MOBILE NETWORK THROUGHPUTCALCULATION: A REVIEW OF
THROUGHPUT CALCULATION
VM. NOUFIDALI, M.Tech
Department of Computer and
Communication Engineering
Karunya University, Coimbatore-641114
vmnoufid@karunya.edu.in
alimon.vm@gmail.com
Phone: 9486383914.
Abstract
Network throughput and packet delay are
the two most important parameters to
evaluate the performance of wireless ad hoc
networks. Generally it is difficult to achieve
both high throughput and low packet
delay.The era of satellite-based mobile
communications systems started with the
first MARISAT satellite which was
launched into a geostationary orbit over the
Pacific Ocean in 1976 to provide
communications between ships and shore
stations. The combination of high cost and
unacceptably large equipment has kept
mobile satellite communications (MSC)
systems from appealing to the wider market
of personal mobile communications.
However, the progress made over the last ten
years in digital voice processing, satellite
technology, and component miniaturization
has resulted in the viability of MSC systems
in responding to the growing market in
personal mobile communications. The main
focus of this paper is describing the
calculation of throughput in mobile
networks. Throughput capacity in mobile ad
hoc networks has been studied extensively
under many different mobility models.
However, most previous research assumes
global mobility, and the results show that a
constant per-node throughput can be
achieved at the cost of very high delay.
Multiantenna or MIMO systems offer great
potential for increasing the throughput of
multihop wireless networks via spatial reuse
and/or spatial multiplexing. In multirate
wireless networks, interference between
higher and lower data rate links is a
notorious factor against performance
improvement. This paper describes new
algorithms for throughput optimization in a
mobile backbone network. This hierarchical
communication framework combines mobile
backbone nodes, which have superior
mobility and communication capability, with
regular nodes, which are constrained in
mobility
and
communication
capability.Finally, i also evaluate the effects
of several network parameters on this
achievable throughput, and show how
throughput
behaves
under
these
effects.[01][02][04][05][08][16]
I. Introduction
Wireless network is becoming more and
more popular in nowadays. Comparing to
the traditional wired network, wireless
network set up the connections through
wireless channel. Generally there are two
kinds of wireless networks. One has a wired
backbone network in which the base stations
are the boundary nodes, and the extended
connections between mobile users and the
base station are wireless channels. This onehop wireless network is very popular
currently, i.e., the cellular networks and
WLANs. The other is wireless ad hoc
network, which has more than one hop
wireless channels in the connection. This
kind of topology is not widely implemented
yet, but it is useful sometimes, especially in
military applications and sensor networks. In
this project, I will focus on the latter
topology, the wireless ad hoc network,
without considering any wired links. As an
extension to the backbone network, wireless
ad hoc network consists of nodes that
communicate with each other through
wireless channels only. I can describe the
system as follows. In these system consists
of only wireless nodes, in which all nodes
can communicate with other nodes in the
range of radio transmission through wireless
channel. Each wireless node can act as a
sender, a receiver or a router. As a sender,
the node can send message to the specified
destination node through some route. As a
receiver, it can receive the message from
other nodes. As a router, it can relay the
packet to the destination or next router in the
route if necessary. Each node can buffer
packets when the packets need to wait for
transmission. We are interested in the
capacity and delay of such kind of network.
In general these two parameters are the most
important performance measurement for any
wireless network systems. The capacity
represents the throughput (bits per second)
of the whole system including all nodes, and
the delay represents the average time
duration of a packet transmitting in the
network from a source to the destination. As
in any other queuing system, there are
tradeoffs between the capacity and the delay.
Intuitively in order to increase the capacity,
we need to keep all nodes busy with
transmitting or receiving packets during all
the time, which means the queue of each
node is always nonempty; obviously this
will lead to a longer delay. On the other
hand, in order to reduce the delay, the
optimal situation is, all nodes along the route
can transmit the packet immediately to the
next node until it reaches the destination,
which means there is no packet competing
for transmissions in the queues, surely this
causes very low throughput. We will see this
tradeoffs in wireless ad hoc networks in the
report. Furthermore, this paper objective is
to find a way that the network can achieve a
high throughput while keeping the delay
under certain threshold. This report will
address the problem. In the following,
section will describe the methodology to
model the problem step by step. Finally this
paper concludes the throughput and delay in
wireless ad hoc networks.
II. Discussion
General:
1. Throughput
In communication network, throughput or
network throughput is the average rate of
successful message delivery over a
communication channel. This data may be
delivered over a physical or logical link, or
pass through a certain Network node. The
throughput is usually measured in bit per
second (bit/s or bps), and sometimes in data
2
packets per second or data packets per time
slot.The system throughput or aggregate
throughput is the sum of the data rates that
are delivered to all terminals in a
network.[32]
system. From a user perspective, this is often
phrased as either "which device will get my
data there most effectively for my needs?",
or "which device will deliver the most data
per unit cost?". Systems designers are often
interested in selecting the most effective
architecture or design constraints for a
system, which drive its final performance.
Four different values have meaning in the
context of "maximum throughput", used in
comparing the 'upper limit' conceptual
performance of multiple systems. They are
'maximum
theoretical
throughput',
'maximum achievable throughput', and 'peak
measured throughput' and 'maximum
sustained throughput'. These represent
different quantities and care must be taken
that the same definitions are used when
comparing different 'maximum throughput'
values. [32]
2. How to Calculate Throughput:
Throughput is used to determine the
maximum rate at which a computer user can
expect data to transfer. Those using fast
Internet connections might see their transfer
levels as less than they expected due to a
low throughput. This document will outline
how to determine the maximum throughput
available to computer users on their given
network. [32]
3.
How
to
Throughput:
Calculate
Network
Network throughput refers to the average
data rate of successful data or message
delivery over a specific communications
link. Network throughput is measured in bits
per second (bps). A common misconception
on measuring network throughput is that
measuring the time it takes to upload or
download a large file is the maximum
throughput of a network. This method does
not take into account communications
overhead such as Network receiver window
size, machine limitations or network latency.
Maximum network throughput equals the
TCP window size divided by the round-trip
time of communications data packets. [32]
5. Maximum theoretical throughput:
This number is closely related to the channel
capacity of the system, and is the maximum
possible quantity of data that can be
transmitted under ideal circumstances. In
some cases this number is reported as equal
to the channel capacity, though this can be
deceptive, as only non-packetized systems
(asynchronous) technologies can achieve
this without data compression. Maximum
theoretical throughput is more accurately
reported to take into account format and
specification protocol over head with best
case assumptions. This number, like the
closely related term 'maximum achievable
throughput' below, is primarily used as a
rough calculated value, such as for
determining
bounds
on
possible
performance early in a system design phase.
[32]
4. Maximum throughput
Users of telecommunications devices,
systems designers, and researchers into
communication theory are often interested in
knowing the expected performance of a
3
Throughput is sometimes normalized
and
measured
in
percentage,
but
normalization
may cause
confusion
regarding what the percentage is related to.
Channel utilization, Channel efficiency and
packet drop rate in percentage are less
ambiguous terms. The channel efficiency,
also known as bandwidth utilization
efficiency, in percentage is the achieved
throughput related to the Net bitrates in bit/s
of a digital communication channel For
example, if the throughput is 70 Mbit/s in a
100 Mbit/s Ethernet connection, the channel
efficiency is 70%. In this example, effective
70Mbits of data are transmitted every
second. Channel utilization is instead a term
related to the use of the channel disregarding
the throughput. It counts not only with the
data bits but also with the overhead that
makes use of the channel. The transmission
overhead consists of preamble sequences,
frame headers and acknowledge packets.
The definitions assume a noiseless channel.
Otherwise, the throughput would not be only
associated to the nature (efficiency) of the
protocol but also to retransmissions resultant
from quality of the channel. In a simplistic
approach, channel efficiency can be equal to
channel
utilization
assuming
that
acknowledge packets are zero-length and
that the communications provider will not
see
any
bandwidth
relative
to
retransmissions or headers. Therefore,
certain texts mark a difference between
channel utilization and protocol efficiency.
