Experimental Study of Hidden-Node Problem in IEEE802.11 Wireless Networks

Experimental Study of Hidden-node Problem in IEEE802.
11
*
Wireless Networks
Ping Chung Ng , Soung Chang Liew, Ka Chi Sha and Wai Ting To
Dept. of Information Engineering, The Chinese University of Hong Kong
{pcng3, soung}@ie.cuhk.edu.hk ; {kcsha2i, wtto2i}@ine.cuhk.edu.hk
ABSTRACT
The Hidden-node problem is a well-known phenomenon in IEEE
802.11 wireless networks. Most previous studies of the hiddennode problem were conducted by simulations, not experiments. In
this paper, we set up real-network experiments to investigate the
impact of hidden nodes in both infrastructure and multi-hop adhoc networks. Our experiments suggest transmission and carriersensing channel models different from those in the default settings
of the NS2 simulator. We also study the RTS/CTS mode as a
solution to the hidden-node problem. We show that using either
2Mbps or 11Mbps to transmit RTS/CTS not only can not solve
the problem, but also further degrades the throughput by
introducing additional overhead. In addition, our experiments
confirm the existence of re-routing instability and optimal
offered load in multi-hop ad-hoc networks which have only been
previously seen in simulations.
shows that the 11Mbps-data-rate throughput drops precipitously

after distance of 2.2m while that of 2Mbps-data-rate degrades
gradually until 10m.
The NS2 simulators default setting assumes 2Mbps has a hard
TXRange (i.e., 250m), below which all transmissions are
successful and above which all transmissions are unsuccessful.
The experimental results indicate that a more appropriate
transmission model will be one in which the probability of
successful transmission is a gradual function of distance.
In terms of throughput, we could devise a transmission channel
model which includes two operating ranges: 1) full-capacity
transmission range and 2) partial-capacity transmission range.
With 11Mbps data rate, 1) is around 2.2m while 2) is around 0.4m.
With 2Mbps data rate, 1) is around 3m while 2) is around 6m.
Keywords
Wireless Networks, Ad hoc Networks, IEEE 802.11, Experiments
1. EXPERIMENT SET-UP
Figure 2. Throughput versus distance y in 1-hop flow
3. CARRIER-SENSING CHANNEL MODEL

Figure 1. Two 1-hop flows and the real experimental network
We set up a real 6-node multi-hop network with six symmetric
DELL Latitude D505 laptop PCs with 1.5GHz Celeron Mobile
CPU and 512MB RAM. Each node has a Buffalo WLI2-CF-S11
IEEE 802.11b Wireless LAN card. All nodes run RedHat Linux 9
with HostAP driver. To facilitate experimentation, the
transmission power of each WLAN card is fixed to the small
value of -38dBm. This has the effect of scaling down the
transmission range. The basic and data rates are set at 11Mbps.
Unless otherwise indicated, the RTS/CTS mechanism is disabled.
The data sources are UDP traffic streams with fixed packet size of
1460bytes.
2. TRANSMISSION CHANNEL MODEL

In the first experiment, we investigate the transmission ranges
(TXRange) by varying the distance between node 1 and node 2,
for data rates of 2Mbps and 11Mbps. Link 2 is inactive. Figure 2
Postgraduate Student,
Undergraduate Student
* This work was sponsored by the Areas of Excellence scheme

established under the University Grant Committee of the Hong Kong
Special Administrative Region, China (Project Number AoE/E-01/99).
Copyright is held by the author/owner(s).
ACM SIGCOMM05, USA.
Figure 3. Throughput versus distance y in two 1-hop flows

The next experiment varies the distance between node 1 and node
3 in Fig. 1 with the distances of links 1 and 2 fixed at 2m. The
throughput results are shown in Fig. 3. The throughputs as
functions of the link separation depend on the carrier-sensing
operation. Note that the distance is small in the experiment
because we have scaled down the transmission power to facilitate
experimentation. Let y be the distance between nodes 1 and 3 and
x be the distance between nodes 2 and 3. The particular shapes of
the curves suggest a carrier-sensing model which includes four
regions: 1) full carrier-sensing region CSRange ( y 6m ) where
all nodes can hear each other, 2) partial carrier-sensing region
( 7m y 9m ) with y>CSRange but x<InRange (Interference
Range), 3) partial carrier-sensing region ( 10m y 12m ) with

