Effective Streaming of Clustered Sensor Data in Harsh Environment

Y. Morgan, T. Kunz & M. El-Gindy
International Journal of Computer Networks (IJCN), Volume (3) : Issue (3) : 2011 178
Effective Streaming of Clustered Sensor Data in Harsh
Environment
Y. Morgan
Software Systems Engineering
University of Regina
Regina, SK, CANADA
yasser.morgan@uregina.ca
T. Kunz
Computer Systems Engineering
Carleton University
Ottawa, ON,CANADA
tkunz@sce.carleton.ca
M. El-Gindy
Automotive Engineering
University of Ontario Institute of Technology
Oshawa, ON, CANADA
moustafa.el-gindy@uoit.ca
Abstract
A milestone of success for any sensor network establishment is successful data streaming over
point-to-point communications (P2P). Typical P2P services would present the descriptive best-
effort (BE) or other Quality of Service (QoS) streams. In this research, we present the Stateless
Wireless Ad-hoc Network model (SWAN), an integrated model that is known to work in typical ad-
hoc configurations like the classical wireless ad-hoc sensor networks. SWAN is a lightweight QoS
model that enables operations over any routing protocol or Media Access Control (MAC) layers
while exhibiting some advantages over competing models. Nevertheless, SWAN is vulnerable to
troubles related to mobility and false admission. The SWAN model design relies on picking a
candidate (victim) data flow to manipulate as a congestion control measure. We extend SWAN by
adding the destination-based regulation and also show the reasons why the destination-based
regulation chooses real-time data victim streams in an accurate mode.
Thus, we propose the use of destination-based regulation to resolve the dynamic congestion
issues. This confines SWANs pertinence for streaming scenarios such as the instance of
essential continuous broadcast (CB-streaming). We consider it is possible to transport CB
communications over substitute MAC layer such as the 802.11 or other wireless technology. To
perform that, we introduce the purpose of SWAN in addition to the suggested enhanced
destination based algorithm to transport the trusted CB-streaming applying P2P communications,
but by rendering it with dissimilar QoS parameters. We believe that the SWAN approach in this
regard is more in line with the nature of sensor networking with the awareness that sensors
normally form and vanish quickly giving small opportunity to reconfigure or reposition the profiles
of dynamically formed networks. Hence, sensing devices maintain low computational, storage
capacities, processing, and battery life1.
Keywords: SWAN, P2P, Streaming.
1. INTRODUCTION
The recent advances in sensing and wireless technologies have facilitated the design of modern
systems that have never been possible. A typical approach uses probe sensors to collect data
1
This article is based in part on Y. L. Morgan, T. Kunz, "The Full ESWAN Destination-Based Approach: Operations And
Evaluation", in the 'discontinued' Journal of Information and Communication Technology, vol. 3, no. 2, pp. 30-41, June
2007.

from harsh industrial or natural environment where extreme pressures and temperatures reach
levels prohibiting common data collection methods. Another example is collecting information
from battlefield or from remote places like the arctic or even deep below ocean surface or
volcanic site. In all those scenarios, scientists are increasingly demanding the streaming of
information where sensors are required to deal with mobility, low battery and periods of long
inactivity. Yet sensor networks must maintain connectivity and lightweight management of it
exceedingly limited resources. Precisely, the instinctive solvent would certainly be to transport
data streams over point-to-point communications (P2P), which happens to be acquiring impulse
as the milestone of success for any sensor network establishment. In an effort to transport data
streams over P2P links, we require to back up more descriptive best-effort (BE) Quality of Service
(QoS). In this research, we spread out an integrated model, which was acquainted to work on
classical wireless ad-hoc sensor networks. SWAN is a lightweight conciliatory QoS model that
enables operations over any routing protocol or Media Access Control (MAC) layers while
rendering some vantages over competitor models. Nevertheless, SWAN is vulnerable to troubles
related to mobility and false admission. The core facet of SWAN (and in fact any QoS model) is
picking a candidate data flow to manipulate during which QoS warrantees are being
transgressed. We broaden SWAN by showing the destination-based algorithm and also show the
reasons why the destination-based algorithm chooses real time data victim streams in an
accurate mode. And so we furnish trial run results to examine and obtain the destination-based
method.
QoS support in sensor networks is certainly a dynamic research subject. Acknowledging the
belief that QoS frameworks as a way for the secured Internet will likely not be suitable for
networks, by using highly active topologies, research workers featured several QoS results as a
way for sensor networks. In the midst of various suggestions, SWAN has demonstrated high
levels of robustness and power onto retrieve right from the contrary mobility states compared with
FQMM, dQoS or INSIGNIA, as a result of its stateless design. SWAN could function over BE
MAC which includes IEEE 802.11 Distributed Coordination Function (DCF), and habituates an
integrated stateless distributed access to work out the active QoS disabilities. SWAN runs in a
fully decentralized mode with the intention to straighten out the sensing dynamics. SWAN
expends informant admission controller to confine the level of acknowledged real time flow rates.
In response to environmental dynamics that leads over to periodic congestion, SWAN expends
explicit congestion notification (ECN) to dynamically modulate real time traffic. Due to the fact that
intermediate sensors will likely not preserve per-flow state data, resolving congestion scenarios
turns a bit challenging. Yet, holding the tenet associated with a stateless model proceeds the
system unproblematic, scalable, robust, and ironically lightweight.
Besides, SWAN also conforms the technique of delicate real time service warrantees. Whenever
a real time flow rate is acknowledged about the network applying admission control, it is potential
over at any detail during the life time of the flow rate to be downgraded to best-effort/ to halt it in
reception over to network dynamics. Each source sensor should re-initiate a novel admission
procedure with the intension to re-establish the flow. This would be a knock down feature of
SWAN due to the delicate real time guarantees, which acts as a conventional reception to self-
healing feature the wireless sensor networks, may go through. Nevertheless, SWAN enforces no
actual service distinction grounded on exploiter profiles. This confines SWANs pertinence for
streaming scenarios such as the instance of essential continuous broadcast (CB-streaming). We
consider it is possible to transport CB communications over substitute MAC layer such as the
802.11 or other wireless technology. To perform that, we introduce the purpose of SWAN in
addition to the suggested enhanced destination based algorithm to transport the trusted CB-
streaming applying P2P communications, but by rendering it with dissimilar QoS parameters. A
number of us reckon that the SWAN approach in this regard is more in line with the nature of
sensor networking with the sensation that sensors normally form and vanish quickly considering
small opportunity to reconfigure or reposition the profiles of dynamically formed clusters. Hence,
sensing devices maintain low computational, storage capacities, processing, and battery life. The
dynamic shaping and diminishing of sensor sub-clusters is necessary to be handled and
modulated dynamically, autonomously and seamlessly without user intervention.