In a point-to-point or poin to multi point
communication link, where only one
terminal is transmitting, the maximum
throughput is often equivalent to or very
near the physical data rate (the Channel
capacity) since the channel utilization can be
almost 100% in such a network, except for a
6. Peak measured throughput
The above values are theoretical or
calculated values. Peak measured throughput
is throughput measured by a real,
implemented system, or a simulated system.
The value is the throughput measured over a
short period of time; mathematically, this is
the limit taken with respect to throughput as
time approaches zero. This term is
synonymous
with
"instantaneous
throughput". This number is useful for
systems that rely on burst data transmission;
however, for systems with a high Duty cycle
this is less likely to be a useful measure of
system performance.
7. Maximum sustained throughput
This value is the throughput averaged or
integrated over a long time (sometimes
considered infinity). For high duty cycle
networks this is likely to be the most
accurate indicator of system performance.
The maximum throughput is defined as the
asymptotic throughput when the load (the
amount of incoming data) is very large. In
systems where the load and the throughput
always are equal packet loss (where does not
occur), the maximum throughput may be
defined as the minimum load in bit/s that
causes the delivery time (the latency) to
become unstable and increase towards
infinity. This value can also be used
deceptively in relation to peak measured
throughput to conceal packet shaping. [32]
8. Channel utilization - Channel efficiency
- Normalized throughput
4
small inter-frame gap. [32]
the objective through exploiting the patterns
in the mobility of nodes. The throughput
achieved by this algorithm is only a polylogarithmic factor off from the optimal.
First of all, we develop the models for
wireless ad hoc networks with static nodes.
Gupta and Kumar [29] propose two models
for such kind of network. For simplicity, the
models scale the space so that n nodes are
located in a region of area 1 m^2. Each node
can transmit at W bits per second over a
common wireless channel. The channel is
divided to several sub-channels, each with
capacity W1 ,W2 ,… ,W m bits per second,
where Σ Wm =W(limit m=1 to M).the
different models are given. [16]
1. Model of arbitrary networks [16]
2. Model of random networks [16]
3. Model with mobile nodes [16]
4. Delay model [16]
III .Methodology
In this part, we will show the methodology
to solve the problem step by step. Recall that
our objective is to achieve high capacity in
wireless ad hoc networks with keeping the
packet delay under a small threshold. We
will model the networks from simple to
complex, from general to specific step by
step. In each model, we will describe the
scenario, the transmission model and the
measurement metrics in details.
1. THROUGHPUT AND DELAY IN
WIRELESS AD HOC NETWORKS
This paper first look at the throughput
capacity, theoretically in mobile ad hoc
networks.Gupta and Kumar [29] show the
average available throughput per node
decreases as 1/sqrt (n) or 1/sqrt (n log n) in a
static ad hoc network, where n is the number
of nodes. That means, the total network
capacity increases as at most sqrt
(n).Furthermore, Grossglauser and Tse [30]
show mobility can improve the capacity.
However, delay is not guaranteed in their
schemes. Actually delay will increase due to
possibly more hops or queuing in order to
increase the throughput.Bansal and Liu [31]
show it is possible to achieve close-to
optimal capacity while keeping the delay
small. In their model, each sender can
achieve an average throughput of (c (W min
(m, n)/n log^3 n)), where W is the maximum
available bandwidth, with the packet delay
at most 2d/v , where d is the diameter of the
network and v is the velocity of the mobile
nodes. Based on this fact, the authors
propose a routing algorithm that achieves
ROUTING ALGORITHM
Step1. Local leader election
A local lead is elected among the static
nodes within each region of size 1/ m ×1/ m
.This leader will be responsible for
communicating all the messages of the static
nodes in its region with the mobile nodes.
Step2. Static to mobile phase
A static node S1 wanting to send
messages to destination R first transfers its
message to its local leader S. S stores the
message and waits for a mobile node M1
such that M1 is close enough to S and
moving approximately along the direction of
R. when such a node is available S hands
over the data from S1 to M1.
5
Step3. Mobile to mobile phase
hierarchical network architecture called
a mobile backbone network, in which
mobile agents are deployed to provide longterm communication support for other
agents in the form of a fixed backbone over
which end-to-end communication can take
place. Mobile backbone networks can be
used to model a variety of multivalent
systems. For example, a heterogeneous
system composed of air and ground vehicles
conducting ground measurements in a
cluttered environment can be appropriately
modelled as a mobile backbone network, as
can a team of mobile robotic agents
deployed to collect streams of data from a
network of stationary sensor nodes.
The mobile nodes relay the packets
towards R amongst all possible mobile
nodes such that the packet moves closer and
closer to the destination.
Step4. Static to static phase
When the mobile node carrying the
packet is close enough to the destination, it
hands off the packet to some leader node.
This packet is then relayed among the static
leader nodes towards the correct leader
node, which can transmit the packet to the
destination node directly. With this routing
algorithm, the wireless ad hoc network can
achieve close-to optimal capacity while
keeping the packet delay small. This
algorithm exploits the mobility patterns of
the nodes to provide guarantees on the
packet delay. The readers can refer to [3] if
interested in the detailed operations and
arguments of the algorithm.
Fig.1. A typical example of an optimal
mobile backbone network. Mobile backbone
nodes, indicated by*, are placed such that
they provide communication support for
regular nodes, shown as 0. Each regular
node is assigned to one mobile backbone
node. Dashed lines indicate the radius of
each cluster of nodes.
2. THROUGHPUT OPTIMIZATION IN
MOBILE BACKBONE NETWORKS
Detection and monitoring of spatially
distributed Phenomena often necessitate the
distribution of sensing platforms. For
example, multiple mobile robots can be used
to explore an area of interest more rapidly
than a single mobile robot, and multiple
sensors can provide simultaneous coverage
of a relatively large area for an extended
period of time . However, in many
applications, the data collected by these
distributed platforms is best utilized after it
has been aggregated, which requires
communication among the robotic or
sensing agents. This paper focuses on a
3. AN ANALYSIS OF TRAFFIC AND
THROUGHPUT FOR UMTS PACKET
CORE NETWORKS
6
can generate enough traffic to congest
UMTS PS networks and impact the majority
of subscribers using interactive Web
browsing and E-mail applications. As a
result, mobile operators must find
algorithms and rules that will dimension
their emerging 3G PS networks, while
addressing their potential 4G deployment
requirements and that will not require a
“forklift” upgrade. In order to accurately
plan, design, and dimension the UMTS PS
network, this paper will develop the
algorithms of traffic and throughput for the
UMTS PS network entities (NEs) described
in Section 3. The analysis will be based on
the live traffic and throughput generated or
absorbed in the interfaces of PS NEs. A case
study is provided to verify the algorithms
created for UMTS PS domain. This paper is
aimed at helping UMTS PS network
operators dimension an optimum network
size and build an optimum network structure
to deliver an optimum quality of service for
users. In addition, the network optimization
and expansion is the further effort for the
mobile operator after the rolling out of
mobile networks. To minimize the
CAPEX/OPEX and maintain the QoS of
mobile core networks, we propose that the
impact of cell cite re-homing on the mobile
core should be studied. It is believed that the
appropriate cell site re-homing in radio
domain, via correct algorithms applied, not
only optimizes the radio network but also
helps improve the QoS of the core network
and minimize the mobile operator’s
CAPEX/OPEX investment in their core
networks.
Packet switched domain of third generation
(3G) UMTS network serves all data related
services for the mobile subscribers.
Nowadays people have a certain expectation
for their experience of mobile data services
that the mobile wireless environment has not
fully met, since the speed at which they can
access their packet switched (PS) services
has been limited. Mobile operators realize
that if they are to succeed in today’s
wireless2 communications landscape, they
must address the quality of service for their
packet service users. Simply adding more
bandwidth to accommodate increased packet
switched traffic is an expensive alternative.
Hence, the mobile operators are faced with
the issue of how to do more with less? The
answer is to ensure a reasonable
dimensioning for UMTS packets switched
(PS) network while maintain the network
quality of service. Radio access solutions are
a primary concern of the UMTS deployment
strategy, as it impacts the mobile operators’
most valuable asset: spectrum. As an equally
important part of this formula, the core
network will play an essential role in
enhancing mobility, service control, efficient
use of network resources and a seamless
migration from 2G/3G to 4G. Hence the
network evolution calls for a transition to a
“flat,” all-IP core network with a simplified
architecture and open interfaces.
