Transcript Multiple Directional Antennas in Suburban Ad

Multiple Directional Antennas in
Suburban Ad-Hoc Networks
Ronald Pose
Muhammad Mahmudul Islam
Carlo Kopp
School of Computer Science & Software Engineering
Monash University
Melbourne, Australia
Outline
 Focus
of this paper
 Overview of SAHN
 Effects of omni-directional and
directional antennas on SAHN
 Conclusions
Focus of this Paper

Routing performance using three antenna schemes





multiple fixed directional
multiple omni-directional
single omni-directional
Estimation of achievable performance in a SAHN
Interference of non-SAHN nodes on a near by SAHN
SAHN (1/3)
How to connect to a corporate network from home
or how to link a community of broadband users
Commercial Wired Services
Direct Dial-up Services
Internet Services
Dial-up
Broadband (cable modems, xDSL, etc)
Ad-Hoc Wireless Networks
Single Hop Solutions
802.11b
Multi Hop Solutions
Nokia Roof-Top
SAHN
MIT Roofnet
SAHN (2/3)
Provides services not offered by commercial service
providers
Bypass expensive centrally owned broadband
infrastructure
Provide symmetric bandwidth
Independent of wired infrastructure
Avoid ongoing service charges for Telco independent
traffic
Features multi-hop QoS routing
Security throughout all layers
Utilizing link states (e.g. available bandwidth, link stability,
latency, jitter and security) to select suitable routes
Avoid selfish routing strategy to avoid congestion
Proper resource access control and management
SAHN (3/3)







Ideal for cooperative nodes. E.g. spread over a suburban area,
connecting houses, businesses, branch offices, etc
Topology is quasi static
Uses wireless technology
Symmetric broadband, multi Mbps bandwidth
No charges for SAHN traffic
Application
Application
SAHN services
Presentation
Presentation
run alongside
Session
S ession
A V O
U E T
TCP/UDP
TCP/UDP
TCP/IP
Transport
Transport
D D H
I
I
E
IP
Conceived in 1997 by
O O R
IP
Network
Network
SAHN Routing
Ronald Pose
Data Link
e.g. IEEE 802.11 variants
Data Link
Carlo Kopp
Physical
e.g. IEEE 802.11 variants
Physical
A Standard SAHN Node





Appears to host like a cable modem
Functionally more like a
RF LAN repeater
Embedded
microprocessor &
protocol engine
that implements all
SAHN protocols, manages
and configures the system
Each SAHN node has at least 2 wireless links
Capable of achieveing link rate throughput
Omni-directional Antennas
Advantages
 Directional orientation is not required
 May provide more connecting links
 Installation is easy and quick
 Ideal for ad-hoc networks with high mobility
Drawbacks
 Power radiates in all directions
 Increases hidden and exposed terminal problems
 Increases multiple access intereferences (MAI)
 Increases collisions and packet loss
 Degrades network performance
 Easy to eavesdrop
Directional Antennas
Advantages
 Power can be beam formed
 Reduces hidden and exposed terminal problems
 Reduces multiple access intereferences (MAI)
 Reduces collisions and packet loss
 Improves network performance
 Eavesdropping is limited to the direction of communication
 Ideal for ad-hoc networks with less mobility
Drawbacks
 Requires antenna direction alignment
 May provide fewer links
 Installation may be complicated
 Network planning is more difficult
Assumptions in this Work








Only the interference related effects on the routing protocol are
presented
Each of the antenna elements in multiple antenna schemes are
allocated distinct non-overlapping frequency channels
Multiple omni-directional antennas represent an omnidirectional antenna scheme that can operate simultaneously in
multiple non-overlapping frequency channels
GloMoSim (version 2.02) has been used for simulating various
layers and wireless media
The radio layer has been modified to use multiple directional
and omni-directional antennas
The effect of secondary lobes on the primary lobe has been
ignored while using directional antennas
A two-ray path loss scheme calculates the propagation path loss
DSR has been used as the routing protocol
Different Stages in Simulation
Simulations have been divided into the following stages:
1.
2.
3.
4.
Find maximum achievable throughput, delivery
ratio and response time in single and multiple
hops
Investigate the effect of different packet sizes on
network performance
Study the average network performance
Investigate the impact on network performance
of the presence of other networks operating
nearby
Simulation Setup in Stage 1 & 2




A chain of nodes and only one pair of nodes were active at
a time
Adjacent nodes have been separated by 350 metres
UDP packets have been used to avoid additional delays for
hand shaking and end-to-end acknowledgements in TCP
Loads at the sources have been changed from 10% to 85%
to get the critical point beyond which performance remains
unchanged
A node operating at 25% load means that it is generating
traffic at 2.75Mbps (maximum bit rate is 11Mbps)
Simulation Results for Stages 1 & 2 (a)
Maximum delivery ratio, throughput and effects of different
packet sizes in a single hop
Observations for Stages 1 & 2 (a)




