SELECTIVE SUSPENSION OF TRANSMISSION FOR AVOIDING

PRIORITY REVERSAL IN MOBILE AD HOC NETWORKS

#R.Gunasekaran, Dr.V.Rhymend Uthariaraj

Department of Information Technology, Anna University, Chennai, India
#gunamit@annauniv.edu

ABSTRACT
Ad hoc wireless networks are a very potential field offering lot of scope for research.
In these networks, the Medium Access Control (MAC) protocols are responsible for
coordinating the access from active nodes. These protocols assume greater
significance since the wireless communication channel is inherently prone to such
problems as hidden terminal, exposed terminal and fading effects .The scheme
proposed here is used to perform priority scheduling in nodes resolving any
contention scenario that can arise for the channel in the best possible manner. Alert
transmission packets are used as a means of notification whenever a high priority
node wants to transmit data. Suspend transmission packets are used to avoid priority
reversal issue and a retry count is implemented to avoid starvation among the nodes.

Keywords: Ad Hoc Networks, Alert Transmission, Suspend Transmission,

Priority, Retry Count

1 INTRODUCTION A and C are hidden from each other. Now consider

another case where B is transmitting to A. Since C is
Contention for channel among the nodes is within B’s range it receives the transmission too and
resolved using Contention based protocols. In a can eventually defer its own transmission which is
heterogeneous network like ad-hoc several problems unnecessary as C’s transmission is in no way going
like hidden terminal and exposed terminal problem to affect A receiving the packets from B. This is
can arise. The popular Carrier Sense Multiple Access known as the exposed terminal problem i.e. C is
MAC scheme and its variations such as CSMA [1] exposed to B.
with Collision Detection (CSMA/CD) developed for
wired networks, cannot be used for wireless
networks. Priority scheduling is a means to avoid
channel contention among the various nodes in the
network. The scheme here proposes a new protocol
for effective priority scheduling. Two new packets
have been designed namely Alert Transmission and
Suspend Transmission packets which form the crux
of the new scheme. A retry count is implemented to
avoid priority starvation. Figure 1: Hidden Terminal Problem

The rest of this paper is organized as follows. Different flows in multi-hop networks have different
Section 2 presents the related work. The proposed degree of contention. Here, the contention degree for
Priority Scheduling scheme with a Suspend a flow is defined as the number of flows with which
Transmission mechanism is explained in Section 3. it is competing for the channel. Two types of MAC
Simulation results are given in section 4. schemes are prominently used. Reservation and
contention based schemes. Reservation based
2 RELATED WORK schemes usually make some assumptions about high
priority traffic. Flow scheduling is done locally
2.1 Classic CSMA problems while contention resolving probabilistically. Black-
In fig 1 Node B is within the range of A and C burst dealt in [8] is a typical example where a high
but nodes A and C are not visible to each other Let priority node transmits this black-burst signal as a
us consider the case where A is transmitting to B. notification for its transmission. Reference [9]
Node C, unaware of the transmission at B can generalizes this for wireless ad-hoc network. That is
transmit data to B thus causing collision at B. This is each station can sense the transmission of the other
referred to as the hidden-terminal problem, as nodes nodes in the network. Reference [6] explains a
dynamic priority scheduling with a CAN MAC low priority packets that hear either BT1 or BT2 will
protocol. defer their transmissions for some duration. In this
way, channel access priority of a high priority node
2.2 IEEE MAC 802.11 DCF can be ensured. Certainly, if there is no high priority
The 802.11 DCF function [5] is subjected to packet backlogged at a high priority node, a low
several research modifications, which is giving a priority node will not receive any busy tone.
back-off counter to each node such a way that every
node can choose a random number between 0 to 3 PROPOSED SCHEME
maximum contention window size. After sensing the
channel to be idle for an inter-frame space the nodes In this scheme priority scheduling in wireless
start counting their back-off counters to zero, and if ad-hoc networks, using alert transmission
the channel is found to be busy they freeze the back- mechanism is implemented. This way contention for
off counters. The value of Contention Window is channel access between nodes is resolved. This is
constrained to be between CWmin and CWmax. A also seen to eliminate the hidden terminal and
source station sends an RTS for which it receives exposed terminal problems occurring frequently in
back CTS following which it transmits data and gets ad-hoc networks. Individual nodes are assigned
an ACK packet back. In the event of CTS or ACK priority ‘Low and High’ based on the back off
not received the source is led to believe that collision counter value. It is computed using the formula
has occurred, so it is imperative that there is shown in Eq. (1)
adequate waiting time for the source before it arrives
at some decision. There are two waiting stages in Back Off = (1%cw)*priority*slot time …..(1)