UMTS Packet Switched (PS) network is a
typical data network in which data traffic,
particularly with streaming media services,
is live, extremely time sensitive to delay,
latency, jitter, and non-tolerant of
congestion. For example, a small minority of
packet service subscribers running File
Transfer Protocol (FTP), streaming video or
peer-to-peer (P2P) file sharing applications
ARCHITECTURE OF UMTS CORE
NETWORKS
7
Packet Switched (PS) domain and
Circuit Switched Domain comprise the Core
Network (CN) of a 2G Global Systems for
Mobile Communications (GSM) or a 3G
UMTS network. Whether in 2G or 3G
phase, the CN plays an essential role in the
mobile network system to provide such
important
capabilities
as
mobility
management, call and session control,
switching and routing, charging and billing,
and security protection.
In R99 version, the first version of 3G
UMTS network, the CN domain still
consists of the same network entities (NE)
and the same network architecture as that in
GSM phase. However, there is a change in
the circuit switched domain of R4, the
second version of UMTS, which supports a
networking mode where bearer is separated
from control. Meanwhile multiple bearer
modes such as ATM/IP/TDM are supported
by CN. Consequently the Mobile Switching
Center (MSC) in GSM/UMTS R99 is split
into two NEs: MSC Server (MSS) and
Media Gateway (MGW). We should note
that no changes happen in packet switched
domain from R99 to R4 except for a new IuPS interface which is used to connect PS
domain with 3G radio access network
(RAN).
The CN in UMTS is logically classified
into the circuit switched domain (CS) and
packet switched domain (PS). The CS
domain includes such logical NEs as MSC
Server, MGW, Visitor Location Register
(VLR) integrated in MSC Server physically,
Home
Location
Register
(HLR),
Authentication
Center
(AUC),
and
Equipment Identity Register (EIR). The
packet switched domain (PS) includes
Serving GPRS Support Node (SGSN) and
Gateway GPRS Support Node (GGSN).
More specifically, PS domain consists of
data service NEs: SGSN and GGSN as well
as auxiliary NEs like Charging Gateway
(CG), Border Gateway (BG) and Domain
Name System Server (DNS), and different
service platforms attached to PS domain.
Figure 1 displays the topology of UMTS CN
with the logical NEs mentioned above.
From 3GPP TS23.060, 3GPP TS24.008,
3GPP TS23.002, Packet Switched domain
physically consists of SGSN, GGSN, and
Charging Gateway. Below is a short
description of these NEs. On the other hand,
the other NEs in CS domain such as HLR,
MGW and MSS coordinate with SGSN or
GGSN to implement some PS related
functions. 5 From 3GPP TS29.060 and
3GPP TS29.061, SGSN is responsible for
the delivery of data packets from and to MSs
8
within its serving area. Its tasks include
packet routing and transfer, mobility
management (attach/detach and location
management), logical link management, and
authentication and charging functions. Its
interfaces include Iu-Ps interface connecting
to RNC, Gn/Gp interface to GGSN, Gr
interface to HLR, Gs interface to MSC
Server or MSC, Gd interface to Short
Message Center (SMC), and Ga interface to
Charging Gateway. GGSN is a gateway
between UMTS PS/GPRS network and
external data networks (e.g. Internet). It
performs such functions as routing and data
encapsulation between a MS and external
data network, security control, network
access control and network management.
From UMTS PS/GPRS aspect, a MS selects
a GGSN as its routing device between itself
and external network in the activation
process of PDP context in which Access
Point Name (APN) defines the access point
to destination data network. From external
data network aspect, GGSN is a router that
can address all MS IPs in UMTS PS/GPRS
network. GGSN provides Gc interface to
connect with HLR, Gn/Gp interface with
SGSN, Gi interface with external data
networks, and Ga interface with CG.
Charging Gateway is the billing unit for PS
domain. Sometimes coupled together with
SGSN, it collects, merges, filters and stores
the original Call Detail Record (CDR) from
SGSN and communicates with billing
centre, and then transfers sorted CDR to
billing centre. HLR is responsible for
storing, updating, revising or deleting
subscriber related information, covering the
basic service subscription information,
supplementary
service
subscription
information and location information of
subscribers. In addition, it also implements
the function of subscriber security
management. From physical connection
aspect, HLR provides D interface to connect
with VLR in MSC Server, C interface to
connect with MSC Server or MSC in GSM
CN, Gr interface with SGSN, and Gc
interface with GGSN. The type of signaling
message delivered from and to HLR is
Mobile Application Part (MAP). In UMTS
circuit switched domain, MSC Server is a
functional entity that implements mobile call
service, mobility management, handover,
and other supplementary services. Due to the
philosophy of separation of control function
from bearer function in UMTS CN, it is
actually a controller of MGW to establish
call routes between Mobile Stations (MS)
via Mc interface. MSC Server also
physically integrates with a VLR to hold
subscriber’s data. MSC Server provides the
optional Gs interface with SGSN. In
addition, a MGW in a UMTS implements
bearer processing functions between
different networks. It implements UMTS
voice communication, multimedia service,
CS domain data service, and interworking
between PSTN and UMTS CN and between
GSM CN and UMTS CN. MGW provides
Iu-CS interface to connect with the Radio
Network Controller (RNC) in the Radio
Access Network (RAN), Nb interfaces with
its peer MGW, E interfaces with 2G MSC,
Mc interfaces with MSC Server, A interface
with BSC, and Ai interface with Public
Switched Telephone Network (PSTN).
ALGORITHMS FOR THROUGHPUT
IN INTERFACES OF UMTS PACKET
CORE NETWORKS
9
transport the data in both control and user
plane via IP over ATM. The total throughput
in Iu-PS interface is the sum of the
throughput of user plane and control plane in
Iu-PS interface. The following paragraphs
will respectively introduce the algorithms of
user plane and control plane of Iu-PS
interface. [23]
Since Iu-PS interface is newly defined in
UMTS CN, this section will first introduce
the algorithms for Iu-PS interface. The
throughput algorithms for the other
interfaces such as Gn, Gi, Gs and Gr
interface, since they have been existing in
GPRS network, will also be introduced
based on a general rule: total traffic (Erlang
or message size) times traffic proportion to
obtain the traffic distribution for each NE
and each link.
4. Modeling Throughput Gain of Network
Coding in Multi-Channel Multi-Radio
Wireless Ad Hoc Networks
Network coding (NC) is receiving more
and more research attention since it is a
promising technique to increase the network
throughput for both wired and wireless
networks. By exploiting the broadcast nature
of the wireless channel, the conventional
wireless NC proposed in can significantly
increase the network throughput as
compared with the traditional non-NC
transmission scheduling based scheme in
multi-hop wireless ad hoc networks. The
essential idea of conventional wireless NC
can be explained as follows using a simple
example. As shown in Fig. 1(a), node A
wants to send a single packet to node
C,while node C wants to send a single
packet to node A. Due to transmission range
limitations, both of these two paths go
through the relay node B. This is the
simplest two-way relay topology. Suppose
that the time axis is divided into time slots
and the transmission of each packet spends
one time slot .Then, if we adopt the non-NC
scheduling based scheme, four Manuscript
received 1 August 2008; revised 20 February
2009. The research reported in this paper
was supported in part by the National
Science oundation CAREER Award under
Grant ECS-0348694.The authors are with
Iu-PS Interface
Iu-PS interface, situated between Radio
Network Controller (RNC) and Serving
GPRS support Node (SGSN) and Iu-CS
interface between RNC and Media Gateway
(MGW) composes the Iu interface. Iu-PS
and Iu-CS interface define the same protocol
stacks of transport network user plane and
control plane, whereas they have the
different transport network user plane.
