minimized.
Discussion-Capacity

H (G ) n C
denotes the AP generates traffics.
C denotes the capacity of the network.
H(G) is the average number of hops from an AP to a gateway.
n is the total number of APs in the network.

Therefore the capacity available to each node,

Therefore, node ,
is bounded by
C C

nH (G ) nH mini (G )

When H(G) is minimized, the capacity of APs can be maximized.

 C = 60 Mbps
p
Discussion Capacity
Discussion-Capacity n=5

1 2 3 4 5

H(G) = 1.2 nH(G) = 6

= 10 Mbps

1 2 3 4 5

H(G) = 2 nH(G) = 10
= 6 Mbps

When H(G) is minimized, the capacity of APs can be maximized.

Conclusion

Optimally placing gateways.

Minimize the average number of hops from APs to gateways.

Minimize the total power consumption.

Maximize the average capacity of each AP.

O S
On Scheduling
h d li V Vehicle-Roadside
hi l R d id
Data Access

Yang Zhang , Jing Zhao and Guohong Cao

Department of Computer Science & Engineering
The Pennsylvania State University

ACM VANET 2007

Introduction The Architecture of Vehicle (Roadside service)
Vehicular Ad-hoc NETworks includes
Moving Vehicles
Vehicles, and
Roadside Units (RSU).
Local broadcasting information
IEEE 802.11 access point

RSU

RSU
Introduction The Architecture of Vehicle (Roadside service)

Applications of the RSU

Commercial Advertisement
R l Ti Traffic
Real-Time T ffi
Digital Map Downloading
Introduction Challenges
Bandwidth Competition
All requests (upload/download) compete for the
same limited bandwidth
Time Constraint
Vehicles are moving and they only stay in the RSU
area for a short period of time
Data Upload / Download
The miss of upload
leads to data staleness
Problem
How to develop a scheduling algorithm for
data access?

Performance
f Metrics
i (Tradeoff )
Service Ratio ()
Ratio of the number of requests served before the service
deadline to the total number of arriving requests.
D t Q
Data Quality
lit ()
Percentage of fresh data access
System Model
Each request is characterized by
v-id : the Vehicle-ID
d-id : the Data-ID
op : upload / download
deadline : Time Constraint

( v-id , d-id , op , deadline )

System Model
Assumptions
Each vehicle knows the service deadline of its request.
Location-aware (GPS)
Deadline-aware (driving velocity)
The RSU maintains a service cycle
Service non-preemptive
 Naive Scheduling Policies
First Come First Serve (FCFS)
Fi Deadline
First D dli FiFirst (FDF)
Smallest Datasize First (SDF)
( )

FDF

SDF

Scheduling Scheme I
D
D*S
S Scheduling
D*S Scheduling
D:D D-List
Li t (Deadline-list)
(D dli li t)
S : S-List (DataSize-list)

D-List S-List
Current Clock = 100
(Deadline) (Data Size)

a 101 a 50 e

b 102 b 100 c

c 105 c 150 b
Sort by Sort by
d 106 Deadline d Data Size 200 f

e 110 e 350 a

f 115 f 500 d

g 116 g 1000 g

Requests in RSU
Scheduling Scheme I
D
D*S
S Scheduling
Basic Idea
Given two requests with the same deadline, the
one asking for a small size data should be served
first.
Given two requests asking for the data items with
same size, the one with an earlier deadline should
be served first.
Assign each arrival request a service value based
on its deadline and data size,
size called DS_value
DS value as
its service priority weight
DS value = (Deadline CurrentClock) * DataSize
DS_value
Scheduling Scheme I
D
D*S
S Scheduling Implementation
Information in RSU

D-List S-List
(Deadline) (Data Size)
a 101 50 e

b 102 100 c

c 105 150 b

d 106 200 f

e 110 350 a

f 115 500 d

g 116 1000 g
Current Clock = 100
 Scheduling Scheme I
D
D*S
S Scheduling Implementation

MinDS Step 1
= (Deadline - CurrentClock) * DataSize
Current Clock = 100
= (101-100) * 350 = 350

