wireless mesh networks: first part: an overview 2-nd workshop 2-nd workshop on on women project...
TRANSCRIPT
Wireless Mesh Networks:Wireless Mesh Networks:
First part:First part:an Overviewan Overview
2-nd Workshop2-nd Workshop
on on
WOMEN ProjectWOMEN Project
Rome September 8-th, 2006
University of Rome “La Sapienza”, INFOCOM Dept. (Faculty of Engineering)
OutlineWireless Mesh Network: Definition and Characteristics
MAC layer solutions currently adopted
Wireless Mesh Networks: application scenarios
Slotted Seeded Channel Hopping (SSCH)
Dynamic Source Routing (DSR)
Conclusions and future researches
Network layer solutions currently adopted
I.F. Akyildiz, X. Wnag, W. Wang, “Wirless Maesh Netowrks: a survey” , Computer Netwroks I.F. Akyildiz, X. Wnag, W. Wang, “Wirless Maesh Netowrks: a survey” , Computer Netwroks No.47, pp. 445-487, 2005.No.47, pp. 445-487, 2005.
Wireless Mesh Networks: Definition
A Wireless Mesh Network is a multi-hop distributed mesh topology system, with self-configuration and self-organization capabilities, where each node is potentially able to forward Informative Units toward other nearby nodes
Wireless Mesh Networks: Characteristics (1/2)
1. Auto-configuration: all network nodes are designed to self-discover their neighbors and paths without needing of any centralized network entity
2. Auto-organization: nodes can autonomously resolve Out-of-Service events, due to temporary off or congested radio links, by exploiting the Mesh Topology
3. Scalability: the covered area can be extended by simply adding new nodes to the current Mesh Network
4. Mobility: the nodes can move on a limited area and keep the connectivity with (at least) a network node
5. Mesh Clients: mobile and peripheral nodes able to communicate with other nodes only through radio interfaces. Minimal routing functions are solved by them. Moreover, they are power constrained, typically low cost and developed on already existing Wireless Cards (e.g., 802.11a/b/g Network Interface Cards (NIC) )
6. Mesh Routers: nodes with minimum (or null) mobility, constituting the network backbone, with radio interfaces towards the mesh clients and mesh routers and wired interfaces towards the outside network. They are not power constrained, can process the most of network traffic and results more expensive than the mesh clients.
7. Additional features of the Wireless Mesh Networks: Currently there is no standard, and open questions are related to the security aspects and to proper MAC protocol developments
Wireless Mesh Networks: Characteristics (2/2)
This architecture is composed by mesh routers which are employed for the wireless backbone and mesh clients are excluded by the mesh topology
Connections among the mesh routers are realized with IEEE802.16 technology
Mesh routers function also as gateway for Internet access
Wireless Mesh Networks: architectures (1/3)(Infrastructure/backbone)
IEEE
802.16
Wireless Mesh Networks: architectures (2/3)(Client-Mesh)
This architecture is composed by self-configured Mesh Clients with routing functions
It represents the mesh network operating in ad-hoc mode
Currently wireless links are IEEE 802.11 based
IEEE 802.11
This architecture given by combing the two previous ones
Mesh Clients can access at the network through mesh routers as well as directly with other mesh clients
IEEE
802.16
IEEE 802.11
Wireless Mesh Networks: architectures (2/3)(Hybrid-Mesh)
Application Scenarios (1/2)
Low cost alternative to link difficult areas to be cabled
Broadband home networking Community Networking
Alternative to
IEEE 802.11 and Bluetooth standards
They can be view as a low cost solution of wide band access networks
Metropolitan Wireless Mesh Networks
Application Scenarios (2/2)
MAC layer solutions
Mac protocol for Wireless Mesh Networks has to consider several differences with those employed by the WLANs:
1) The multi-hop environment 2) All the architectures are distributed and each node is involved to the cooperation of
the network traffic management
Currently, the proposed MAC protocols are mainly based on two methods:
“Virtual” MAC protocols working on the top of existing MAC protocols
P.Bahl, R. Chandra, “SSCH: Slotted Seeded Channel Hopping for Capacity
Improvement in IEEE 802.11 Wireless ad-hoc networks”
Innovative MAC protocols with features similar to those proposed for ad-hoc wireless networks