Ouyang. Y. and Fallah M.H. (2009)
illustrate the throughput algorithm for Iu-CS
interface. Table 1 displays the protocol
stacks of Iu-CS interface. Defined by 3GPP
TS 25.401, ITU-T I.363.2, 3GPP TS 25.415,
and 3GPP TS 25.413, the data of user plane
in Iu-CS interface is transparently
transported and carried by ATM Adaption
Layer 2 (AAL2) while the voice data such as
Adaptive Multi Rate (AMR) frame is
supported by User Plane Protocol (Iu-UP)
stands on the top layer and follows by AAL2
and ATM. According to 3GPP TS23.060,
3GPP TS 32.015, and 3GPP TS 25.413, the
protocol stacks of Iu-PS interface are shown
in Table 2, in which a significant difference
is AAL5 rather than AAL2 in Iu-CS
interface is adopted in layer 2 of Iu-PS to
10
the Networking and Information Systems
Laboratory.The following is a possible
sequence of these transmissions: 1) node A
sends a packet to node B while node C
remains silent;2) node B relays A’s packet to
node C;3) node C sends its packet to node B
while A remains silent; 4) node B relays C’s
packet to node A. In contrast to the non- NC
scheme, the conventional wireless NC
scheme only needs three time slots to
complete the two-way relay transmissions .
The following is a possible sequence of
these transmissions: 1) node A sends a
packet to node B; 2) node C sends a packet
to node B; 3) node B transmits a new packet
obtained by performing an XOR of A’s and
C’s packets. Node A can XOR the received
new packet from node B with its own packet
to obtain C’s packet. In the same way, node
C can get A’s packet. Besides the broadcast
nature of the wireless channel, is there
anything else that can be exploited by the
NC to further increase the network
throughput? The answer is positive. thee
authors of [3], [4] developed the analog NC,
which can even take advantage of the native
physical-layer coding ability by analogously
mixing simultaneously arrived radio waves
at the relay nodes, to further increase the
network throughput. Specifically, under the
analog NC, the two-way relay transmissions
can be completed in just two time slots. In
the first time slot, A and C transmit their
packets to node B simultaneously, resulting
in interfere of their transmissions at the relay
node B. Due to interference, the relay node
receives the sum of A’s and C’s analog
signals. This is a collision and the relay node
B cannot decode the bits. In the second time
slot, the relay node B simply amplifies and
forwards the received interfered signal at the
physical layer without decoding it. Since
node A knows the packet it sent, it also
knows the packet’s corresponding analog
signal. It can thus subtract its original signal
from the received interfered signal to get the
signal of C’s packet, from which it can
decode C’s packet. Likewise, C can decode
A’s packet. The promising analog NC
technique motivates us to investigate how to
practically apply the analog NC and how
well the analog NC can perform in the
multi-hop, multi-channel and multi-radio
wireless ad hoc networks with multiple
unicast sessions. The main goal of this paper
is to analytically model the network
throughput improvements of the above
mentioned two types of wireless NC over
wireless ad hoc networks. To our best
knowledge, there was no existing work
reported yet in the literature, which
compares the network throughputs achieved
by the non-NC scheme, conventional NC
scheme, 0733-8716/09/$25.00 _c 2009 IEEE
594 IEEE JOURNAL ON SELECTED
AREAS IN COMMUNICATIONS, VOL.
27, NO. 5, JUNE 2009
Fig. 1. Diagrams of the relay network
topologies. (a) An example of the 2-way
relay network. (b) An example of the n-way
relay network. and analog NC scheme,
11
respectively,
from
a
theoretical
perspective.Our contributions can be
summarized as follows.We show that the
network
throughput
gains
of
the
conventional NC and the analog NC are
(2n)/(2n− 1) and n/(n − 1), respectively, for
the n-way relay networks where n ≥ 2.
Applying
the
linear
programming/optimization technique, we
formulate a general framework, which is
applicable to any transmission schemes with
or without NC, to maximize the network
throughput for any wireless network
topologies. Our framework is featured with
multi-path routing that efficiently seizes the
wireless NC opportunities. Under the
developed framework, we propose a joint
link scheduling, channel assignment and
routing algorithm to approach the obtained
optimal network throughput.
• We utilize the developed framework to
quantitatively characterize the network
throughput gains of the conventional NC
and the analogy NC in various network
topologies where the type of routing
schemes, the number of channels, and the
number of radios may vary. We
also conduct extensive simulations to
evaluate the performance of our proposed
joint link scheduling, channel assignment,
and routing algorithm.
radios, it can communicate with more than
one neighbor at the same time over different
orthogonal channels. To analyze the
throughput gain of wireless NC over the
non-NC scheme for general network
topologies, we start with the system models
including network model and the wireless
channel/interference models.
A. The Network Model
We model the wireless network topology,
characterized by the nodes and the links
corresponding to pairs of nodes within direct
communication range and interference
range, as a directed graph G = (V, E, I),
where V represents the set of nodes, E is the
set of data links, and I is the set of
interference links. Note that E is the set of
links that can carry data, while I is the set of
links that can sense signals but not decode
the data. Let E−(v) and E+(v) be the sets of
incoming and outgoing links of node v with
v ∈ V , respectively. Denote by
e = (u, v) the directed link in the network k
from node u to nodev with u, v ∈ V . Let t(e)
and r(e) be the transmitting and receiving
nodes, respectively, of link e. Also, let e =
(v, u) be the reverse link of e = (u, v). There
are multiple orthogonal channels over each
link in the network. Let M and _M_ be the
set and the number of these channels,
respectively, over each link in the network.
The network is exploited by a number of
sessions to transport data packets. Denote
the set of sessions by A. A session a, with a
∈ A, is characterized by a triplet {s(a), d(a),
θ(a)}, where s(a), d(a), and θ(a) are the
source node, destination node, and
throughput, respectively, of session a.
Packets of session a with a ∈ A are routed
SYSTEM MODELS FOR GENERAL
NETWORK TOPOLOGIES
We then consider the wireless ad hoc
networks with general network topologies,
where there exist multiple unicast sessions.
The wireless spectrum is divided into a set
of orthogonal channels. Each node is
equipped with either a single radio or
multiple radios. If a node has multiple
12
from s(a) to d(a) in multiple hops if there is
no directed link between the source and
destination nodes. Every node in the
wireless network can be a source or
destination, i.e., s(a), d(a) ∈ V, ∀a ∈ A.
There may be multiple routes for session a
from s(a) to d(a). Let Pa be the set of
available routes/paths for session a. For a
path P ∈ Pa of session a, it can be
considered as an ordered subset of links, P =
{e0, e1, ・, end }, such that t(e0) = s(a) and
r(eNa) = d(a). For any given path P, link e,
and node v, we use e ∈ P to represent that
link e is on path P and v ∈ P to represent
that node v is on path P. Furthermore, we
use e1e2 ∈ P to denote that the path P
includes links e1 and e2, and the link e2 is
immediately behind e1, i.e., r(e1) = t(e2). B.
Wireless Channel/Interference Model We
denote by DT and DI the transmission range
and interference range, respectively. Because
DI is always larger than or equal to DT in
practice, let DI = αDT with α ≥ 1. Let h(u, v)
be the Euclidean distance between nodes u
and v. This paper adopts the protocol model
of interference [24]. There exists an edge e =
(u, v) ∈ E, if and only if h(u, v) ≤ DT , which
implies that nodes u and v can communicate
directly in one hop. Let cm(e) be the date
rate of link e over channel m. This is the
maximum data rate at which node t(e) can
communicate with node r(e). There exists an
edge
i = (u, v) ∈ I, if and only if DT ≤ h(u, v) ≤
DI, which implies that nodes u and v cannot
communicate directly in one hop, but can
interfere with each other. The definition of
interference link set I captures the behavior
of the carrier sense multiple access with
collision avoidance (CSMA/CA) featured by
IEEE 802.11 medium access control (MAC)
[25]. In light of carrier sensing, a
communication between nodes u and v can
block all transmissions within distance DI
away from either u or v.[3]
5. Throughput Analysis of Cooperative
Mobile Content Distribution in Vehicular
Network using Symbol Level Network
Coding.