Due to single hop scenarios the impact of directional and
omni-directional antenna was all the same
At around 55% load, the communicating link seems to
saturate
Above 55% load, the minimum time required to serve each
data frame becomes more than the time slot needed to
sustain the data rate
Smaller packets (e.g. 500 bytes) can reduce the peak
performance of the network by almost 50%
Since each data frame involves various delays (e.g. time for
RTS, CTS, DIFS, SIFS etc), smaller packets increase the
delay overhead per bit, hence reduce network efficiency
Simulation Results for Stages 1 & 2 (b)
Maximum delivery ratio (1500 bytes/packet) with multiple hops
Traffic load 25%
Traffic load 55%
Simulation Results for Stages 1 & 2 (c)
Maximum throughput (1500 bytes/packet) with multiple hops
Traffic load 25%
Traffic load 55%
Observations for Stages 1 & 2 (b, c)




Performance with single and multiple omni-directional
antennas reduced almost by 60% in most cases whereas
the performance achieved with multiple directional
antennas remained almost unchanged
Due the mechanism of DCF, some of the nodes along a
route have to wait while others are transmitting if omnidirectional antennas are used
As a result data transmission via multiple hops suffers
more back-off delays and collisions than single hop
communication
Directional antennas solved this problem with the sacrifice
of a range of directions
Simulation Results for Stages 1 & 2 (d)
Minimum response time (1500 bytes/packet)
Observations for Stages 1 & 2 (d)







Response time can be as small as 2.6 milliseconds for single
hop communication
With multiple directional antennas, a response time of 6.4
millisecond is possible at the 11th hop
At this distance, response times for a single and multiple
omni-directional antennas are 13.6 milliseconds and 8.4
milliseconds respectively
With the increase of number of hops, distance traveled by a
packet increases, hence more time is needed to get a reply
for a request
Moreover, time required for resolving interference can
make a response more delayed
The later problem is more common for omni-directional
antennas than directional ones
TDMA based schemes may exhibit better results
Simulation Setup in Stage 3










77 nodes were placed on a 3000x3000 sq km flat terrain where
each node had at most 6 neighbors
Separation between neighboring nodes was 350 metres
All antenna configurations had the same transmission range
Channel allocation to multiple antenna elements was random
On average each antenna channel connected 2 neighbors
Multiple directional antennas were allowed to communicate to
at most 3 neighbors at 3 different frequency channels which
effectively reduced the degree of connectivity per node
CBR and interactive applications generated random traffic
The number of nodes for interactive traffic was 6
To vary traffic, the number of nodes for CBR terminals were
increased in 5 steps (4, 8, 12, 16 and 20). For each
configuration, loads at CBR sources were varied at 4 different
levels (10%, 25%, 40% and 55%)
Each data packet was 1500 bytes long
Simulation Results for Stage 3 (a)
Average delivery ratio (1500 bytes/packet) with multiple hops
Traffic load 5.19%
Traffic load 20.78%
Simulation Results for Stage 3 (b)
Average throughput (1500 bytes/packet) with multiple hops
Traffic load 5.19%
Traffic load 20.78%
Simulation Results for Stage 3 (c)
Average response time (250 bytes/packet) with multiple hops
Traffic load 5.19%
Traffic load 20.78%
Observations for Stage 3




At low load and for the same number of hops, multiple
omni-directional and multiple directional antennas
perform similarly
As the loads at the CBR sources increase, their
performance start to differ significantly up to a certain
limit (i.e. for moderate traffic)
With the increase of the number of nodes and the rate of
traffic generation, fewer routes remain unsaturated to
balance the aggregated network load
As a result, a small performance gain can be achieved with
multiple directional antennas over multiple or single omnidirectional antennas
Simulation Setup in Stage 4


In omni-directional mode,
both networks use the same
frequency channel
In the multiple omnidirectional antenna scheme,
SAHN uses a frequency
channel different from the
neighboring network
Simulation Results for Stage 4
Effect on throughput (1500 bytes/packet) due to interference
from other networks
Observations for Stage 4




If the nearby node belongs to the same network, they are
supposed to co-operate, e.g. route others' packets
A node can decide not to allow packets coming from other
nodes belonging to a different network
A node cannot stop nodes of a different network from
transmitting. Instead, it can stop listening from that
direction to avoid interference
Two ways to do this



operating in different frequency channel or
using directional antennas
Directional and multiple omni-directional antenna
achieved similar performance (i.e. throughput was
constant) despite the increasing load at the nearby non
SAHN node whereas the omni-directional antenna suffered
from interference
Conclusions
 If no route exists in configured directions antennas may
need to be redirected and it may be difficult with multiple
fixed directional antennas
 Multiple fixed directional antennas may be expensive to
buy and install
 A smart directional antenna can be an alternative
solution at low cost
 We plan to optimize a routing protocol and the MAC
layer to efficiently handle the real life problems with
smart antennas in the context of SAHN
References
 R. Pose and C. Kopp. Bypassing the Home Computing
Bottleneck: The Suburban Area Network. 3rd Australasian
Comp. Architecture Conf. (ACAC). February, 1998. pp.87100.
 A. Bickerstaffe, E. Makalic and S. Garic. CS honours theses.
Monash University. www.csse.monash.edu.au/~rdp/SAN/ 2001
 MIT Roofnet. http://www.pdos.lcs.mit.edu/roofnet/
Thank You
?