J.W. Kim, N. Bambos, “Power Efficient MAC scheme using Channel Probing in Multirate Wireless Ad-hoc Networks“
Slotted Seeded Channel Hopping(SSCHSSCH)
SSCH is a protocol for ad-hoc wireless networks using IEEE 802.11 standard and exploits IEEE 802.11 MAC layer service
It can be simply implemented via software into a device equipped with a Wireless Network Interface Card (NIC) and IEEE 802.11 standard compliant
Its task is to extend the channelization of the IEEE 802.11a (13 channels), IEEE 802.11 b/g (11 channels) standard to the ad-hoc networks so to increase the throughput of each node
Each node is equipped with a Channel Scheduler for the channel/frequency hopping
Each node is equipped with a Packet Scheduler where the flow management is given by per-neighbor FIFO queues which are maintained in a priority queue ordered by perceived neighbor reachability
As the number of flows increases, SSCH considerably exceeds the IEEE 802.11a performances
SSCH Performances
Non-disjoint flows
Disjoint flows
Dynamic Source Routing (1/2)
Routing protocol for ad-hoc wireless networks with low mobility nodes
Differently to “Distance vector” or “Link State” based protocols, Dynamic Source Routing does not use periodic routing advertisement messages
It is based on Source Routing technique: before transmitting, each node evaluates the nodes’ sequence (hop) through which the packets are forwarded toward the destination node
Route Discovery
A control procedure is adopted for the correct packet reception and is based on data link acknowledgement between two adjacent nodes
Route Maintenance
DSR Performances
Routing length between a factor 1.01 and 1.09 from the optimal case
Overhead: ratio from 1.01 to 2.6 from the optimal case
ConclusionsConclusions
Wireless Mesh Networks are considered as a flexible, performing and low cost alternative to current WLANs
Currenlty, the proposed solutions are of proprietary type (MIT, Roofnet, Nokia, Mesh Connectivity Layer) and are essentially based on the IEEE 802.11 a/b/g standards
The MAC (SSCH) and routing (DSR) protocols currently adopted result to be extremely simple to be implemented
By the end of 2006 IEEE 802.11s Mesh standard is expected to be ratified
Wireless Mesh Networks:Wireless Mesh Networks:
Second Part:Second Part:PublicationsPublications
Title of Paper : “Optimized Power Allocation for Multi-Antenna Systems impaired by Multiple-
Access Interference and Imperfect Channel-Estimation”
Accepted on IEEE Tr. On Vehicular Technology
Authors: E.Baccarelli, M.Biagi, C.Pelizzoni, N.Cordeschi
WOmEn Project First Publication
Outline
• System Model (Wireless MIMO channel)• Mean Mutual Information• Power-Constrained Maximization of the Mean Mutual Information• System Nodes Interaction: the Game Theory Approach
• Spatial-Power Allocation Multi-Antenna (SPAM) Game for Ad -hoc
networks • SPAM game-vs.-collision-free Access strategies• Conclusions
System Model-(MIMO Wireless Ad-Hoc Network) (1/2)
Tx0 Rx0
Tx2
Tx1 Rx2
Rx1
System Model (2/2)
Spatial Covariance Matrix.
Channel Matrix.
K
H
d
Multiple Access Interference (MAI)
Tx0-Rx0 Reference link
• The overall observed signal vector during the payload phase
T
L tr
1( ) ( ) ( ), T T 1 Tn n n n
t y H d
†whereTra{ } tP, E{ (n) (n)}R R
• The information stream is power constrained as:
1[ ] d T
Tpay
t I H
������������������������������������������y
†E{ ( ) } payT Pt
Payload Phase (Tx0-Rx0 reference link)
User Information Throughput and Capacity
The choice of is finalized to reach the system capacity { ( ) }n
†
pay: { } T pay
1( ) sup I( ; | )
TE t P
C
H y H
}
pay
1( ) sup I( ; | )
TG GPttra{R
H y H
T
•
•
We adopt Gaussian distributed input signals for computing the following information throughput:
( ) ( )G CH HT
Under some conditions we have derived the Gaussian Throughput is equal to the Capacity
Mean Mutual Information
T *-1/2 -1/2 2 -1
pay r d d d
2
-1 *pay
tr d
1I( ; | ) T lg det + + P
t
Tlg det + ( )
t[( )]
([ )]
-
G
y H I K H H
R
KR K
I K
Such expression is valid under some conditions we have derived and reported into the Paper
Maximization of the User Information Throughput
Problem: evaluate and
(opt)Rˆ( )G HT
†( ) U *(1),..., *( ),0 UA t s Aopt diag P P s R
• It has been derived the Power Allocation Algorithm in order to find the optimal expressions of P*(1)….P*(s). It reduces to the Water Filling Approach when perfect Channel estimation is considered.
*-1/2 †
d d A A
t-sA 1 s
,
=diag{k ,...,k , } , s min{r,t}.
