Study of the Capacity of multihop Cellular networks - Semantic Scholar
Study of the Capacity of multihop Cellular networks*
Antonis Panagakis, Elias Balafoutis, Ioannis Stavrakakis Dept. of Informatics and Telecommunications, University of Athens, Athens, Greece {grad0260, balaf, istavrak}@di.uoa.gr *This work has been supported in part by the IST Programme under contract IST-2001-32686 (BROADWAY).
NATIONAL & KAPODISTRIAN UNIVERSITY OF ATHENS
1
Outline
• Cellular Network Multihop Cellular Network (MCN) • Evaluation of MCN throughput via simulations • Discussion and alternative architectures
NATIONAL & KAPODISTRIAN UNIVERSITY OF ATHENS
2
Cellular Network
• AP has a “distinguished” role NATIONAL & KAPODISTRIAN UNIVERSITY OF ATHENS
3
Enhancing (??) the Capacity of a Cellular Network (via the Multi-hop Cellular Network -MCN) Allow for AÆB communication without engaging the AP Reduce the transmission range to allow for spatial reuse and potentially improve the capacity
(“side-effects”: + decreased power consumption / - connectivity, QoS)
AP is “downgraded” to (almost) an ordinary node NATIONAL & KAPODISTRIAN UNIVERSITY OF ATHENS
4
Common Belief: Spatial Reuse Î Higher Capacity
True only if transmissions are spread geographically (traffic is almost entirely intra-cell) NATIONAL & KAPODISTRIAN UNIVERSITY OF ATHENS
5
Realistic WLAN traffic is Mostly Inter-cell
¾ ¾
AP throughput cannot exceed channel capacity Interference will result in a throughput well below it!
NATIONAL & KAPODISTRIAN UNIVERSITY OF ATHENS
6
“Common” Sense:
Spatial Reuse provides no Throughput Advantage under Inter-cell traffic AP is the gateway to the world for a MCN Inter-cell traffic will dominate
NATIONAL & KAPODISTRIAN UNIVERSITY OF ATHENS
7
Evaluation of MCN (simulation environment) • • • • • • •
ns simulator (version 2.1b7) MAC : IEEE 802.11 DCF mode Propagation model: two ray path loss model Routing: DSR Transport protocol: UDP packet size: 512 bytes default parameters of the simulator (WaveLan Card) (except transmission range, range of interference)
NATIONAL & KAPODISTRIAN UNIVERSITY OF ATHENS
8
Evaluation of MCN (simulated topology) • 11x11 (121-node) grid – AP : the node at the centre – side of the grid = 40m – transmission range = 60m
• Limitations: – uplink – inter-cell traffic (directed to the AP) – Static nodes (no mobility)
NATIONAL & KAPODISTRIAN UNIVERSITY OF ATHENS
9
Throughput Efficiency of MCN (channel capacity 1300Kbps - a=2.6)
Thhroughput (Kbps)
Interference factor a: ratio of the interference range over the transmission range (for example, for default values of simulator: a ≈ 550/250 = 2.2) 500 450 400 350 300 250 200 150 100 50 0
a = 2.6
0
1000
2000
3000
4000
5000
6000
Total Load (Kbps)
Throughput < 250Kbps for Load < 1300Kbps !!! (paradoxes discussed later) NATIONAL & KAPODISTRIAN UNIVERSITY OF ATHENS
10
Throughput Efficiency of MCN (channel capacity 1300Kbps)
a=1
"
'
%
(
)
*
+
'
,
-
Reduce interference to the min : a=1 No significant improvement
!
"
"
#
$
%
&
a = 2.6
MCN architecture limits throughput not the interference level NATIONAL & KAPODISTRIAN UNIVERSITY OF ATHENS
11
Explaining the Paradox
Load
Antonis Panagakis, Elias Balafoutis, Ioannis Stavrakakis Dept. of Informatics and Telecommunications, University of Athens, Athens, Greece {grad0260, balaf, istavrak}@di.uoa.gr *This work has been supported in part by the IST Programme under contract IST-2001-32686 (BROADWAY).
NATIONAL & KAPODISTRIAN UNIVERSITY OF ATHENS
1
Outline
• Cellular Network Multihop Cellular Network (MCN) • Evaluation of MCN throughput via simulations • Discussion and alternative architectures
NATIONAL & KAPODISTRIAN UNIVERSITY OF ATHENS
2
Cellular Network
• AP has a “distinguished” role NATIONAL & KAPODISTRIAN UNIVERSITY OF ATHENS
3
Enhancing (??) the Capacity of a Cellular Network (via the Multi-hop Cellular Network -MCN) Allow for AÆB communication without engaging the AP Reduce the transmission range to allow for spatial reuse and potentially improve the capacity
(“side-effects”: + decreased power consumption / - connectivity, QoS)
AP is “downgraded” to (almost) an ordinary node NATIONAL & KAPODISTRIAN UNIVERSITY OF ATHENS
4
Common Belief: Spatial Reuse Î Higher Capacity
True only if transmissions are spread geographically (traffic is almost entirely intra-cell) NATIONAL & KAPODISTRIAN UNIVERSITY OF ATHENS
5
Realistic WLAN traffic is Mostly Inter-cell
¾ ¾
AP throughput cannot exceed channel capacity Interference will result in a throughput well below it!
NATIONAL & KAPODISTRIAN UNIVERSITY OF ATHENS
6
“Common” Sense:
Spatial Reuse provides no Throughput Advantage under Inter-cell traffic AP is the gateway to the world for a MCN Inter-cell traffic will dominate
NATIONAL & KAPODISTRIAN UNIVERSITY OF ATHENS
7
Evaluation of MCN (simulation environment) • • • • • • •
ns simulator (version 2.1b7) MAC : IEEE 802.11 DCF mode Propagation model: two ray path loss model Routing: DSR Transport protocol: UDP packet size: 512 bytes default parameters of the simulator (WaveLan Card) (except transmission range, range of interference)
NATIONAL & KAPODISTRIAN UNIVERSITY OF ATHENS
8
Evaluation of MCN (simulated topology) • 11x11 (121-node) grid – AP : the node at the centre – side of the grid = 40m – transmission range = 60m
• Limitations: – uplink – inter-cell traffic (directed to the AP) – Static nodes (no mobility)
NATIONAL & KAPODISTRIAN UNIVERSITY OF ATHENS
9
Throughput Efficiency of MCN (channel capacity 1300Kbps - a=2.6)
Thhroughput (Kbps)
Interference factor a: ratio of the interference range over the transmission range (for example, for default values of simulator: a ≈ 550/250 = 2.2) 500 450 400 350 300 250 200 150 100 50 0
a = 2.6
0
1000
2000
3000
4000
5000
6000
Total Load (Kbps)
Throughput < 250Kbps for Load < 1300Kbps !!! (paradoxes discussed later) NATIONAL & KAPODISTRIAN UNIVERSITY OF ATHENS
10
Throughput Efficiency of MCN (channel capacity 1300Kbps)
a=1
"
'
%
(
)
*
+
'
,
-
Reduce interference to the min : a=1 No significant improvement
!
"
"
#
$
%
&
a = 2.6
MCN architecture limits throughput not the interference level NATIONAL & KAPODISTRIAN UNIVERSITY OF ATHENS
11
Explaining the Paradox
Load
