jianliang xu, dik l. lee, and bo li dept. of computer science hong kong univ. of science &...
Post on 22-Dec-2015
218 views
TRANSCRIPT
Jianliang XU, Dik L. Lee, and Bo Li
Dept. of Computer ScienceHong Kong Univ. of Science & Technology
April 2002
On Bandwidth Allocation for Data Dissemination in Mobile Cellular Networks
04/19/23 2
Outline
• Introduction
• Problem formulation
• Bandwidth allocation techniques
• Performance evaluation
• Wrap up
04/19/23 3
• On-Demand Access– A client sends data requests uplink to
the server, and the server returns the
results to the client individually
– Fast response for a light-load system
Wireless Data Dissemination
• Data Broadcast– The server periodically broadcasts
info to the entire client population, and the
clients monitor the broadcast channel
to retrieve the data of their interest
– Scale to an unlimited client population
04/19/23 4
Bandwidth Allocation – Previous Work
• Data dissemination– Band allocation among data items [Acharya et al., SIGMOD’95;
Hameed and Vaidya, ACM WINET]
– Band allocation between on-demand access and data broadcast [Acharya et al., SIGMOD’97; Lee et al., ACM MONET]
– Confined to a single-cell environment
• Voice communications– Pretty many studies [Oh & Tcha, IEEE ToC; Zhang & Yum, IEEE
ToVT; Li et al., ACM WINET; etc.]
– Minimize call blocking/dropping prob. or improve carried traffic while ensuring QoS
– Different objective with data dissemination (access delay)
• Our study: data dissemination & multi-cell
04/19/23 5
A Motivating Example
• Cellular: 2 cells; 6 Kbps shared bandwidth• Database: 4 items, each of 1K bits• Data access rate: Cell A: 1 Cell B: 4• Flat broadcast
Band Alloc(Kbps) Exp. Latency(s) Overall Exp.
Latency(s)Cell A Cell B Cell A Cell B
Uniform 3.0 3.0 1.0 1.0 1.0
Proportional 1.2 4.8 2.5 0.625 1.0
Best 2.0 4.0 1.5 0.75 0.9
1 2 3 4
04/19/23 6
Bandwidth Allocation Problem
Input parameters
Traffic pattern for each cell (long-term steady state) Total amount of bandwidth Frequency reuse pattern
Problem: how much bandwidth is assigned to each cell Objective: to minimize the overall access latency for a
multi-cell wireless data dissemination system
04/19/23 7
Definitions
Min reuse distance: the min. distance at which frequencies can be reused at acceptable interference
Interference cluster: a maximal subset of cells which are within the distance of mutual interference
1
32
3R
04/19/23 8
Cost Model• Latency for on-demand access
• Latency for data broadcast
iii
i
ii
ii
o
lb
l
lb
bit
1
),(
)2
(),( ,,
1,
i
ji
i
jiM
jjii
b
b
l
b
spbit
i
• Overall expected latency
),(),(1
11
1
io
N
NiN
j j
ii
bN
iN
j j
i bitbitb
b
Notationsbi amount of band to cell i data access rate for cell ili average item size for cell i
li,j size of item j in cell isi, j space distance between instances of item j in cell ili, j size of item j in cell i
pi, j access prob for item j in cell i
Mi num of items in cell i’s db
N num of cells in the systemNb num of cells using broadcast
i
04/19/23 9
Problem Formulation
• Optimal bandwidth allocation problem
Find out (b1, b2, …bN)
Minimize
Subject to for any interference cluster Q
BbQb
k
k
),(),(1
11
1
io
N
NiN
j j
ii
bN
iN
j j
i bitbitb
b
04/19/23 10
• An interference cluster: Nc cells
• Problem formulation
Find out (b1, b2, …bc)
Minimize
Subject to BbcN
ii
1
• A constrained-minimum problem
Optimal Allocation for Interference Cluster
),(),(1
11
io
N
NiN
j j
ii
bN
iN
j j
i bitbitc
b
c
b
c
04/19/23 11
Optimal Allocation for Interference Cluster (cont’d)
• Solved the optimization problem using the Lagrange
multiplier theorem
Theorem 1 The min overall expected latency is achieved when the bandwidth allocated to cell i, is given by
cbiiN
i
N
Ni iib
i
N
Ni iib
i
bN
i
N
Ni iib
i
N
Ni iib
i
i
NiNllit
lBit
Nii lit
lBit
b
b c
b
c
b
b c
b
c
b
1 if )1,(
)()1,(
if )1,(
)()1,(
1 1
1
1 1
1
04/19/23 14
Allocation Techniques for Cellular Networks
• Challenges– Frequency reuse – non-uniform traffic loads
• Heuristic allocation techniques– Compact allocation
– Cluster-step allocation
04/19/23 15
Compact Allocation
• Equivalence class: the cells at min reuse distance• Assign the same bandwidth to all the cells in an
equivalence class• Reduced problem: bandwidth allocation for a cluster
consisting of the cells with one from each equivalence class– Take the average parameter values
– Apply the optimal allocation technique
– Favor homogeneous loads for the cells in an equivalence class
04/19/23 16
Cluster-Step Allocation
• For all possible interference clusters, assign bandwidth one cluster after one depending on their importance– Importance is determined by the aggregate deserved
allocation factor ( for broadcast and for on-demand access)
– For each cluster only unassigned cells are considered: the total bandwidth is a min of B1 and B2
• B1: remaining available bandwidth for the cluster
• B2 : deserved allocated bandwidth for the unassigned
cells
– Apply the optimal allocation technique for each cluster
)1,(itb
i iil
04/19/23 17
Simulation Setup
• Database size: 1,000 data items• Total available bandwidth: 672 Kbps • Mobile cellular network: 7 × 7 cells • Min reuse distance: • Access rates
• RAND: uniformly distributed between 1 and 50
• HOMO: for the cells in equivalence class i, uniformly distributed between 1 and 2i+1
• HETERO: uniformly distributed between 1 and 10 except some have 100
R21
04/19/23 18
0
5
10
15
20
25
30
35
40
Exp
ecte
d A
cces
s La
tenc
y (s
)
RAND HOMO HETERO
Uniform Alloc
Compact Alloc
Cluster-step Alloc
Numerical Result – Data Broadcast
04/19/23 19
1.E-01
1.E+00
1.E+01
1.E+02
1.E+03
1.E+04
Exp
ecte
d A
cces
s La
tenc
y (s
)
RAND HOMO HETERO
Uniform Alloc
Compact Alloc
Cluster-step Alloc
Numerical Result – On-Demand Access
04/19/23 20
1.E-01
1.E+00
1.E+01
1.E+02
1.E+03
1.E+04
Exp
ecte
d A
cces
s La
tenc
y (s
)
RAND HOMO HETERO
Uniform Alloc
Compact Alloc
Cluster-step Alloc
Numerical Result – Hybrid Access
04/19/23 21
• Limitations
- On-demand access with an M/M/1 model
- Bandwidth allocation granularity
• Formulated the band allocation problem for a wireless data dissemination system
• Analyzed the optimal bandwidth allocation technique for an interference cluster
• Proposed two heuristics for bandwidth allocation in a mobile cellular network
• Numerical results showed the superiority of the proposed solutions
Wrap Up …