Mobile content distribution (MCD)
is a promising rvice in VANETs, where
multimedia contents are distributed from
one or more fixed access points (APs) to the
vehicles driving through an area of interest
(AoI). Examples of MCD services include:
live video broadcast of road traffic and
weather conditions; periodic broadcast of
multimedia advertisements from local
businesses; updates of the GPS map about a
city. However, the rovisioning of MCD in
VANETs meets several challenges. On one
hand, multimedia contents containing audio
and video usually require high downloading
rate or end-to-end throughput. On the other
hand, wireless is a well-known lossy
medium with limited bandwidth, in which
channel fading and various interference
dramatically reduce the hroughput. The high
mobility of VANETs, leading to fast and
unpredictable topological changes, will
further exacerbate the frequent packet losses
and collisions.
Network coding (NC) is a common
technique adopted in
MCD. NC basically breaks the store-andforward packet forwarding paradigm by
allowing intermediate nodes to combine
received packets. Since each coded packet
could
benefit
multiple
receivers
simultaneously, the bandwidth efficiency
can be improved. More recently, symbol13
level network coding (SLNC) has been
proposed . By combining packets at symbol
level, SLNC allows a node to recover
correctly received symbols from erroneous
packets. Hence, SLNC provides increased
successful packet reception rate due to its
better error tolerance. However, there is a
lack of theoretical foundation and
understanding on the performance limits
such as achievable throughput by SLNC,
especially in realistic conditions such as high
mobility MCD scenario in VANETs. In this
paper, we endeavor towards bridging the
above gap by first establishing a theoretical
framework to analyze the achievable
throughput of SLNC in general wireless
networks, and then apply our model to study
MCD in a highway VANET scenario. We
consider a line-shaped road topology where
the vehicles are positioned according to
some distribution, and they cooperatively
broadcast continuous contents coming from
the AP to all the vehicles inside an AoI
using SLNC. We are interested in answering
two questions: 1) Regarding realistic issues
of channel fading, interference and node
distribution, how does the achievable
throughput at a node change with its
distance from the AP? 2) Given a specific
vehicle mobility pattern, what is the
downloading volume a vehicle can obtain
from the AP during the time period it drives
through the AoI? The main contributions of
this paper can be summarized as
follows: (1) We propose an analytical model
to compute the achievable throughput of
single-flow unicast and multicast using
SLNC in general wireless networks, based
on network flow and queueing theory. (2)
We apply the above model to derive the
expected
achievable
throughput
of
cooperative MCD system at a certain
distance from the source using PLNC or
SLNC in VANETs. Then, we obtain the
expected downloading volume of a vehicle
with a certain vehicular mobility pattern. (3)
We demonstrate numerical results from our
model. Our findings provide insights on
optimized choices for cooperative MCD
system design in VANETs.
RELATED WORK
A. Capacity Scaling Law of Wireless
Networks with Network Coding
The previous works focused on
deriving the information theoretic capacity
of wireless network with NC. An2 other
similar approach reveals the capacity of NC
in a wireless network via the “asymptotic
throughput”. and studied a dense network
model considering interference and noise to
obtain its asymptotical throughput. In
contrast, in this paper we consider an
extended network model with constant node
density, which is more realistic for the
VANETs Lun et al.proposed a theoretical
model to compute the exact capacity region
of random linear network coding in both
wired and wireless networks. This is the
most related work in terms of technical
approach, but we have several key
differences: 1) we model the achievable
throughput of SLNC instead of PLNC; 2) we
provide a method to derive the achievable
throughput for MCD in VANETs under
practical conditions including channel
fading,
medium
contention,
symbol
correlation and vehicle distribution; 3) from
our results, we give several insights on the
design of cooperative MCD systems in
VANETs.
14
Unlike PLNC, SLNC operates in the
granularity of symbols. The readers may
refer to for more detailed description of
SLNC coding operation. Before we elicit the
system model, first of all, we must have a
scheduling strategy to determine how
packets are injected onto each arc. However,
the packet scheduling problem is a difficult
problem
with
many
existing
works]contributing to address this problem.
In our generic framework, we assume a
scheduling scheme is predefined. The packet
transmission in our model works as follows:
each node stores packets that it receives,
then new packets are generated for injection
by SLNC whenever the node obtains a
transmission opportunity through the
scheduling scheme. Consequently, each
node potentially always has packets stored in
its memory, and injection can happen
whenever it gets a chance. Wireless network
is modeled as a directed hypergraph H =
(N,A), where N is the set of nodes, A is the
set of(a) Hypergraph (for modelling PLNC)
(b) Multi-Hypergraph (for modelling SLNC)
B. Achievable Throughput of MCD in
VANETs
Recently, SLNC is introduced into
cooperative MCD to further improve the
downloading performance [9], [10].
However, there still lacks an analytical
model to compute the achievable throughput
of SLNC in MCD system and to quantify the
gain of it compared with PLNC. There are
only a few works studying the capacity of
unicast or multicast in a VANET setting.
The asymptotic transport capacity of
VANETs was studied in , and they derived
the achievable throughput in VANETs
without considering interference and vehicle
distribution. In, Johnson et al. considered a
similar scenario to ours, and derived the
achievable throughput without taking into
account channel fading or vehicle
distribution; as a result, their achievable
throughput has no connection with the
source destination distance and node density.
ACHIEVABLE THROUGHPUT OF
SYMBOL-LEVEL
NETWORK CODING IN WIRELESS
NETWORK
In this section, we first put forward a
system model to analyze SLNC in generic
wireless network. Then based on this model,
we propose a theoretical framework using
queuing theory and flow network to obtain
the achievable throughput of SLNC.
Fig. 1: Graph models for PLNC and SLNC
A. System model for SLNC in Wireless
Network
15
hyperarc is decomposed into M injections of
symbols to M hyperarcs. Before delving into
the achievable throughput of SLNC with the
multi-hypergraph model, we present the
definition of achievable throughput:
Definition 1 (Achievable throughput): The
feasible flow rate a source node can forward
to its destinations over a long term.
B. Achievable Throughput for SLNC in
Wireless Network
Fig. 2: two-link tandem network. Left:
packet level queueing network; right:symbol
level multi-queueing network. hyperarcs. A
hyperarc (i, J) represents a lossy broadcast
link, where i is a node from N, J is a nonempty subset of N. Some injected packets on
(i, J) are received by a set K which is a
subset of J. We define the average rate at
which packets are injected to the hyperarc (i,
J) and is received by exactly K ⊂ J as ziJK.
We believe the above hypergraph model is
an accurate abstraction of packet-level
wireless network, thus is a suitable model
for analyzing PLNC. On the other hand,
SLNC encodes the information at symbol
level, which makes the hypergraph model
inappropriate
without
capturing
the
reception status of each symbol in the
packet. In order to fully capture SLNC
performance, we convert the hypergraph
model into a multi − hypergraph HM =
(N,AM), where M is the number of symbols
contained
in
each
packet.
In
multihypergraph, as in Fig. 1(b), there are M
corresponding hyperarcs {a1, a2, ..., aM} in
the hyperarc sets AM all with the same
starting node i and end sets J. We call these
M related hyperarcs as multi − hyperarcs.
Conceptually, hyperarc am corresponds to
the transmission link of the mth coded
symbol. By virtual of multi-hypergraph, the
process of one packet injected to one
In this section, we give our general
results for achievable throughput of SLNC
starting from special cases. We first consider
a two-link tandem network, which is a
multigraph shown in Fig. 2. The tandem
network with PLNC has been studied in , in
which the propagation of innovative packets
through a node follows the propagation of
jobs through a single-server queueing
station, which can be analyzed using fluid
approximation for discrete-flow networks.