Regrettably, SWAN has disregarded the attribute of real time flowing delays as SWAN framework
utilizes per-link delays simply to observe congestion, and uses bandwidth states to alleviate
admission control. End-to-end delays have simply not been regarded in evaluating the quality of
real time flows in SWAN. In cases where source and destination sensors are away off each other,
in terms of number of hops, real time packets go through increased end-to-end delays. We found
our destination-based access on the basis that destination sensors could make better
assessments of the rate of obtained real time flows by means of end-to-end delays. Thus, we
propose the use of destination-based algorithm to resolve the dynamic regulation matters. In the
research, we summarize the relevant factors of the original SWAN model in Section 2, and then
in Section 3, we depict the setbacks of active regulation of real time flows. In Section 4, we
demonstrate the two proposals rendered throughout the original SWAN framework like, source-
based and network-based algorithms, and then criticize both proposals. In Section 5, we present
some terminologies necessary to discussing the active regulation and present the destination-
based algorithm. Furthermore, we describe the intellectual behavior associated with our
destination-based algorithm and explain the mechanism to limit the set of victim flows. In Section
6, the test-bed used to formalize the destination-based algorithm, and exemplify the test-bed, the
actual result, analyze, and glance at the suggested enhancements. Eventually, in Section 7 we
conclude the evidence and propose potential future research extensions.
2. COMMON SWAN OPERATIONS
SWAN can be described over two major components. The Admission Controller (AC) is
chargeable for including any novel flow to the sensor network. Admission control is carried out at
the original source sensor that basically initiates real time flows. The Rate Controller (RC) is
accountable for modulating BE traffic and keeping traffic loads at medium sensors and other
factors are a Classifier, that chooses real time packets on to get around the shaper, and a
Shaper, that stand for an simple leaky bucket traffic shaper. The purpose of the shaper would be
to hold up BE packets in compliance together with the pace calculated by the RC. Figure 1
instances the architecture of the SWAN framework.
Classifier
Shaper
MAC
IP admission
controller
rate
controller
shared media channel
admit /
reject
API
request
pre-marked
/ unmarked
send probe
receive probe
rate
markedpkts
unmarked
pkts
marked /
packet delay
utilization of real-time traffic
pkts
unmarked / ECN
FIGURE 1: Block Diagram of the SWAN Architecture.
Admission Controller (AC)
A mobile sensor would originate a real time flow rate just soon after finding a consent message in
SWAN. SWAN-AC follows the popular Additive Increase Multiple Decrease (AIMD) algorithm,
that's been utilized through the TCP congestion controller for many years. The TCP congestion
control algorithm make sure that the scheme does work around or rather about the cliff as

instanced on Figure 2, that guarantees maximum scheme throughput at the cost of having
prominent queues, and in addition longer average delays. The SWAN AIMD admission controller
algorithm complies a relatively cautious approach: SWAN maintains the scheme at the time-lag
knee, the spot where the system throughput is virtually identical to the cliff, but queues are
importantly far lower charged. SWAN employs MAC delay as a responses alternative to packet
loss because of the fact that typically happens at the cliff when delays evolve at the knee. This
approach is exemplified in Morgan 03, Xue 10, and fully depicted in Morgan 11.
Because of this cautious approach, SWAN resource utilization is lower than the available and
assuming that the remainder bandwidth will likely run through by BE traffic. Additionally, this drop
off bandwidth is usually considered as a safety step against the network dynamics which include
bandwidth fluctuation and mobility.
The AC sets out by directing a probe request on to the destination sensor and the medium
sensors intercept the request also, renovate it with the bottleneck bandwidth. Intermediate
sensors employees their AC onto estimate available bandwidth for newer real time flows, but they
cannot enforce resource, or bandwidth allocation. The destination sensor responds through a
probe reply to set the bottleneck bandwidth. Whenever the source sensor obtains the probing
response packet, it can perform the source-based admission control by contemplating end-to-end
bandwidth accessibility with the bandwidth demand as a way for the novel real time flow.
load
throughput
delay
delay
"knee"
congestion
control "cliff"
FIGURE 2: General Behavior of a Congestion Controlled System.
Classifier
The source sensor grades packets consorted with acknowledged real time flow as real time. The
classifier directs marked packets to the MAC layer instantly, getting around the shaper. SWAN
implicitly presumes that real time flows need not to be policed.
Shaper
The shaper is a simple leaky delay queue that enforces delays on BE packets based on feedback
from the Rate Controller (RC).
Rate Controller (RC)
The RC determines the link status utilizing link delay quantified by the MAC layer. For example,
the delay can be extracted expending the IEEE 802.11 DCF mechanism. The RC finds
extravagant link delay when one or maybe more packets have greater time-lags than a threshold
link delay d (sec). The threshold delay d lies in the real time delay demands as mentioned in Ahn
02. When substantial link delay is noticed, RC backs off the rate by r %. RC should re-adjust
parameters (d and r) each T seconds. The diagram shown in Figure 2 and most of this section
refer to Morgan 11.
3. DYNAMIC REGULATION ISSUES
SWAN presents active direction components in reply to stipulations produced by network
characteristics like sensor mobility and off key admission. It is substantial to describe the
presence of both topics on network resources.