where cw is the size of the contention window for

each node. And, priority is a user defined integer
value. For each node slot-time is 20µs, and CWmin <
CW < CWmax, where CWmin is the minimum CW,
and it is usually set to 32 and CWmax is the
maximum CW and often set to 1024.

Figure 2: Distributed Coordination Function (DCF)

ad hoc network, the inter frame space (IFS) stage

and the back off stage. The back off counter is a
random value between zero and the Contention
Window. For example high priority source stations
randomly choose the back off interval from [0, 2i+1
-1] and low priority source stations choose from
[2i+1, 2i+2 -1], where i is the number of consecutive
times a station attempts to send a packet. Two
different values of CWmin and CWmax are set for
different priority classes. It proposes an exponential
increase by a factor of 2 in the event of collision.

2.3 Existing Scheme

In order to effectively perform a priority
scheduling among these nodes in the network, a
priority scheduling scheme was proposed. Whenever
a high priority packet is backlogged at some high
priority node 0, it will send a primary busy tone
signal every M slots before it acquires the channel,
where M is a parameter of the proposed scheme.
When another node1 of lower priority hears this
primary busy tone signal (BT1), it will send a
secondary busy tone signal (BT2). All nodes with
Figure 3: Transmission of AT1 and AT2 packets Figure 5: Suspend Transmission Packet during
and Data between the sender, receiver and neighbors. Priority Reversal
After DIFS idle time, the station senses the Therefore in order to eliminate such a scenario,
medium to determine whether or not it is idle. If it is whenever a high priority node receives a Alert
idle, then the station decrements its back off value by transmission packet either directly or via indirect
a slot time, otherwise the back off value stays the means it can compare the initial back off value in the
same. When the back off value of a station reaches 0, Alert transmission packet to check if the source node
the station sends an alert transmission packet AT1 to is of higher priority or lower than its own. The high
its immediate neighbors, which again sends priority node will immediately send a SUSPEND
secondary AT2 packet to its neighbors and so on TRANSMISSION (ST) packet for suspending the
such that the hidden terminal problem is effectively transmission this will be directed at the source node.
overcome. The transmission of AT1 and AT2
packets is shown in Fig 3. All the nodes in the However not all high priority nodes can transmit
network are thus conveyed of the node’s intention to the ST packet. This transmission of ST is decided
transmit data. A typical Alert transmission packet based on the following criteria: original priority of
shown in Fig 4 will contain the following the node, priority threshold determined through
information: Sender Address, Initial Back Off average packet transmission time. Only if a high
counter of the original sender node, Receiver priority node satisfies these conditions it can transmit
Addresses, and The time of transmission. the ST packet. The ST packet contains the following
fields: Sender Address, Receiver Address, and Initial
The frame format of the Alert Transmission Packet Back off counter, Time sent.
(AT) is as follows:
The frame format of the Suspend Transmission
Packet (ST) is as follows:

Figure 4: Frame Format of Alert Transmission

Packet (AT1 or AT2)
Figure 6: Frame Format of Suspend Transmission
Duration represents the time of sending and TA Packet
is the sender address while RA is the receiver
address. Other lower priority nodes sensing the Duration represents the time of sending and TA
transmission immediately freeze their back off is the sender address while RA is the receiver
counters and defer their transmission to a later period. address. The right to send an ST packet for nodes
will not remain constant it can be subjected to
3.1 Priority Reversal changes based on network characteristics.
A priority reversal occurs when a low priority
node has its back off at zero when nodes at a higher An example scenario is depicted in Fig 7.
priority are in contention. This can lead to a situation Nodes 1, 4 are high priority nodes and nodes 3, 5 are
where the lower priority node grabbing the channel of low priority. At t1, the initial back off values of
before the Higher priority nodes. nodes 1, 3, 5 are 10, 17 and 18.