The achievable throughput with PLNC is
then proven to be determined by the average
packet arrival rate on each arc. On the other
hand, the tandem network with SLNC has M
arcs between two nodes (See Fig. 2). We
virtually maintain 3 one queue for each
symbol position at the queueing station of
each node, so that each node maintains M
multiple queues. Then, the propagation of
innovative symbols through arc (i, j) can be
regarded as the propagation of jobs through
a multiple single-server queueing system
consisting of M single-server queueing
stations. In the following, we demonstrate
that the single-server queueing station at
each symbol position works the same way as
the single-server queueing station of
PLNC.As we mentioned above, [8] shows
that the propagation of packets with
16
innovative information can be modeled as a
single-server queueing station, which results
in the achievable throughput determined by
the average packet arrival rate on each arc, if
the average rate exists. So assume the
average packet arrival rate exists, if the
arrival process for each symbol has a same
average rate, the relationship between
achievable throughput and average arrival
rate still holds for each symbol.[2]
properties in achieving the maximum
network throughput, and develop a lineartime greedy algorithm, named Top (𝐾−1)
Full-Speed Channel Assignment (TFS-CA),
accordingly. Moreover, TFS-CA is topology
preserving, i.e., all links preserve the
connectivity from a single channel scenario,
which could facilitate individual routing
protocols [3]. Extensive simulations are
conducted to assess its performance in
comparison with the algorithm of Niranjan,
random channel assignment, and a tight
upper bound we derived . The simulation
results have demonstrated that TFS-CA is
far superior to two considered algorithms
and much improves the network throughput.
Lin and Chou conducted simple experiments
to clarify the multi-rate sharing problem and
measure aggregate throughputs during the
contention period. Based on the property of
fair medium access in CSMA/CA, they
derived a numerical formulation to estimate
the long-term achievable link throughput
with the interference consideration. Their
formulation is not only useful to understand
the effects of interferences, but also the
maximum objective in this paper.To
describe the formulation and make the
coming presentation clear, we introduce a
number of notations first: 𝑁 Number of links
𝐾 Number of orthogonal channels 𝑑𝑖 Data
rate of link 𝑖 𝐿𝑆𝑘 Set of links on channel 𝑘
𝑛𝑘 Number of links on channel 𝑘 𝑇ℎ𝑘
Throughput of all links on channel 𝑘
𝑇ℎ𝑡𝑜𝑡𝑎𝑙 Total network throughput In the
single-hop environment, all links interfere
with each other, and each node is assumed to
have the whole information of network
topology. The interference formulation
introduced by Lin and Chou stated that if
link 𝑖 uses channel 𝑘, then its data rate 𝑑𝑖
6. TOWARD MAXIMIZING
HROUGHPUT IN MULTICHANNEL
MULTIRATE WIRELESS NETWORKS
IEEE
802.11
based
wireless
networks have received considerable
attention in recent years, primarily owing to
less cost and easier deployment without a
pre-existing
infrastructure.
IEEE
802.11a/b/g standard supports multichannel
and multirate settings to facilitate network
transmissions. For example, IEEE 802.11a
and b/g provide 12 and 3 orthogonal
channels, respectively. A great number of
works aim at maximizing the network
throughput by considering concurrent
transmission and automatic configuration of
nodes to use more channels. In multirate
wireless networks, interference on cochannel is a notorious factor against
performance improvement as it leads to the
so-called performance anomaly , which
makes higher data rate links act like lower
ones.One common technique to overcome
performance anomaly is an appropriate
allocation of channels to communicating
links. In this letter, we consider the singlehop multichannel multirate networks, and
adopt the throughput formulation derived
from to calculate the channel throughput
under interferences. We show useful
17
will be degraded to (Σ 𝑗∈𝐿𝑆𝑘 1/𝑑𝑗 )−1,
where 𝐿𝑆𝑘 is the index set of links
communicating on channel 𝑘. Since there
are ∣𝐿𝑆𝑘∣ = 𝑛𝑘 links using channel 𝑘, the
channel throughput 𝑇ℎ𝑘 equals to 𝑛𝑘 (Σ
𝑗∈𝐿𝑆𝑘 1/𝑑𝑗 )−1 . Then, our objective is to
find 𝐿𝑆1,𝐿𝑆2, . . . ,𝐿𝑆𝐾 such that 𝑇 ℎ𝑡𝑜𝑡𝑎𝑙 =
𝐾 𝑘=1 𝑇ℎ𝑘 can be maximized in a singlehop multichannel multirate wireless
network.[5]
case for wireless sensor networks which
need to collect a large amount of data, such
as multimedia sensor networks. Wireless
mesh networks in developing countries,
whose nodes have occasional access to the
power grid, serve as another example. We
focus on the case of fixed scheduled wireless
networks. More precisely, we consider
wireless networks that are operated by
scheduling link transmissions to be conflictfree, as opposed to a random access medium
access control (MAC) protocol. There are
several reasons for focusing on such
networks. First, upcoming standards, such as
IEEE 802.16 and Long-Term Evolution
(LTE) for fourth generation (4G) cellular
networking, support completely scheduled
modes of operation, whereby link
transmissions in the network can be
precisely controlled and scheduled to be
conflict-free. Second, our centralized and
scheduled solutions bound the performance
of distributed solutions and serve as
benchmarks
for
designing
wireless
networks. We also focus on the case where
the traffic requirements are static. Both
IEEE 802.16 and LTE are envisaged for
static deployment of wireless networks,
which would carry aggregated traffic from
several users, thereby motivating aggregated
(and relatively static) traffic requirements.
Hence, we use a fluid model of data, i.e., we
offer long-term guarantees on network
throughput. We assume that the network is
specified in terms of a set of nodes and a set
of flows described in terms of their origin
and destination. We use a realistic
interference model based on the Signal-toInterference-and-Noise Ratio (SINR) for
modeling the conflicts to avoid when
scheduling the wireless links. This conflict
set model (also used in [12], [14]) captures
6. Throughput-Lifetime Trade-Offs in
Multihop Wireless Networks under
an SINR-Based Interference Model
Imroving the network throughput
(how fast the network may deliver data) and
improving the network lifetime (how long
the network may last) are two important
design objectives for multihop wireless
networks. These two objectives appear to be
in conflict with each other intuitively, higher
throughput
means
greater
energy
consumption, and hence reduced network
lifetime. In order to be able to balance these
two design objectives, it becomes important
to identify the trade-offs between throughput
and lifetime. In particular, it is not clear if it
is possible to set the transmit power in a
network, so as to improve both throughput
and lifetime. Further, how much throughput
needs to be traded off to achieve a certain
improvement in lifetime? Our aim is to
investigate the trade-off between throughput
and lifetime in multihop wireless networks,
to address questions of the above nature. We
consider scenarios where it is required to
achieve an adequate network throughput
(not necessarily the maximum) as well as a
sufficiently long lifetime. This would be the
18
the fact that the interference to a certain link
is the cumulative interference from the
multiple links that are activated during the
same period of time . Our notion of the
following three optimization problems: P1.
to maximize the network lifetime while
achieving the max-min network throughput,
P2. to maximize the network throughput
while achieving a pre-specified network
lifetime, and P3. to maximize the network
lifetime while achieving a fraction of the
max-min throughput. A solution to any of
the problems above is a network
configuration,
which
achieves
the
corresponding objective. By network
configuration, we mean the complete choice
of parameters for operating the network
including the set of links, the link
transmission schedules, and the routes for
the flows. In other words, we jointly select
flow routes and link schedules to achieve the
desired objective. In order to solve problems
P1, P2, and P3, and study the trend of their
solutions as a function of transmit power, we
adopt the following approach. We assume a
network-wide reference power level. All the
nodes may use the reference power, or a
finite number of power levels which have a
fixed offset from the reference power. Nodes
may also use a finite set of modulation and
coding schemes. We solve the problems P1,
P2, and P3 for a fixed setting of the
network-wide
reference power (and
possibly, a fixed set of offsets), and study
the trend of the solution by varying the
reference power. We consider this model to
be a realistic representation of the
capabilities of modern wireless radios, as
compared to using the Shannon capacity
formula to model wireless link rates. For
ease of exposition, we assume first that all
the nodes use a single power level (the
reference power) and a single modulation
and coding scheme. We will later show how
to extend this case to multiple power offsets
and modulation and coding schemes.Even in
the case of a single power and
modulation,solving these problems requires
searching among combinatorially many
configurations due to the intricate conflict
structure, we have developed computational
tools based on column generation to
circumvent this issue, which we use to
obtain all the numerical examples in this
work. It must be emphasized that all our
solutions are exact. Whereas our formulation
makes no assumptions on the traffic flows,
for our numerical results we consider traffic
patterns that converge to a gateway/sink.