s
n1
d
n2
moving out
moving in
s d
n2
Bandwidth rerouted
to n2
without admission
FIGURE 3: Congestion/Overload Due to Mobility.
Mobility
Figure 3 provides us the real time flow rates between sensors s and d can be rerouted off from
sensor n1 to sensor n2 because of mobility and precisely the underlying routing algorithm
executes the necessary rerouting. Even though sensor n2 will certainly go through growth in real
time traffic, it could not execute any admission procedure to allow the novel flow. Particularly,
backing up the new rerouted flows could induce the sensor to go through the congestion and this
state represents the congestion due to mobility.
False Admission
As instanced in Figure 4, sensors s1, s2, and s3 typically originates an investigation request to
direct real time flow rates to sensors d1, d2, and d3 via sensor n. If sensor n performs the three
requests after a short time, the admission controller on sensor n can admit the three flows though
it fairly has room for just one flow and that is because of the shortage of resource reservation in
SWAN. Right until real time packets take in usable bandwidth, sensor n may always acknowledge
novel real time flows. The resulting congestion at sensor n is known as congestion due to false
admission.
s1 d1
n
s2
s3 d3
d2
many probe
req. accepted
less can
be served
FIGURE 4: Congestion/Overload Due to False Admission.
It is vital to recognize the actual mobility and false admission but, it simply exemplifies two issues
around other issues related to network dynamics. SWAN follows the explicit congestion
notification (ECN) regulation algorithm to retrieve out from congestion terms induced by network
dynamics. Because sensors are continuously and independently keeping track of their bandwidth
usage all sensors would be able to notice violations. Clogged sensors are then going to apply the
ECN bits through the IP header of typical real time packets to endure destinations of the
presence of congestion. Every individual destination sensor is going to supply a modulate
message onto the relevant source sensor and the source sensors will instantly re-initiate new
probe demands to seek out an improved service route to the destination.
4. COMMON DYNAMIC REGULATIONS
The determination of congested sensors to tag packets using ECN is absurdly decisive; because
flows that are tagged with ECN may precede their QoS privileges. SWAN suggested two rules,
namely source and network-based regulations. Both regulations access mark ECN packets
differently.

1. Source-Based Regulation:
A congested sensor tags RT flows with Congestion Experienced (CE) applying the ECN bits.
When destination sensors run into packets with CE bit marked they direct the modulate content to
the associated source sensors. Source sensors instantly execute a multiplicative drop off on
relevant RT flows. As an effect, the congested sensor goes through a steady drop off in the
amount of RT traffic right until the congestion status is withdrawn and the intermediate sensor
stops marking CE bits. If the uncommitted reduced bandwidth of RT flow is ineffective to its
source sensor, it must substitute a random short of time and then re-initiate a probe asking to re-
establish the wanted level of service.
Source sensors have to distribute the re-initiation to prevent a flash-crowd status during which
sensors normally fall into other false admission once again, hence, the random backup time is
necessary. Source-based regulation powers RT flows going through one congested sensor on to
regulate. This method appears to be aggressive and could force more flows to be regulated even
if the period of bandwidth violation was constricted. Furthermore, it does not discriminate between
dissimilar RT flows.
2. Network-Based Regulation:
A congested sensor chooses a subset out of all the real time flows to be a victim flow set in a
network-based regulation and the congested sensor grades packets related with victim flows. It is
potential for a congested sensor to categorize a particular set of RT flows by implementing a
useful hash function with no demand to hold flow information. Packets of victim RT flows will
attain relevant destination sensors marked based on CE, later the network-based outlook
complies the similar procedure as depicted for source-based regulation. Supposing a congested
sensor does not go through any drop off with the period of real time traffic soon after a period of
time T seconds and it estimates a raw set of victim flows. SWAN adds some intelligence at the
congested sensor in order to choose the set of victim flows. For E.g., supposing source sensors
interpose RT-flows by tagging them as RT-old/RT-new employing the IP-TOS field.
Congested sensors could use a biased function to organize the set of victim flows out of novel
flows desiring to lessen false admission. Network-based regulation chooses precisely the victim
flows set at random and in the good instance it separates against newly acknowledged flows.
5. DESTINATION-BASED REGULATION
In the following subsections, we compare the projected destination-based regulation to other
destination-based components and appear the difference. We identify the proposed plan and
refine upon its preemptive and retrieval behaviors.
Common Destination-based Approaches
The idea of destination-based access is normally applied by a stack of QoS routing algorithms so
as to boost the classic per-source-destination QoS routing coarseness. The most specific
destination-based QoS routing algorithm is one during which QoS flows are tagged on a per-
destination granularity as a way for ease in a compromise to cut back the algorithm complexity at
average sensors. Altogether destination-based QoS routing algorithms, QoS flow rates which are
discovered on a per-destination granularity so as to mimic ATM solutions. The classic role of
destination-based approaches could be extended to function other targets like in which the
technique is applied to scale bandwidth brokerage and service provisioning. The destination-
based approach suggested in this research is different. Several QoS mechanisms are dependent
on the source to spot the per-flow QoS parameters that substitute the end-to-end resources, and
then transfer. Our proposition employs feedback data linked to the destination to aid congested,
intermediate sensors identifying QoS flows that are enduring from network mobility. Therefore, it
can be withdrawn to release the congestion and the destination-based approach provided at this
point is unique.