DIFS+
AT1 SIFS DATA
WAIT
Node 1
t1 (bc=10) t2(bc=0) DIFS+
SIFS
ST AT1 WAIT DATA
Node 4
t3 t4 t5(bc=9)t6(bc=2) t7 t9(bc=0)
AT1 DIFS +
Node 3 SIFS WAIT

t1 (bc=17) t2(bc=7) t5(bc=7) t6(bc=0) t8(bc=3)

Node 5
t1 (bc=18) t2(bc=8) t5(bc=8) t6(bc=1)

Figure 7: Scenario explaining the Priority Reversal

issue with 5 nodes. excessive starvation of a low priority node. This can
At t1, nodes 1, 3, 5 compete for the channel be fixed based on the number of nodes in the
access while 4 stays away from contention. Once the network and network characteristics. The number of
DIFS time expires, the back off time of node 1 times the back off counter is frozen is the retry count.
counts to zero and then, it sends an alert transmission The backpri value of the low priority node comes in
packet AT1 to all its neighbors. The nodes which handy whenever its RC value reaches a threshold.
receive AT1 send AT2 packet to its neighbors. After Once a node’s retry count reaches this threshold the
the SIFS period expires, nodes freeze their back off following occurs. The initial back off counter value
counters. Nodes 3, 5 have their back off counters is replaced now with backpri value and a slot time is
frozen at 7, 8 respectively and their retry counts are added to it. But before overwriting the initial back
increased by 1. off value it is imperative that a copy of it is stored as
backup in initial backup variable defined. Now the
Node 1 after sending AT1 waits for a backpri value is used to overwrite the new back off
DIFS+SIFS period and then takes control of the value which denotes the current or active back off
channel for data transmission. At t4, nodes 4 (BC value of the low priority node. That is once RC
=9), node 3, 5 contend for the channel access with 3 threshold is reached do the following
beating 4 leading to a priority reversal. To overcome
this, once the AT1 packet of 3 reaches node 4, it Initial backup=Initial back off (3)
realizes that it has high priority than node 3. Hence,
it disregards the AT1 packet and transmits a Initial back off=backpri+ slot time (4)
SUSPEND TRANSMISSION packet ST to node 3
for suspending the transmission. The ST packet New back off= backpri (5)
contains the current back off value, curr of the
sender node i.e., node 4 in our case. Once node 3 The backpri value denotes the lowest current
receives the ST packet from node 4, it resets it back back off value of the high priority nodes that have
off value according to the formula. beaten the current node to access the channel. This
means the low priority node is promoted to a high
New Back off = curr + slot time (2) priority status temporarily, this is only fair because it
has starved so long a period defined by RC threshold
Where, curr is the current back off value of the to transmit the current packets, and it is necessary
node sending ST. After sending ST, node 4 waits for that some means are done to promote its priority
an SIFS period and then starts counting its back off status, to minimize the further backlogging of these
timer to zero. Once a node receives ST it is packets. The low priority node will now enter
necessary to do the following apart from resetting its contention as a high priority node since it has its
back off value. Firstly, there is a variable backpri initial back off value reset. Now the initial back off
initialized with the value of initial back off counter value will remain as backpri + slot-time only till the
of the node. From the ST packet received, the current node transmits the current packets backlogged. The
back off value of the high priority sender is obtained. RC value is reset to zero as shown in Eq. (6) and
This value is compared with the existing value in initial back off value is set to the initial backup in Eq.
backpri and the smaller value is stored in backpri. (7) after current packets are transmitted.
Simultaneously, the back off counter is frozen.
RC=0 (6)
If current back off time of sender < backpri, then
Overwrite backpri as follows Initial back off= initial backup (7)
Backpri=current back off time of sender
Else This means after the nodes are transmitted the node’s
Do Nothing priority is reverted back to its original status, which
is only agreeable as it cannot be promoted all the
Similarly, when this low priority node gets into time. This scheme would thus be helpful in avoiding
contention in the next idle phase and if it loses starvation of low priority nodes for the channel
contention again by receiving an ST from a high access.
priority node, it will compare the backpri value with
the current back off value of the sender and store the 4. SIMULATION
smaller value into backpri. Thus, the priority reversal
issue is dealt with using Suspend Transmission The proposed scheme is implemented with the
packets. help of ns2 and the results of the implementation
analysis are illustrated in the following graphical
3.2 Starvation Avoidance representations. The Random Way Point model [10]
A Retry Count (RC) is used to prevent the is used in ns2 simulation. Figure 8 indicates the
comparisons in aggregate throughput between the observed that with the increase in the number of
proposed Alert Transmission schemes to the existing nodes in the network, the throughput increases.
IEEE MAC 802.11 scheme. The number of nodes is
used as the measuring criteria. The simulation is 5. CONCLUSION
carried out with 20, 40, 60, 80, 100 nodes. The
results show that the proposed scheme produces A new priority scheduling scheme (Alert
better average throughput when the number of nodes Transmission Scheme) is proposed for ad hoc
increases. networks. With the use of AT1, AT2 the Alert
Transmission Scheme ensures the channel access of
high priority data packets. Priority reversal is also
IEEE MAC 802.11
AT-ST SCHEME avoided by the use of Suspend Transmission ST
3000
packets. To avoid the starvation of lower priority
2500
packets and to ensure a fair scheduling, retry count is
Aggregate Throughput