Examples of such traffic patterns can be
found in wireless mesh and sensor networks,
where only a few gateways or sinks attract or
initiate traffic. We have also studied cases
where all flows are initiated by the
gateway/sink. We have not observed any
major differences in trends, and hence we do
not present them for lack of space. Note that
our analytical results are very general, and
would apply regardless of the traffic pattern.
Both our analytical and numerical results
show that the optimal trade-offs between
throughput and lifetime are usually not
obtained at the minimum power that enables
network connectivity. In addition, our results
show that, by fixing the throughput
requirement, the lifetime is not a
monotonically decreasing function of the
transmit power. Finally, for a given power
level, our results indicate the existence of a
throughput threshold, below which a small
sacrifice of throughput leads to a large (more
than proportional) improvement of lifetime
and beyond which a reduction in throughput
only leads to a proportional improvement in
19
lifetime. We provide both theoretical and
intuitive explanations for these phenomena
in the paper. In addition to the above, we
highlight the importance of the interference
model and point out why results based on
the interference range model (to be defined
later) may be misleading.
7. Throughput Calculation in a
HIPERLAN Type 2 Network Considering
Power Control and Link Adaptation
An important quality parameter in
engineering wireless voice networks such as
the
global
system
for
mobile
communications network is the percentage
of coverage area where the achievable
signal-to-interference ratio (SIR) is above
the minimum value. This minimum value is
usually defined by a maximum allowed bit
or packet error rate (PER) that still
guarantees a minimum level of perceived
communication quality. This can be
translated into a utility function, i.e., a
function that indicates the usefulness of an
SIR level to the user. An example utility
function for voice services is shown in Fig. 1
as, e.g., proposed in [1]. The statement of
this utility function is that full utility is given
above the SIR threshold whereas the utility
drops abruptly to zero below the SIR
threshold. Note that this may not be an exact
model, but it is the one used to dimension
voice networks. The inverse of the covered
area with sufficient SIR is the area where
users cannot get a service with sufficient
quality, the so-called outage areas. which
uses the utility function for voice in Fig. 1,
can be used to determine the required
frequency reuse distance between radio cells
with a certain radius, a given required SIR,
and an outage probability. It is a popular
initial estimate in voice radio network
planning. The IRF, however, is not well
suited for many wireless data networks. This
is especially true if the perceived quality
degrades smoothly with decreasing SIR. A
possible quality measure is the user
throughput. With an automatic repeat
request (ARQ) scheme, the throughput T is
given as a function of the PER, which is a
function of the SIR, and the raw data rate on
the physical layer R. An example utility
function for data networking, based on the
throughput, has been proposed in and is
shown in Fig. 1. It expresses the fact that,
e.g., a web user is entirely satisfied as long
as the web pages appear “quick enough.” He
gets more and more unsatisfied as the delay
becomes longer, until he finally stops using
the web because the waiting time becomes
unacceptable. This satisfaction is closely
connected with the achievable throughput
and the resulting delay.
20
can be used to derive an indication of radio
network behaviour but is not suited for the
dimensioning of network installations. A
comparable measure in the domain of voice
networks is the IRF, which is based on
similar assumptions as the model in this
paper. The IRF is commonly used for an
initial indication of basic network
parameters but is too simplistic for final
network planning. The advantage of such
analytical models is that, compared with,
e.g., simulation models, they are simple and
require only little time to yield initial results.
So the methods to engineer wireless voice
networks are not necessarily suited for
wireless data networks. This paper proposes
a new evaluation method using the
throughput as utility. It relies on HIPERLAN
type 2 (H2) as the underlying technology [3].
However, the basic approach is widely
generic and can be reused for other network
technologies. The main achievement of this
paper is a theoretical model to evaluate
network throughput. Transmit power control
(TPC) parameters are fixed for each
experiment, where the influence of power
control is considered by performing many
experiments with different TPC parameter
combinations, i.e., it is only link adaptation
(LA) that is dynamic in this paper. As is well
known, radio network investigations have to
take into account a huge number of
parameters, making it difficult, if not
impossible, to find analytical models. The
analytical model proposed in this paper,
therefore, has been developed for a
simplified regular scenario with circular
radio cells, equal spatial distribution of
terminals inside cells, and with a
deterministic propagation model. The
chosen scenario reflects most properties of
radio networks. This means that the model
SOME BASICS AND RELATED WORK
A. Basics of H2
A number of publications about H2
are available.The explanations given here
are those that are necessary for the
understanding of this paper. LA is the ability
to transmit high data rates when the channel
is good, i.e., at high SIR, and low data rates
at bad channel conditions, i.e., at low SIR. It
is mainly the physical layer of H2 that
supports LA. Seven Phy modes exist, as
defined in. Each Phy mode PM has its raw
data rate RPM. However, with decreasing
SIR, the share of packets that are received
with errors increases. H2 provides a
selective ARQ that retransmits incorrect
packets. With the PER and RPM, the useful
throughput after ARQ TPM is given by TPM
= RPM(1 − PER). (1) The throughput curves
that are used throughout this paper are taken
from [3] and are shown in Fig. 2. They are
approximated with the functions in [6] for
all investigations in this paper. Note that the
envelope of the curves is similar to the
utility function for wireless data networks in
Fig. 1. Obviously, the highest possible
21
throughput in a system with interference can
be achieved by always selecting the Phy
mode with the highest 1The Phy mode with
9 Mb/s is not used here because it has worse
performance than the one with 12 Mb/s. Fig.
2.
Theoretically
achievable
useful
throughput with the Phy mode PM as
parameter.throughput for the current γ,
denoted as Tmaxγ(PM), where γ is
another notation for SIR.We define R
maxγ(PM) as the function that always
selects the Phy mode with the highest
throughput for a given γ. H2 is a connectionoriented system with a central medium
access control (MAC), where an access
point (AP) determines and announces its
transmissions in the downlink (DL) and
those of mobile terminals (MTs) in the
uplink (UL). The air interface is divided into
MAC frames of equal duration and the
scheduled transmissions can vary from
MAC frame to MAC frame. The AP
scheduler has knowledge about the amount
of data in each buffer that waits to be
transmitted and allocates transmission
opportunities to connections. During
scheduling, the AP has to take into account
the Phy mode each connection has in UL
and DL. Since scheduling is not the focus of
this paper, we will not go into the details of
this subject. For further discussion, see, e.g.,
.The scheduling scheme used in this
publication tries to find a balance between
allocation of time shares and amount of data.
We assume Nc active connections and each
connection i, i ∈ 1, . . . , Nc, uses the Phy
mode PMi with raw data rate RPMi . The
scheduling strategy works such that the time
share of connection i is proportional to
FPMi , where FPMi = 54 Mb/s RPMi . (2)
As we will see later, FPMi contains all we
need in the mathematical models to
represent the scheduling scheme, so the
models are applicable to any scheduler for
which FPMi can be computed. The only
justification for the specific choice in this
paper is that some scheduling scheme had to
be chosen among others with the purpose of
making simulation and theoretical models
comparable. It is worth mentioning that the
results in subsequent sections have been
obtained for this specific choice of
scheduling scheme and that other scheduling
schemes may lead to different outcomes. For
the purpose of TPC, the AP transmits with
constant transmit power Pt,AP during a
MAC frame, where Pt,AP can vary from
MAC frame to MAC frame. Moreover, the
AP broadcasts Pt,AP in each MAC frame
together with its expected receive power
Pe,AP.[19]
III. SUMMARY AND CONCLUSIONS
In this paper, explore the throughput and
delay in wireless ad hoc networks. Our
objective is to achieve high throughput
while keeping the acket delay relatively
small. In order to solve this problem, we
start from the simplest model, compute the
capacity only, then add more assumptions
step by step, and finally find out a routing
algorithm which can chieve our objectives.
