Throughput-Range Tradeoff of Wireless Mesh Backhaul

1
Throughput-Range Tradeoff of Wireless Mesh
Backhaul

• IEEE Journal on Selected Areas in Communications,
  2006

Presented by Hermes Y. H. Liu Institute of IM, NTU
2
Authors

• Harish Viswanathan
• EE Ph.D. from Cornell University
• Bell Lab. Lucent Technologies
• Sayandev Mukherjee
• EE Ph.D. from Cornell University
• Bell Lab. Lucent Technologies

3
Outlines

• Introduction
• Wireless Network Flow Problem
• Throughput Range Tradeoff in Regular Network
• Load Balancing Through Mesh Backhaul
• Summary

4
Outlines

• Introduction
• Wireless Network Flow Problem
• Throughput Range Tradeoff in Regular Network
• Load Balancing Through Mesh Backhaul
• Summary

5
Introduction

• Wide-area wireless broadband access becomes
  ubiquitous with the technologies of WiFi
  (802.11), WiMAX (802.16) and 3G cellular system
• Considering data rate and base station (BS) range
  which imposed by transmit power result in large
  number of BSs to cover a given area
• Wireless backhaul (e.g., 802.11s and 802.16d) is
  proper for the network

6
Introduction

• Mesh networking has advantages over single-hop
  networking in spatial reuse for increased
  capacity, coverage enhancement, and load
  balancing (LB) through route diversity, and
  extended to wireless ad hoc networks
• Examine wireless mesh networking from providing
  wired network connection to wireless BSs
• Routing problem how different flows are routed
  from the wired BS
• Scheduling problem duration for each
  transmission scenario (a given set of
  transmitter-receiver pairs) should be active

7
Outlines

• Introduction
• Wireless Network Flow Problem
• Throughput Range Tradeoff in Regular Network
• Load Balancing Through Mesh Backhaul
• Summary

8
Wireless Network Flow Problem

• A. System Model
• 1.Locations of Node network with multiple
  nodes (BS serving end users)
• a. Access Points (APs) have wired connection to
  the backhaul network
• b. Extension Points (BPs) extend the range of
  the wired BSs (APs)
• All APs and EPs are assumed identical
• 2.Channel Model
• a. Receg-Greenstein model (pass loss)
• b. No fast fading since all nodes are all
  stationary
• ?So the maximum rate is related to
• transmit power, distance, shadow fading,
  interference
• ?Set transmit power to be equal at all APs
  and EPs, but different between. The shadow
  fading is the customary log-normal model

9
Path Loss
10
Path Loss
11
Path Loss
12
Path Loss
13
Path Loss
14
Wireless Network Flow Problem

• 3.Traffic Model
• The aggregate traffic demand per cell which
  presents the traffic for all
  
  subscribers served by that EP is constant over
  time
• B. Statement of the Wireless Network Flow Problem
• The bits intended for the destination EP reside
  in buffers at the intermediate EPs between AP and
  destination EP, and these buffers are assumed to
  be infinite
• Flow on i the portion of the total bits travels
  through a given link i
• Commodity traffic intended for each EP which
  indexed by the label of EP
• Throughput to each destination EP total bits
  received/ transmit time allocated for these bits
  at AP and intermediate EPs

15
Wireless Network Flow Problem

• The network has n EPs, each EP d having the same
  demand f
• Link denotes a wireless link between
  node i and j
• is the portion of the total number of bits
  between AP and EP d through the edge
• Each link has a finite capacity
  represents the maximum rate (bit/s) of
  transmission on that link
• However, the capacity on each link depends upon
  the set of other simultaneously transmitting
  links due to interference
• Transmission scenario the subset of
  simultaneously active links

16
Wireless Network Flow Problem

• Transmission scenario the subset of
  simultaneously active links

17
Wireless Network Flow Problem

• N possible transmission scenarios
• represents the flow through the link
  to destination d in transmission scenario t
• Total transmission time over all N transmission
  scenarios is
• where is the transmission time allocated to
  the links on transmission scenario t