used.
2000

1500 5 NODES
10 NODES
15 NODES
1000
120

500 100

THRO UG HP UT
0 80
0 20 40 60 80 100
60
Number of nodes
40

Figure 8: Comparisons in aggregate throughput 20

between proposed Alert Transmission scheme and 0
IEEE MAC 802.11 scheme. 0 0.2 0.4 0.6 0.8 1 1.2
ARRIVAL RATE
Figure 9 shows the comparison results in the
delivery ratio of high priority packets between the Figure 10: Throughput as a function of delay in the
proposed scheme and the IEEE MAC 802.11 scheme. arrival of packets with different number of
The results show that the high priority packets are nodes in network
delivered at a much better rate in the proposed
scheme. The average throughput is compared with the
number of nodes in the network. The delivery ratio
IEEE MAC 802.11 of high priority packets is also observed to be better
AT-ST SCHEME in the proposed scheme, further the throughput is
Delivery Ratio of High Priority Packets

1.2 illustrated as a function of delay in arrival rate of

1
packets for varying number of nodes in the networks.
The simulation results ascertain that the overall
0.8 average throughput, delivery of packets in the
0.6
network implementing the proposed scheme is better
than the IEEE MAC 802.11 scheme.
0.4

0.2
REFERENCES
0
0 20 40 60 80 100
[1] Andrew Muir, J.J. Garcia Luna Aceves, “An
Number of Nodes
Efficient Packet Sensing MAC protocol for Wireless
Networks”, 1998
[2] Chunhung Richard Lin and Mario Gerla, “Real-
Figure 9: Comparison of delivery Ratio of High time support in multihop wireless networks. Wireless
Priority Packets between the proposed Alert Networks”, 1999
Transmission scheme and the IEEE MAC 802.11 [3] Wei Liu, Yuguang Fang, “Courtesy
Piggybacking, “Supporting Differentiated Services
Figure 10 shows the throughput as a function of in Multi-Hop Mobile Ad Hoc Networks”, April 2004
delay in the arrival rate of packets, the number of [4] Xue Yang, Nitin H. Vaidya, “Priority Scheduling
nodes used here are 20, 40 and 60, 80, 100 and it is in Ad Hoc Networks“, July 2006 (ACM)
[5] Sunil Kumar, Vineet Raghavan, Jing Deng, [9] J.L.Sobrinho and A.S.Krishnakumar. Quality-of-
“Medium Access control Protocols for ad hoc Service in Ad Hoc Carrier Sense Multiple Access
wireless networks” A survey”, June 2004 Wireless Networks. IEEE Journal on Selected Areas
[6] Oleko Odongo, “Dynamic Priority Scheduling in Communications, 17(8), August 1999.
with Can MAC protocol”, December 2006 [10] Tracy Camp, Jeff Boleng and Vanessa Davies.”
[7] Yu Wang, Brahim Bensaou, “Priority Based A Survey of Mobility Models Ad Hoc Network
Multiple Access for Service Differentiation in Research”, 10 September 2002
MANETS”, July 2005
[8] J.L.Sobrinho and A.S.Krishnakumar, “Real-time
traffic over the IEEE 802.11 medium access control
layer, Bell Labs Technical Journal, pages 172187,
Autumn 1996.