DS_Value
DS Value 350
D-List S-List
D-List entry request b
(Deadline) (Data Size)
request b DS_Value
a 101 50 e DataSize MinS

b 102 100 c MinS = 350 / (102-100) = 175

c 105 150 b
MinS = 175
d 106 200 f

e 110 350 a

f 115 500 d

g 116 1000 g
 Scheduling Scheme I
D
D*S
S Scheduling Implementation

MinDS Step 1
= (Deadline - CurrentClock) * DataSize
= (101-100) * 350 = 350

D-List S-List
(Deadline) (Data Size) Current Clock = 100

checked a 101 50 e

cut standard b 102 100 c

c 105 150 b
MinS = 175
d 106 200 f

e 110 350 a

f 115 500 d

g 116 1000 g
Scheduling Scheme I
D*S Scheduling Implementation

Step 2
Current Clock = 100
DS_Value (request e)
DS_Value 350 = (Deadline - CurrentClock) * DataSize
SS-List
List entry request cc = (110-100)
(110 100) * 50 = 500 > MinDS = 350
request c DS_Value
Deadline MinD D-List S-List
350 > ( MinD
in 100
00 ) * 100
00 (Deadline) (Data Size)

MinD = 350 / 100 + 100 = 103.5 a 101 50 e

b 102 100 c
Mi D = 103.5
MinD 103 5
c 105 150 b

d 106 200 f

e 110 350 a

f 115 500 d

g 116 1000 g
Scheduling Scheme I
D
D*S
S Scheduling Implementation

Step 2

DS_Value (request e)
= (Deadline - CurrentClock) * DataSize
= (110-100)
(110 100) * 50 = 500 > MinDS = 350

D-List S-List
Current Clock = 100 (Deadline) (Data Size)

a 101 50 e

b 102 100 c
Mi D = 103.5
MinD 103 5
c 105 150 b

d 106 200 f

e 110 350 a

f 115 500 d

g 116 1000 g
Scheduling Scheme II
Download Optimization: Broadcasting

Observation
several requests may ask for downloading the same
data item
wireless communication is broadcast in nature
Basic Idea
delay some requested data and broadcast it before
the deadlines,, then several requests
q mayy be served
via a single broadcast
the data with more pending requests should be
servedd first
fi t
DSN_value = ( Deadline CurrentClock ) * DataSize/Number

The number of pending

requests for the same data (N)
Scheduling Scheme II
Download Optimization: Broadcasting

When calculating their DSN value, we need to assign each

pending request group a single deadline to estimate the urgency
of the whole group
group.

Not too much impact !

Scheduling Scheme II
The
Th PProblem
bl off DSN

Data Quality will be blocked

DSN_value = (Deadline CurrentClock) * DataSize / Number
For upload request
request, it is not necessary to maintain several update
requests for one data item since only the last update is useful.
Number value of update requests is always 1, which makes it not
fair for update requests to compete for the bandwidth .
D*S/N can improve the system service ratio but block the service
opportunity of update requests, which degrades
d d the h ddata
quality for downloading.
Schedulingg Scheme III
Upload Optimization (2Step Scheduling)

Basic Idea
two priority queues: one for the update requests and
the other for the download requests.
the data server provides two queues with different
bandwidth (i.e., service probability)
Simulation Setup

NS-2
NS 2
400m*400m square street scenario
One RSU server is located at the center of two 2-way y roads
40 vehicles randomly deployed on each lane
Each vehicle issues request with a probability p
Access pattern of each data item follows Zipf distribution
Effect of Workload

As workload
A kl d iincreases, D*S/N can achieve
hi the
th hi
highest
h t
service ratio while its data quality degrades dramatically
Effect of Workload

As workload
A kl d iincreases, D*S/N can achieve
hi the
th hi
highest
h t
service ratio while its data quality degrades dramatically
Effect of Workload

As workload
A kl d iincreases, D*S/N can achieve
hi the
th hi
highest
h t
service ratio while its data quality degrades dramatically
 C l i
Conclusion
This paper proposed a basic scheduling scheme called D*S to
consider both service deadline and data size when making
scheduling
h d li decisions.
d ii
To make use of the wireless broadcasting, this paper proposed
a new scheduling
h d li scheme
h called
ll d D*S/N tot serve multiple
lti l
requests with a single broadcast.
This paper also identified the effects of upload requests on
data quality, and proposed a Two-Step scheduling scheme to
pprovide a balance between servingg download and update
p
requests.