0%
10%
20%
30%
40%
50%
60%
70%
80%
0.30
0.25
0.20
0.15
0.10
0.05
0.00
Realtime Delay (sec)%Packets
Expired packets
Congested
packets
FIGURE 5: SWAN RT packet delay histogram.
Basic Definitions
The advised intensified SWAN along destination-based regulation developed from the slack of
SWAN design regarding model behavior in reaction to average traffic loads burden. The above-
average traffic load with at least one third of available usable bandwidth is ingested by real time
traffic and one third by best-effort traffic. The time-lag histogram presented in Figure 5 exemplifies
the majority of RT packets go through a time-lag of under 35 msec through the above-average
traffic load condition.
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
0.30
0.25
0.20
0.15
0.10
0.05
0.00
Realtime Delay (sec)
%CumulativePackets
About 10% of
delivered RT pkts
are not usable
FIGURE 6: Cumulative RT packet delay %.
Nevertheless, a substantial percentage of packets seem to go through delay higher than 175
msec. This is well defined on the cumulative graph in Figure 6. Where 9% of RT packets that
appear to go through delays over the 175 msec. For example, interactive VoIP flows will
disregard packets considering delays further than a stipulated threshold of 150 msec. By iterating
the identical test for various assorted mobility scenarios, SWAN model would systematically
causes about 9% of the of a typical RT packets to run out (9.04 %, 9.64 %, 9.97 %, 9.72 %, 10.03
%, and 9.87 %). So, the bandwidth gone through by extremely delayed packets is mostly a
bandwidth that inappropriately consumes network resources and degrades services rendered to
other early RT flows.
Destination-based regulation relies on destination sensors to find an additional gain in expired
bandwidth and then it modulates each flow consequently. This particular preemptive behavior
complies with preserving network resources. The destination-based regulation chooses a subset
of the congested flows with an intention to regulate in case of congestion. This subset is chosen
primarily based on flow stream quality beginning with the lowest quality flows for the initial time.
The grade of each flow is measured out as a function of the packet delays.
Maximum acceptable packet delay (MAPD) is the key threshold packet time-lag valuations/sec
required to guarantee an end in the destination behavior of disregarding packets of a finite flow.

MAPD is considered to be a flow-specific value that is referred by the destination sensor, which
also compares to the end-to-end packet time-lag.
Expired packets are usually the real time flow packets that, basically, display a time-lag over
MAPD. Identically, the expired bandwidth equals the bandwidth downed by expired packets.
Effective bandwidth = encountered bandwidth - expired bandwidth. Thus, effective bandwidth is
precisely the bandwidth recognized over at destination sensor. It can be used by destination
applications to rematch the real time cascade over a short time T.
Limited QoS renders a flow to, more or less, the postulated bandwidth. However, a substantial
portion of the obtained bandwidth is absolutely not functional as a result of excessive packet
delays. The effective bandwidth comprehended as of destination sensor on a period of time T is
scantily sufficient for the application to effectively replay the real time flow.
Effective bandwidth ratio (EBR ββββ) The percentage of average effective packet time-lags
obtained in real time by the destination sensor being finite real-time flow over a period of time T.
Due to the fact that the bandwidth ratio (β) lies on effective bandwidth and then , 0 ≤ β ≤ 1. Figure
7 along with equation 1 describe the EBR, and limited QoS definitions.
EBR β =
effectiveBW
receivedBW
overtimeT … (1)
EBR (β) measures out the grade of a RT flow rate during which the values closer to 1 would
suggest high flow quality and the values closer to 0 argue modified flow quality with ineffective
bandwidth use. RT flow-specific EBR rates (βH, βL) exemplify wanted measures of real time flow
rate quality, during which the values below βL stand for a junk in network resources that needs
regulation. The important investigation request message would communicate both βH and βL
values on to the destination sensor.
High
QoS
Limited
QoS
BE
No
QoS
1
2
3 4
5
1
0
EΒΒΒΒRββββ
6
ββββΗΗΗΗ
ββββL
FIGURE 7: (EBR β) Service view in a loaded intermediate sensor.
Effective delay ratio (EDR δδδδ) is known to be the percentage of intermediate, effective packet
time-lags at a desired destination sensor to MAPD over a period of time T. Simply because delay
ratio (δ) lies on effective bandwidth only, 0 ≤ δ ≤ 1.
EDRδ =
Avg.eff .pktDelay
MAPD
overtimeT … (2)
Effective delay ratio quantifies the sum of the flow. EDR measures closer to 0 suggesting
enhanced QoS. Whereas high EDR values closer to 1 would suggest a flow rate that is prone to
high time-lag averages, although the quality is yet suitable as depicted in Equation 2.
As mentioned in Figures 7 and Figure 8, a real time flow rate like flow 1 (symbolized by circle 1)
can be a flow that gets better than average bandwidth and consequently features high QoS.
Alternatively, flows 2, 3, and 4 are featuring confined QoS simply because they are obtaining the

postulated bandwidth. On the contrary, their effective bandwidth is barely at the desired limit.
Flow 5 demotes down to non-QoS and is unmarked through all intermediate sensors as a
discrete RT-flow any longer. Flow 1 holds a greater quality when compared to flow 6. The
remainder could be stated with the flow rates of and.
Preemptive Behavior
The destination-based algorithm is built over two behaviors, first, preemptive behavior that
screens QoS obtained by the network and upgrades services when provided service is
unsatisfying. The other is the recovery behavior, which is accomplished when intermediate
sensors encounter congestion by modulating limited QoS flows in advance of regulating higher
QoS flows.
The destination sensors modulate flows that are receiving limited QoS with short of noticing
congestion status. Essentially, when an adequate number of packets reach destination sensor
along with delays higher than MAPD over a time period T , the destination sensor finds a modified
QoS status (β < βL), and inserts a modulate message to the appropriate source sensor. The
source sensor then actuates a re-initiate process to identify a different route with more beneficial
quality.
High
QoS
Limited
QoS
BE
No
QoS
1
2
3 4
5
0
1
EDRδδδδ
6
δδδδΗΗΗΗ
FIGURE 8: (EDR δ) Service view in a loaded intermediate sensor.
Recovery Behavior
Recovery behavior is activated by destination sensor that enforces the following mechanisms:
a) Whenever an intermediate sensor is congested, it tags every individual RT packets along with
CE applying the ECN bits and packet markings would run right until the intermediate sensor
experiences an adequate drop in the incoming bandwidth.
b) Destination sensors experiencing (δ ≥ δ1) are going to release a modulate message instantly.
c) Early destination sensors are sure to wait for time period T. Supposing packets continue
getting marked with CE, after that destination sensors with (δ ≥ δ2) might issue setback a
modulate message instantly.
d) Traditionally, congestion gets settled by eliminating flows with higher comparative time-lags
and the values of δ is going to be constant for the network, and then possess to be chosen such
that (δi > δi+1).
This mechanism allows several destination sensors to modulate related real time flows on their
own by regulating low quality flows for start. When congestion is not decided flows featuring
slightly satisfactory quality are modulated till congestion is resolve.