y>CSRange and CSRange>x>InRange, and 4) decoupled region
( y > 12m ) where all nodes can not sense each other. In region 3),
the arrival orders of signals determine whether a signal can be
received successfully.
Figure 4 shows the simulation results of NS2 with default settings.
The nodes of each link are separated by 250m. From Fig. 4, four
regions can also be observed. However, all of them have
deterministic barriers. These hard regions do not exist in realnetwork experiments as shown by Fig. 3. Thus, we believe a
probabilistic model in which the probability of carrier-sensing is a
gradual function of distance will be a more accurate model. In
addition, the throughputs of link 1 in region 2) ( 560m y 690m )
severely degrade to zero which are different from experimental
results. Due to space limitation, we defer detailed explanations of
our transmission and carrier-sensing channel models to the final
poster.
a.
b.
Figure 6. Throughput versus time a) with and b) without
AODV routing protocol
6. OPTIMAL OFFERED LOAD
Figure 4. Simulation results of throughput versus distance y in

two 1-hop flows
4. RTS/CTS
a.
b.
Figure 5. Throughput versus distance y when using a) 2Mbps

and b) 11Mbps for transmitting RTS/CTS
Next, we turn on the RTS/CTS mechanism. Figures 5a and 5b
show that using either 2Mbps or 11Mbps to transmit RTS/CTS
can not prevent the hidden-terminal problem since RTS from
node 1 will collide with DATA from node 3 at node 2. In addition,
the RTS/CTS mechanism introduces additional overheads. The
throughput of link 1 degrades to less than 2Mbps, which is even
worse than that without RTS/CTS.
5. RE-ROUTING INSTABILITY
Reference [1] reported the re-routing instability problem in
wireless ad-hoc networks observed from simulation. Here we
investigate whether such instability occurs in real networks. In
this experiment, both links 1 and 2 use AODV as the routing
protocol. Figure 6a shows link 2 breaks down for 10 seconds from
50ths to 60ths due to the triggering of the re-routing function in
AODV. The associated throughput oscillations are not acceptable
for real-time applications such as video conferencing and voice
over IP. When AODV is turned off, as shown in Fig. 6b, link 1
has a relatively more stable throughput, although the hidden-node
phenomenon still exists.
Figure 7. End-to-end throughput versus offered load in a 6node flow

Reference [2] proposed to control the offered load at the source of
a multi-hop traffic flow to eliminate high packet-loss rate and
unfairness problems due to self interference. The existence of an
optimal offered load was derived by simulation and analysis.
However, it is not clear whether this exists in real multi-hop
networks. Here, we set up a 6-node multi-hop ad-hoc network and
our experiment confirms the results of [2]. As shown in Fig. 7, the
experimental optimal offered load is obtained at 1.25Mbps.
Beyond this value, the offered load is unsustainable high
packet-loss rate is incurred and the throughput obtained is lower
than the maximum possible throughput.
7. CONCLUSION
We have performed experiments to verify 1) the existence of
throughput degradation and re-routing instability due to the
hidden-node problem, 2) the ineffectiveness of RTS/CTS to solve
the problem, and 3) the existence of an optimal offered load in a
self-interfered multi-hop traffic flow. All these results have only
previously been demonstrated in simulations and analyses.
Besides confirmation of the above results, additional experiments
by us suggest transmission and carrier-sensing models that differ
from the default settings in the popular NS2 simulator. Due to
space limitation, we defer most details of our work to the final
poster.
8. REFERENCE
[1] P. C. Ng, S. C. Liew, Re-routing Instability in IEEE 802.11 Multi-hop
Ad-hoc Networks, IEEE WLN04, Nov. 2004, Tampa, USA.
[2] P. C. Ng, S. C. Liew, Offered Load Control in IEEE 802.11 Multi-hop
Ad-hoc Networks, IEEE MASS04, Nov. 2004, Florida, USA.

Experimental Study of Hidden-Node Problem in IEEE802.11 Wireless Networks

Uploaded by

Copyright:

Available Formats

Experimental Study of Hidden-Node Problem in IEEE802.11 Wireless Networks

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Experimental Study of Hidden-Node Problem in IEEE802.11 Wireless Networks

Uploaded by

Copyright:

Available Formats

Experimental Study of Hidden-node Problem in IEEE802.

shows that the 11Mbps-data-rate throughput drops precipitously

Figure 2. Throughput versus distance y in 1-hop flow

3. CARRIER-SENSING CHANNEL MODEL

2. TRANSMISSION CHANNEL MODEL

* This work was sponsored by the Areas of Excellence scheme

Figure 3. Throughput versus distance y in two 1-hop flows

Range), 3) partial carrier-sensing region ( 10m y 12m ) with

6. OPTIMAL OFFERED LOAD

Figure 4. Simulation results of throughput versus distance y in

Figure 5. Throughput versus distance y when using a) 2Mbps

Figure 7. End-to-end throughput versus offered load in a 6node flow

You might also like