18
Wireless Network Flow Problem
(1-a)

• Objective function
• Subject to

(1-b)
and
(1-c)
Where is the link capacity
19
Wireless Network Flow Problem
Where P is the transmit power is the
distance is the shadow fading between node
i and j is the path loss exponent
represents any implementation margin relative to
the rate given by the Shannon Formula (2)
20
Wireless Network Flow Problem

• 1. The interference due to far-away transmitters
  is not exactly zero
• 2. The Shannon Formula (2) is an upper limit on
  the rate achievable for a given SINR
• So the throughput to each EP solved by
  (1-a)- (1-c)is an upper limit
• The solution to the LP problem (1-a)-(1-c) may be
  impossible over all transmission scenarios n EPs
  and 1 AP? L links and N
  transmission scenarios
• The LP problem must be solved over a reduced set
  of transmission scenarios and is only a lower
  bound on the maximum throughput attainable

21
Wireless Network Flow Problem

• C. Greedy Algorithm for Selection of Transmission
  Scenarios
• All transmissions are point-to-point and no
  transmitting node can receive simultaneously
• A good transmission scenario should consist as
  many simultaneous transmissions as possible,
  while keeping the loss in SINR due to
  interference small
• Denote
• Source and destination of link is
• the total transmission rate in the transmission
  scenario t is

22
Wireless Network Flow Problem

• C. Greedy Algorithm for Selection of Transmission
  Scenarios

23
Priority Issue
S
D2
D1
24
Wireless Network Flow Problem
25
Outlines

• Introduction
• Wireless Network Flow Problem
• Throughput Range Tradeoff in Regular Network
• Load Balancing Through Mesh Backhaul
• Summary

26
Throughput Range Tradeoff in Regular Network

• A. Network Model for Evaluation
• To save the infrastructure expenses of laying
  cable or fiber to each BS, extend the range of a
  given BA (AP) wired by using several other BSs
  (EP) unwired to the backhaul
• Multi-hop routing versus single-hop routing (no
  EP transmits to any other EP)
• In Fig. 5, the throughput per EP only
• depends on the distance from the AP

27
Throughput Range Tradeoff in Regular Network

• B. Enumerating the Scenarios for the Multi-hop
  Optimization
• Every transmitter-receiver pair must have either
• (1) the AP as the transmitter, and an EP in the
  first ring as the receiver or
• (2) the transmitter EP and receiver EP located
  in adjacent rings, receiver in the outer ring
  and adjacent to the cell with the transmitter EP
• Maximal transmission scenarios subsets of
  simultaneously active links that no new links can
  be added
• Previous greedy algorithm can enumerate the
  scenarios but did not use the spatial symmetry
  property of this network

28
Throughput Range Tradeoff in Regular Network

• Two extreme cases in antennas
• Narrow-Beam Antennas The interference is assumed
  negligible. Only one restriction
  
  The maximal transmission scenarios are
  symmetric fro the given geometry
• Omnidirectional Antennas
• No neighbor of a transmitter can be a receiver
  and no neighbor of a receiver can be a
  transmitter
• The interference due to transmitting is not felt
  more than one cell away
• Maximal transmission scenarios are symmetric for
  the given geometry
  
  
  

29
Throughput Range Tradeoff in Regular Network

• C. Results Comparing Multi-hop With Single-Hop
• Solve the problem in (1-a)-(1-c) for the network
  in Fig. 5
• The distance of the farthest EP yields the range
  of the region served by the AP with multi-hop
  network
• (1) Solution to the Single-Hop Network Flow
  Problem There is exactly one active link in each
  scenario (AP as transmitter and one EP as
  receiver). The number of transmission scenarios
  are as many as EPs (Nn)
• The time EP i to receive its demanded bits f is
  , where is the rate from AP
  s to the EP i
• The total time required for all EPs to receive
  their demand is
• The throughput is

30
Throughput Range Tradeoff in Regular Network

• (2) Description of the Numerical Parameters Used
  in the Problem
• There are two possible combinations of antenna
  heights and transmit powers for AP and EPs

AP height AP power EP height EP power
Situation 1 Single-hop 20m 43dBm 10m 30dBm
Situation 1 Multi-hop 10m 30dBm 10m 30dBm
Situation 2 Single-hop and Multi-hop 30dBm 10m 30dBm
31
Throughput Range Tradeoff in Regular Network