6. EVALUATION AND ANALYSES
The primary rating perspective normally appears to compare the suggested destination-based
method using both source-based and network-based regulations. Although the source-based
regulation is simply an exceptional example of network-based regulation in which the subset of
dupe flows is the global set. Concerned reviewers would be able to review an evaluation between
source and network-based regulations. Our examination demonstrated the assertive and striking
conduct of source-based regulation which could be easily upgraded by adopting network-based
regulation as proposed by Ahn 02 or as adopting destination-based regulation as illustrated here.
Test-bed Description
Therefore, this Part equates destination-based regulation via a matured network-based
regulation. Ns-2 simulator is used to examine the destination-based approach. The test-bed
presumes a square field along with 20 sensors traversing with a maximum speed of 10 meter per
seconds with a random pause time of 2 sec. The test-bed functions parametric quantities out of a
Lucent WaveLAN card to organize a radio link applying 802.11 as MAC layer during which all
moving sensors holds a transmission range of 250 m. Sensor network On-demand Distance
Vector (AODV) and Dynamic Source Routing (DSR) protocols are the two routing protocols
arbitrary selected for evaluation. The traffic generator expends both constant and variable bit rate
RT-applications (CBR/VBR). Furthermore, we apply TCP links by simulating greedy FTP
applications for packet size 512 bytes. The TCP connections render (BE) packets that will not
likely require QoS services. Real time VoIP runs set up a MAPD of 150 msec and the burst video
flows takes 450 msec. Lastly, the count up simulation time is 300 secs, EBR (βH, βL) values are
(0.97, 0.95) and EDR( ) set values are {0.9, 0.8, …………, 0.2, 0.1}as a way for all the RT flows.
Bandwidth Efficiency
To be able to glance out the destination-based approach SWAN and ESWAN test-bed are
susceptible to the exact same traffic patterns and mobility scenario. The quantity of subjected and
presented RT bandwidth is supervised throughout the simulation time. The actual results are
displayed in Figure 9 and Figure 10 for SWAN and ESWAN. Hence, from the Figure 9 and Figure
10 it is straightforward to understand that the subjected bandwidths are quite identical in both
SWAN and the ESWAN test-bed. The presented bandwidth has a tight match to the subjected
bandwidth in ESWAN than the SWAN.
ESWANs destination-based approach mainly relies on the quality and also the usability of
traversed RT bandwidth. It is extremely important to seek the expired RT bandwidth with the EBR
ratio as specified earlier to be able to study the efficient bandwidth delivery on destination
sensors.
FIGURE 9: Efficiency of Bandwidth Usage in SWAN

Bandwidth Evaluation View
Figure 11 illustrates the distribution of run out RT bandwidth in SWAN and ESWAN. The chosen
traffic pattern sets off the ESWAN pre-emptive conduct at approximately 60 secs soon after
simulation begins it stimulates congestion on both SWAN and ESWAN models around 230secs
from simulation start time. Operating the depicted test-bed applying SWAN and ESWAN models
we are going to notice a substantial drop off in the period of expired RT bandwidth that improves
network resources usage.
FIGURE 10: Efficiency of Bandwidth Usage in ESWAN.
It is simple to realize that the fact of preemptive conduct stimulates a practical ceiling of a period
of expired RT bandwidth from a Figure 11. SWAN then again causes no enforce on the amount of
expired RT bandwidth alternatively it modulates flows merely soon after congestion takes place
by feeling the link. Shortly after 230 sec of simulation time congestion happens and drives the
regulation of RT flows in SWAN and also ESWAN simulations. The tip amount of expired RT
bandwidth in SWAN does not relate to congestion as illustrated in Figure 11.
Expired RT BW
0
2
4
6
8
10
12
14
16
30
60
90
120
150
180
210
240
270
300
Simulation Time in sec
ExpiredRTBW(kbps)
RT SWAN
RT eSWAN
Peak RT
pkt drop
Preemptive behavior
forced early regulation
Congestion caused
regulation
FIGURE 11: Distribution of Expired RT Bandwidth.
Figure 12 illustrates the statistical distribution of EBR for SWAN and ESWAN. The preemptive
doings of ESWAN observes a limited QoS status at approximately 60 sec of simulation time and
then ESWAN pushes a re-regulation of relevant RT flows. Hence, ESWAN presents modified
fluctuations of the EBR value. SWAN on the other hand features no restrictions on the EBR
values. So, the EBR value sinks to the small sum of 90 %. During 230 sec of simulation time
SWAN and ESWAN notice congestion, where as SWAN applies network-based recovery and
ESWAN employs destination-based recovery. This explanation would be clear from Figure 12.
Figure 12 shows how ESWAN goes back slower and stimulates fewer disruptions to RT flows in
comparison with the radical recovery of SWAN. It is crucial to understand that before the onset of