This is a very mportant methodology for any
kind of research.For wireless ad hoc
networks with only static nodes, the capacity
per node is
bits per second for Arbitrary Network model
and for Random Network model. If mobility
is considered in the network, the capacity
22
can be dramatically improved to 1per
S-D pair. Furthermore, if more assumptions
on the traffic pattern and mobility pattern are
introduced, the proposed routing algorithm
can guarantee the packet delay and achieve a
close-to optimal capacity, which is only a
poly-logarithmic factor off from the optimal
algorithm. Note that we have no
considerations on the energy
limitation of the nodes in the network, which
is another important constraint actually
existing in the wireless ad hoc networks. We
show that the difference in achievable
throughput between SLNC and PLNC is
determined by symbol-level diversity. We
then propose a method to analyze the
expected
achievable
throughput
of
cooperative MCD system with SLNC in
VANETs by considering realistic factors.
Furthermore, by considering vehicular
mobility pattern, we compute the expected
downloading volume of a vehicle passing
through an AoI. Through numerical results,
we reveal the impacts of using PLNC &
SLNC under different conditions to the
throughput limits of MCD, which provide
valuable insights for the cooperative MCD
system design.
2.
Throughput Analysis of Cooperative
Mobile Content Distribution in
Vehicular Network using Symbol Level
Network Coding (Qiben Yan, Student
Member, IEEE, Ming Li, Member,
IEEE, Zhenyu Yang, Member, IEEE,
Wenjing Lou, Senior Member, IEEE,
Hongqiang Zhai, Member, IEEE)
3.
IEEE JOURNAL ON SELECTED
AREAS IN COMMUNICATIONS,
VOL. 27, NO. 5, JUNE 2009
(Modelling Throughput Gain of
Network Coding Multi-ChannelmultiRadio Wireless Ad Hoc Networks)
Hang Su, Student Member, IEEE, and
Xi Zhang, Senior Member, IEEE
4.
IEEE TRANSACTIONS ON MOBILE
COMPUTING, VOL. 8, NO. 11,
NOVEMBER
2009
(Throughput
Behavior in Multihop Multiantenna
Wireless Networks Bechir Hamdaoui,
Member, IEEE, and Kang G. Shin,
Fellow, IEEE)
5.
Toward Maximizing Throughput in
Multichannel
Multirate
Wireless
Networks Ying Chih Lin, Richard
Chun-Hung Lin, and Tai-Wei Kuo
D.C., 20004-2505.
6.
Restricted Mobility Improves DelayThroughput Tradeoffs in Mobile Ad
Hoc Networks Michele Garetto,
Member, IEEE, and Emilio Leonardi,
Senior Member, IEEE
7.
Throughput-Lifetime Trade-Offs in
Multihop Wireless Networks under an
SINR-Based Interference Model Jun
References
1.
IEEE TRANSACTIONS ON MOBILE
COMPUTING, VOL. 11, NO. 3,
MARCH 2012 SmoothTrade-Offs
between Throughput and Delay in
Mobile Ad Hoc Networks (Pan Li,
Member, IEEE, Yuguang Fang, Fellow,
IEEE, Jie Li, Senior Member, IEEE,
and Xiaoxia Huang, Member, IEEE.)
23
Luo, Member, IEEE, Aravind Iyer, and
Catherine Rosenberg, Fellow, IEEE
8.
9.
Member, IEEE, Ming Li, Member,
IEEE, Zhenyu Yang, Member, IEEE,
Wenjing Lou, Senior Member, IEEE,
Hongqiang Zhai, Member, IEEE.
Throughput Optimization in Mobile
Backbone Networks Emily M. Craparo,
Member, IEEE, Jonathan P. How,
Senior Member, IEEE, and Eytan
Modiano, Senior Member, IEEE
14. On the Throughput Capacity of
Wireless Sensor Networks with Mobile
Relays Wang Liu, Kejie Lu, Jianping
Wang, Liusheng Huang, and Dapeng
Oliver Wu
Cross-Layer Throughput Optimization
With Power Control in Sensor
Networks Maggie X. Cheng, Xuan
Gong, Lin Cai, Senior Member, IEEE,
and Xiaohua Jia, Senior Member, IEEE
15. On the Delay-Throughput Tradeoff in
Distributed Wireless Networks Jamshid
Abouei, Member, IEEE, Alireza
Bayesteh, Member, IEEE, and Amir K.
Khandani, Member, IEEE
10. Cognitive
Networks
Achieve
Throughput Scaling of a Homogeneous
Network Sang-Woon Jeon, Student
Member, IEEE, Natasha Devroye,
Member, IEEE, MaiVu, Member, IEEE,
Sae-Young Chung, Senior Member,
IEEE, and Vahid Tarokh, Fellow, IEEE
16. Final Report of EE359 Class Project
Throughput and Delay in Wireless Ad
Hoc
Networks
Changhua
He
changhua@stanford.edu.
11. Using Partially Overlapping Channels
to Improve Throughput in Wireless
Mesh Networks Yong Ding, Yi Huang,
Guokai Zeng, Li Xiao Department of
Computer Science and Engineering
Michigan State University
17.
12. MAC Centered Cooperation Synergistic
Design of Network Coding, MultiPacket Reception, and Improved
Fairness
to
Increase
Network
Throughput Jason Cloud, Student
Member, IEEE, Linda M. Zeger, Senior
Member, IEEE, and Muriel M´edard,
Fellow, IEEE
13. Throughput Analysis of Cooperative
Mobile Content Distribution in
Vehicular Network using Symbol Level
Network Coding Qiben Yan, Student
Throughput
of
CDMA
Data
NetworksWith Multiuser Detection,
ARQ, and Packet Combining Ben Lu,
Member, IEEE, Xiaodong Wang,
Member, IEEE, and Junshan Zhang,
Member, IEEE
18. Computing Throughput Capacity for
Realistic Wireless Multihop Networks
Patrick
Stuedi
Gustavo
Alonso
epartment of Computer Science Swiss
Federal Institute of Technology (ETH
urich) 8092, Zurich, Switzerland.
19. Throughput Calculation in a HIPERLAN
Type 2 Network Considering Power
Control and Link Adaptation Markus
Radimirsch, Member, IEEE, and Klaus
24
Jobmann, Member, IEEE
20.
IEEE
26. Optimal Scheduling and Routing for
Maximum Network Throughput Emilio
Leonardi, Member, IEEE, Marco
Mellia, Member, IEEE, Marco Ajmone
Marsan, Fellow, IEEE, and Fabio Neri,
Member, IEEE
Cell Dimensioning and Network
Throughput in Cellular Multi-Hop
Relay Networks K.R. Jacobson, W.A.
Krzymie´n
TRLabs/Electrical
and
Computer Engineering, University of
Alberta Edmonton, Alberta, Canada
27. Optimal Delay–Throughput Tradeoffs in
Mobile Ad Hoc Networks Lei Ying,
Member, IEEE, Sichao Yang, and R.
Srikant, Fellow, IEEE
21. Analysis and Modeling of Upstream
Throughput in Multihop Packet CDMA
Cellular Networks Ali Nabi Zadeh,
Member, IEEE, and Bijan Jabbari,
Senior Member, IEEE
28. Smart-Antenna System for Mobile
Communication Networks Part 2:
Beamforming and Network Throughput
Salvatore Bellofiore, Jeffrey Foutz,
Constantine A. Balanis, and Andreas S.
Spanias Department of Electrical
Engineering,
Telecommunications
Research
Center
Arizona
State
University, Tempe, AZ 85287-7206
USA
22. Link Scheduling With Power Control for
Throughput Enhancement in Multihop
Wireless Networks Jian Tang, Guoliang
Xue,
Senior
Member,
IEEE,
Christopher Chandler, and Weiyi Zhang
23. An Analysis of Traffic and Throughput
for UMTS Packet Core Networks Ye
Ouyang and M. Hosein Fallah,
Ph.D.,P.E. Howe School of Technology
Management, Stevens Institute of
Technology, NJ, USA
24. Optimal Throughput-Delay Scaling in
Wireless Networks – Part I: The Fluid
Model Abbas El Gamal, James ammen,
Balaji Prabhakar, and Devavrat Shah
25. On the Calculation of the Maximal
MOT Throughput in T-DMBJihoon
Choi,
Student
Member,
IEEE,
Donghwan Lee, Student Member,
IEEE,Jieun Yu, Student Member, IEEE,
Kyunghwi Kim, Student Member,
IEEE,and Wonjun Lee, Senior Member,
25
26