• (3) Multi-hop With Omnidirectional Antennas
  Versus Single-Hop

Multi gt Single With interference
Single gt Multi Increase with the cell size
Situation 1
Situation 2
32
Throughput Range Tradeoff in Regular Network

• (3) Multi-hop With Omnidirectional Antennas
  Versus Single-Hop
• Situation 1, If the power available to a wired BS
  is large enough to cover a given area, there is
  no advantage to introduce EPs for more throughput
• Situation 2, multi-hop routing is superior to
  single-hop routing, though the advantage is not
  overwhelming due to the interference

33
Throughput Range Tradeoff in Regular Network

• (4) Multi-hop With Narrow-Beam antennas Versus
  Single-Hop

Single gt Multi Increase with the cell size
Multi gt Single Without interference
Situation 1
Situation 2
34
Outlines

• Introduction
• Wireless Network Flow Problem
• Throughput Range Tradeoff in Regular Network
• Load Balancing Through Mesh Backhaul
• Summary

35
Load Balancing Through Mesh Backhaul

• A. Introduction
• The difference traffic generated in different BSs
  provides an opportunity to route traffic from
  heavily loaded BSs to the wired network through
  lightly loaded BSs
• The wireless mesh backhaul allows reconfiguration
  of routes and traffic flow to maximize the
  traffic carried into the wireless network
• Objective determining optimum routing and
  scheduling of flows also taking into account the
  traffic generation rate and the maximum wired
  backhaul transmission rate

36
Load Balancing Through Mesh Backhaul

• B. Linear Programming Formulation
• under only one transmission scenario

(4)
Where is the link rate on the wireless
links (i,j) is the flow on the
wired link between node i and the wired network
is the common transmission rate
of the wired links (i, w) T is
some fixed duration of time such as the slot time

(5) analogous to (1-b)
is the access
traffic generation rate at node j This is not an
equality constraint because in general the access
traffic generation rate can be larger than the
rate traffic can be transmitted over wireless or
wired links Objective
(LP problem)
(6)
37
Load Balancing Through Mesh Backhaul

• Extend the formula to all transmission scenarios
  with respect to interference and variable-rate
  transmissions

(7-a)
Subject to
(7-b)
(7-c)
(7-d)
Where is the time portion of transmission
scenario i The total throughput is given by
, without loss of
generality, T can be Set to 1
38
Each of the BS-EPs is also connected to The
wired network through a wired backhaul
Connection with maximum transmission rate Our
goal is to determine the MAX access traffic that
can be carried into the wired network using the
available backhaul links through the best
utilization of the wireless mesh
backhaul capabilities
39
Load Balancing Through Mesh Backhaul

• C. Evaluation of the Benefit of LB
• Access traffic generation rate decrease linearly
  from the center outward

40
Load Balancing Through Mesh Backhaul

• The total access traffic generated from all BSs
  is set equal to the total backhaul bandwidth rate
  to the wired network to evaluate the benefit of LB

LB shows a significant improvement and the Larger
the benefit the more unbalanced the Traffic
pattern
41
Load Balancing Through Mesh Backhaul
42
Load Balancing Through Mesh Backhaul
43
Outlines

• Introduction
• Wireless Network Flow Problem
• Throughput Range Tradeoff in Regular Network
• Load Balancing Through Mesh Backhaul
• Summary

44
Summary

• Proposed a LP model for optimum routing and
  scheduling of flow in a wireless mesh network
  include the effect of interference and
  variable-rate transmissions
• Required the enumeration of reduced transmission
  scenarios and associated transmission rates due
  to the complexity (large number of N)
• In given hexagonal network, the throughput is
  determined by the transmission power, antenna
  height, and cell radius
• The multiple-hop outperforms the single-hop when
  the antenna height and transmit power are the
  same
• The total traffic to the wired network is
  maximized by routing traffic to underutilized
  backhaul links (Load Balance)

45
Future Work

• Sequence of the transmission scenarios
  (scheduling at source node) while transmitting
  data in a wireless mesh backhaul network
• Finite buffer limitation
• Load balancing in mesh backhaul network where
  only the source node is connected to the wired
  network through a wired backhaul link

46

• The End
• Thanks for Your Listening!!