congestion SWAN simply didn't identify any trouble with RT flows and consequently RT flows was
not regulated. As an effect RT flows on SWAN went through less quality which network would
supply.
EBR (Effective BW Ratio) for RT Traffic
90%
92%
94%
96%
98%
100%
102%
30
60
90
120
150
180
210
240
270
300
EBRB
RT SWAN
RT eSWAN
Preemptive
behavior
Congestion
& recovery
behavior
FIGURE 12: EBR Distribution for SWAN and ESWAN.
Delay Evaluation View
Figure 13 depicts the preemptive conduct tripped on 60 sec of simulation time forces and also,
presents the minor delays in ESWAN compared to SWAN. Congestion encounters at 230 sec of
simulation time which is quite unique. Flows go through relatively less congested sensor than
SWAN in ESWAN and that is the impact of preemptive behavior. The retrieval behavior on
ESWAN is a trifle lower radical.
EDR (Effective Delay Ratio) for RT Traffic
0%
20%
40%
60%
80%
100%
30
60
90
120
150
180
210
240
270
300
EDRd
RT SWAN
RT eSWAN
Preemptive
behavior
Congestion
& recovery
behavior
FIGURE 13: EDR Distribution for SWAN and ESWAN.
Figure 14 illustrates the necessary instance on consequence of destination-based algorithm on
BE packet time-lags. The preemptive behavior is evoked around 60 sec of simulation time which
induces the BE intermediate delays to peak soon after. Typically, the delay fluctuation of BE
traffic is higher in ESWAN when compared to SWAN. Even though this issue might not create
functioning issues as BE traffic belongs to elastic applications that tolerate such fluctuations.
Avg BE packet Delay
0.0
0.5
1.0
1.5
2.0
30
60
90
120
150
180
210
240
270
300
AvgBEDelayinsec
BE SWAN
BE eSWAN
Slow recovery
of eSWAN
Preemptive
behavior
Delay peak
FIGURE 14: Distribution of Average BE Packet Delay.

Figures 15 and 16 depict the histogram with accumulative statistical distribution of RT packet
delay in ESWAN. These figures are quite comparable to Figures 5 and 6. In ESWAN, not as
much as 1.2 % of the rendered RT packets expired. The ESWAN model systematically turned up
only 1.2 % of RT packets to expire 1.13 %, 1.16 %, 1.17 %, 1.15 %, 1.16 %,and 1.14 % while test
result repetitions for several mobility scenarios.
0%
10%
20%
30%
40%
50%
60%
70%
80%
0.30
0.25
0.20
0.15
0.10
0.05
0.00
Realtime Delay (sec)
%Packets
Fewer
expired packets
Fewer
congested
packets
FIGURE 15: RT packet delay histogram in ESWAN.
Destination sensors apply this modified percentage to supervise the services coming from
network and force regulation when needed. When compared the figures with comparable results
from Part 5, ESWAN model holds decreased percentage of run out RT packets by 7.91 %, 8.48
%, 8.80 %, 8.57 %, 8.87 %, and 8.73 %. The Confidence Interval for the series can be estimated
by applying Equation (3). 0%10%20%30%40%50%60%70%80%90%100%
0.30
0.25
0.20
0.15
0.10
0.05
0.00
Cumulative%
About 1.2 % only of
delivered RT pkts
are un-usable
> MAPD
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
FIGURE 16: Cumulative RT packet delay % in ESWAN.
ConfidenceInterval= X ± t
(
a
2
)
σ
n
.………… (3)
Where: X = The mean difference between SWAN and
ESWAN observations
n = Number of samples (n=6).
σ = The standard deviation of the difference between
SWAN and ESWAN observations.
(1-α) = Confidence level (α=0.05).

t
(
a
2
)
=The upper critical value of the t distribution (=2.45).
The confidence interval for the percentage of decrease in expired RT packets when using
ESWAN compared to when using SWAN is calculated based on the 6 observation samples. The
95% confidence interval is [8.21%, 8.91%]. As this interval does not include 0, the performance
improvement by ESWAN is statistically significant, even with our somewhat limited sample size of
6. This percentage gain in RT packets has a significant impact on the delivered RT quality and on
the user perception.
The Effect of Mobility
Sensor mobility is an important factor in the design and evaluation of VANET based technologies.
The speed of mobile sensors and their pause time are commonly used attributes to define
mobility. The test-bed used pause time of 2 seconds, and when changing the pause time, both
SWAN and ESWAN showed little changes in behavior. When running the same test-bed with
sensor speed of {10, …………, 50} meter per second, both SWAN and ESWAN maintained the same
level of average packet loss as illustrated in Figure 17 up to sensor speed of about 35 meter per
second.
Effect of Mobility on RT -Packet Loss
100
200
300
400
500
600
700
800
10 15 20 25 30 35 40 45 50
Node Speed (meter/sec)
Avg#ofLostPackets/node
eSWAN
SWAN
FIGURE 17: The effect of sensor mobility on average packet loss.
When mobile sensors move faster than 35 meter per second, deterioration in radio link quality
takes effect. ESWAN shows a higher number of packet losses, and the losses grow much faster
compared to SWAN. The reason is the preemptive behavior in ESWAN, which responds to the
limited QoS perceived at destination sensors by forcing too many re-initiate probe requests
flooding the relevant routes and causing congestion, and packet loss. SWAN, on the other hand,
relies on re-routing, which is sufficient in high mobility scenarios.
Therefore, ESWAN is recommended in installations involving limited mobility (i.e. ≤ 35 m/s). We
believe this is not a major restriction since the threshold speed here is beyond vehicular speed
limits (i.e. ≤ 125 km/hr).
Overall Evaluation
In order to investigate the behavior of EBR (β), and EDR (δ) ratios, we apply some changes to the
test-bed.
An increasing traffic load is applied to a five mobile sensors test-bed, and the total consumed
bandwidth is measured then normalized over a period of time T sec. The mobile sensors are
forced to a no mobility condition, and the values of (β and δ) ratios of a VoIP flow are observed
against the increasing RT traffic load of the network. Figure 18 and Figure 19 represent the
results under these conditions. In both figures, the horizontal axes (average load per sensor)
represent the normalized collective bandwidth consumed by all five sensors for RT flows.
Therefore, the exact values of the network RT load will vary based on the test-bed topology, flow
directions, setup, and configurations; however, the shape of the curves will remain the same.

Change in EBR Under Increasing NW Load
92%
93%
94%
95%
96%
97%
98%
99%
100%
5,000
4,900
4,800
4,700
4,600
4,500
4,400
4,300
4,200
4,100
4,000
3,900
3,800
3,700
3,600
3,500
3,400
3,300
3,200
3,100
3,000
2,900
2,800
2,700
2,600
2,500
2,400
2,300
2,200
2,100
2,000
Network Load (bps/node)
SWAN
eSWAN
due to
preemptive
behavior
FIGURE 18: The effect of network load on EBR.
Figure 18 illustrates the impact of increasing overall network RT load on the EBR (β). Due to the
preemptive behavior, ESWAN tends to show higher EBR (β) values than classical SWAN. EBR
(β) values lower than 95% are regulated by ESWAN, and the re-initiation of RT flows provides
either higher EBR (β) value, or the RT flow will be denied service, and hence, have no EBR (β).
Change in EDR Under Increasing NW Load
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
5,000
4,800
4,600
4,400
4,200
4,000
3,800
3,600
3,400
3,200
3,000
2,800
2,600
2,400
2,200
2,000
1,800
1,600
1,400
1,200
1,000
800
600
400
200
0
Network Load (bps/node)
eSWAN
SWAN
due to
recovery
behavior
FIGURE 19: The effect of network load on EDR.
Figure 19 illustrates the impact of increasing overall network RT load on the EDR (δ). When
network RT load increases, the effective average packet delay increases, hence the EDR (δ).
There are relatively lower values for EDR (δ) on ESWAN than SWAN due to the recovery
behavior. High values of δ (> 70%) are commonly associated with congestion, while low values (<
10%) are associated with healthy RT flows.
7. COMMENTS AND CONCLUSION
The actual SWAN model presents only the source and network regulations algorithms as
solutions for regulating real time flows and brings out the two rules to deliver total congestion
recovery. SWAN results put on random array of victim flows and consequently offer little value to
the model. This research presents a novel destination-based regulation to boost the congestion
retrieval of real time flows instead of the source or network based regulations. The destination-
based regulation habituates a predetermined rule to choose victim flows and appends a
preemptive behavior to decrease the frequent occurrence of congestion. Packets tripping over
prominent sensors network may go through tenacious time-lags for the reason that travel over
more hops. Applying the MAPD threshold expends the EDR (δ), allows the network to restrain run
out bandwidth that liberates portion of the traffic load and finally increases bandwidth availability
and efficient use of RT bandwidth. This augmentation arrives at the expense of BE traffic that
realizes comparatively high pitched average delays, but has got only a minor influence on the BE
bandwidth.
The preemptive behavior is depicted to polish the resource utilization gradually and to diminish
probability of congestion. Furthermore, it allows destination sensors to supervise the exact level
of service and request a service ascent when the rendered service is unsatisfactory. The

recovery behavior of the destination-based approach renders quicker recovery when compared to
the network-based approach introduced by SWAN. A reliable recuperation is efficient in
interrupting a lesser amount of RT flows, and provides less average variation in RT delays.
Furthermore, ESWAN is proven to diminish the amount of expired RT packets by about 8.5 %,
which stands for a significant improvement compared to the original SWAN execution.
A number of the components specified in SWAN and in our research work are based on the
shaping rate (T seconds). The time period T is actually minor decent to interact with network
dynamics and prominent enough to average out the high traffic volumes generated by burst
traffic. SWAN and ESWAN test-beds operate the value of 2 seconds for T. Destination sensors
can comfortably allot a value for MAPD based on info from the application layer.
The MAPD values alter substantially with regards to the application, but is very substantive in
customizing the QoS demands for every flow. The destination-based method acquaints EBR (β),
and EDR (δ) as two significant parameters to measure real time flow quality. These parameters
want to be configured at the session start utilizing probe requests. Exploiter satisfaction is a vital
factor in delimiting acceptable thresholds for both these parameters. For example, streaming real
time flows may be able to tolerate larger jitter buffers than interactive real time flows; as a result,
we expect the values to be more stringent for interactive real time flows.
Consequently, further research needs to be performed on the ideal and acceptable values of both
parameters. An alternative crucial factor in measuring the SWAN model is investigating the
pragmatic approach for bandwidth utilization. SWAN follows a materialistic view of bandwidth
availability when admitting new real time flows, assuming that the tenacious slack of bandwidth
may be used by best-effort traffic. Hence, SWAN accomplishes comparatively higher resource
utilization in installations that has equivalent real time, best-effort volume of traffic and in
installations with limited variations in radio link quality. Supplementary research is required to
measure and tune SWAN for environments with skewed percentages of traffic types and highly
variable radio link quality. Likewise an evaluation with Shah Approach is very significant in
measuring the end-to-end performance.
Abbreviations and Acronyms
BE Best Effort
CBR Constant Bit Rate
DiffServ Differentiated
Service Model
DNST Downstream
(destination is a
vehicle in the V2V
network)
DSRC Dedicated Short-
Range
Communications
EF Expedited
Forwarding
ESWAN Enhanced SWAN
QoS model
for V2V networks
FTP File Transfer
Protocol
HC Hop Count
HTTP HyperText Transfer
Protocol
IEEE Institute of
Electrical and
Electronics
Engineering
INSIGNIA In-band Signaling
QoS model for V2V
networks
IP Internet Protocol
MPEG The Moving Picture
Experts Group
digital video format
MTM Mobile Truck Model
NS2 The Network
Simulator version 2
QoS Quality of Service

RT Real-time
SWAN Stateless Wireless
V2V Networks
TCP Transmission
Control Protocol
UPST Upstream
(source is a vehicle
in the V2V network)
V2V Vehicle-to-Vehicle
Communication
VBR Variable Bit Rate
8. REFERENCES
[1] Gahng-Seop Ahn, Andrew T. Campbell, Andras Veres, and Li-Hsiang Sun, “SWAN:
Service Differentiation in Stateless Wireless VANET Networks”, Proceeding of the IEEE
Infocom, June 2002.
[2] Gahng-Seop Ahn, Andrew T. Campbell, Andras Veres, and Li-Hsiang Sun, "Supporting
Service Differentiation for Real-Time and Best-Effort Traffic in Stateless Wireless VANET
Networks (SWAN)", IEEE Transactions on Mobile Computing, September 2002.
[3] Gahng-Seop Ahn, Andrew T. Campbell, Andras Veres, and Li-Hsiang Sun, "SWAN", IETF
draft, MANET workgroup, October 2002.
[4] Josh Broch, David B. Johnson, Yih-Chun Hu, and Jorjeta Jetcheva, "A Performance
Comparison of Multi-hop Wireless VANET Networking Routing Protocols", IEEE
Proceeding of the ACM/IEEE 4th. International Conference on Mobile Computing and
Networking (MobiCom' 98), Dallas TX, USA, October 1998.
[5] Shigang Chen, and Klara Nahrstedt, “Distributed Quality-of-Service Routing in VANET
Networks”, IEEE Journal on Selected Areas in Communication, vol. 17, no. 8, pp. 1488-
1505, August 1999.
[6] D. Chiu and R. Jain, “Analysis of the Increase and Decrease Algorithms for Congestion
Avoidance in Computer Networks”, Computer Networks, 1989.
[7] S. Corson, and J. Macker, “Mobile VANET Networks: Routing Protocol Performance Issues
and Evaluation Considerations”, IETF RFC 2501, January 1999.
[8] David B. Johnson, David A. Maltz, Yih-Chun Hu, and Jorjeta G. Jetcheva, "The Dynamic
Source Routing Protocol for Mobile VANET Networks (DSR)", IETF draft, MANET
workgroup, February 2002.
[9] Seoung-Bum Lee, Gahng-Seop Ahn, Xiaowei Zhang and Andrew T. Campbell, “INSIGNIA:
An IP-Based QoS framework for Mobile VANET Networks”, Journal of Parallel and
Distributed Computing, vol. 60, no. 4, pp 374-406, April 2000.
[10] Mohammad Mirhakkak, Nancy Schult, and Duncan Thomson “Dynamic QoS and Adoptive
Applications for variable Bandwidth Environment”, MITRE-DoD project paper, URL:
<http://www.mitre.org/support/papers/archive99_00.shtml>, April 2000.
[11] Y. Morgan, and T. Kunz, "PYLON: An Architectural Framework for VANET QoS
Interconnectivity with Access Domains", Proceedings of the 36th Hawaii International
Conference on System Sciences, Hawaii USA, January 2003.
[12] Charles E. Perkins and Elizabeth M. Royer, “VANET On Demand Distance Vector Routing”,
Proc. IEEE Workshop Mobile Computing Systems and Applications, Feb. 1999.
[13] Charles E. Perkins, Elizabeth M. Belding-Royer, and Samir R. Das, "VANET On-Demand
Distance Vector (AODV) Routing", IETF draft, MANET workgroup, June 2002.

[14] Y. Morgan and T. Kunz, "An Applied Mobile Truck Model (MTM) for on-the-Road Data-
Probing," in the International Journal for Heavy Vehicle Technologies, April 2011, vol. 18,
no. 2.
[15] Samarth H. Shah, Kai Chen, Klara Nahrstedt, "Dynamic Bandwidth Management for
Single-hop VANET Wireless Networks", ACM/Kluwer Mobile Networks and Applications
(MONET) Journal, Special Issue on Algorithmic Solutions for Wireless, Mobile, VANET ,
vol. 10, num. 1, 2009.
[16] Joao L. Sobrinho and A. S. Krishnakumar, “Quality-of-Service in VANET Carrier Sense
Multiple Access Networks”, IEEE Journal on Selected Areas in Communications, Vol. 17,
No. 8, pp. 1353-1368, August 1999.
[17] Andras Veres, Andrew T. Campbell, Michael Barry and Li-Hsiang Sun, “Supporting Service
Differentiation in Wireless Packet Networks Using Distributed Control”, IEEE Journal of
Selected Areas in Communications, Special Issue on Mobility and Resource Management
in Next-Generation Wireless Systems, Vol. 19, No. 10, pp. 2094-2104, October 2001.
[18] Hannan Xiao, Winston K.G. Seah, Anthony Lo, and Kee Chaing, “Flexible QoS Model for
Mobile VANET Networks”, IEEE Vehicular Technology Conference, VTC 20009-spring,
IEEE 51st, vol. 1, pp 445-449, Tokyo, May 2009.
[19] Yuan Xue, Baochun Li, Klara Nahrstedt, "Optimal Resource Allocation in Wireless VANET
Networks: A Price-based Approach", in the IEEE Transactions on Mobile Computing, 2010.
[20] Network Simulator version 2 <URL: http://www.isi.edu/nsnam/ns/>.
[21] Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications,
IEEE Standard 802.11, June 1999.
[22] H. Mantar, J. Hwang, I. Okumus, and S. Chapin "A Scalable Model for Interbandwidth
Broker Resource Reservation and Provisioning", in the IEEE Journal on Selected Areas in
Communications, vol. 22, no. 10, December 2004.
[23] T. Korkmaz and J. Guntaka "State-Path Decoupled QoS-based Routing Framework", in the
Proceedings of the IEEE Communication Society, Global Communication Conference,
Globocom 04, vol. 3, pp. 1515-1519, Dallas, TX, USA, December 2004.
[24] A. Riedl "A Hybrid Genetic Algorithm for Routing Optimization in IP Networks Utilizing
Bandwidth and Delay Metrics", in the IEEE Workshop on IP Operations and Management,
IPOM 02, pp. 166-170, Dallas, TX, USA, October 2002.
[25] Y. Chen, R. Hwang and Y. Lin "Multipath QoS Routing with Bandwidth Guarantee", in the
Proceedings of the IEEE Communication Society, Global Telecommunications Conference,
Globocom 01, vol. 4, pp. 2199-2203, San Antonio, TX, USA, November 2001.

Effective Streaming of Clustered Sensor Data in Harsh Environment

More Related Content

What's hot (19)

Similar to Effective Streaming of Clustered Sensor Data in Harsh Environment (20)

Recently uploaded (20)

Effective Streaming of Clustered Sensor Data in Harsh Environment