A A H K U A = U D V
D 0
•
2
*
1 1 1
1ˆ(H) lg 1 lg 1 ( ) lg 1 *( )T
r s r
G m lm m lm pay
PP m P m
T
Modelling of the Nodes Interaction (1/2) Game Theory Approach
F.R. Farrokhi, etc…
“Link-Optimal Space-Time Processing with Multiple Transmit and Receive Antennas”
IEEE Commmunications Letters, Vol.5 March 2001.
The Game Theory is adopted in order to consider the node interaction and the dynamic ad –hoc network topology
Modelling of the Nodes Interaction (2/2)Game Theory Approach
MIMO ad-hoc network may be modelled as Noncooperative Strategic Game
• - set of pair; (players set)
• - Action Set of node ;
• - Utility Function of node .
( ) ( ) *
{ : 0 [ ] }, 1,...,g g
g g gA Tra t P g n R R
I( ; |1
( ) ( ,..., ,..., ) )Gg g
pay
u uT
(1) (g) (n*) (g) (g)gy Ha R R R
����������������������������
x(g) x(g)T -R
xgT
xgT
{1,...,n*} N
There have been found Existence and Uniqueness Conditions for the Nash Equilibrum
1. Setup Phase (Eigenvalues and Nash Equilibrium )
2. While
Evaluate
3. Shape
4.
5. If go to 6
else go to 2
6. Evaluate the Throughput
Spatial Power Allocation for Multi-Antenna (SPAM) Systems
* budgetm
P m P
I
2 -1m
t1,...,s: 1+ K +d
m
Pm k tra
I
( )newR
( ) ( )new old Ψ R R
+
220.05 ( )
E EoldΨ R ( ) : ( )new oldR R
*( ), P m m I
SPAM Game-vs.-collision free access strategies
(Examples of Throughput Regions for an hexagonal network)
SNR=5dB, t=r=4.F.R. Farrokhi, etc…
“Link-Optimal Space-Time Processing with Multiple Transmit and Receive Antennas”
IEEE Commmunications Letters, Vol.5, March 2001.
1. The information throughput has been expressed in closed form for the general case of imperfect channel estimations and spatially colored MAI.
2. The power allocation and spatial shaping have been accomplished via the SPAM Game.
3. The SPAM game is fully distributed, asyncronous, scalable access schemes.
4. The SPAM game allows point-to-point throughput higher than those attainable via conventional orthogonal (e.g., collision-free) access schemes.
Conclusions
Title of Paper : “Interference Suppression in MIMO
Systems for ThroughputEnhancement and Error Reduction”
Proc. of IEEE International Wireless Communications and Mobile Computing Conference, 3-6 July 2006, Vancouver, pp.611-616.
Authors: E.Baccarelli, M.Biagi, C.Pelizzoni, N.Cordeschi
WOmEn Project Second Publication
Outline
• Reference MIMO Model• Transmission Rate and Error Rate in multi-user environment• Throughput enhancement and error reduction via interference cancellation• Performances• Impact on MAC and Routing of the pursued aproach•Conclusions and work in progress
Reference MIMO Model
• We consider a scenario where a mesh router receives information bits in the presence of multi-user interference from different mesh clients
Reference MIMO Model• The packet structure
Ttr (header) Tpay (payload)
T (packet)
.......
TL (learning)
Reference MIMO Model
• The received sequence, once acquired information about channel state (perfect CSI assumed) and interference is (under no CSI at the Tx)
0 ( ) ( )0 0
10
n
Us sl l
n nn n
E E
t t
Y Φ H Φ H N
V
Reference MIMO Model• The interference is “generally” spatially colored since it depends on topology, so this last heavily influences the behavior of the system
( )
1
n
Us l
n nn n
E
t
V Φ H N
Orthogonal STBCof n-th user
Possible spatial colorationInduced by the channel
spatially white noise
Transmission Rate and Error Rate in multi user environment
(1 )EP R T
Double goal:
High throughput with
Low BER
Requirement:High interference suppression capability since the transmission is simultaneous (collision)
• The final goal should be to transmit at high rate with very low bit error rates• By defining the “net throughput” also known as “gooput”
We have to maximize throughput and minimize error probability
Transmission Rate and Error Rate in multi user environment
• The throughput enhancement in the sense of “information rate” can be achieved by estimating interference at receiver side and by subtract it since interference reduces the capacity region
At the same time, the error reduction can be obtained by reducing the effect of interference that, generally, increases BER
1.00E-05
1.00E-04
1.00E-03
1.00E-02
1.00E-01
1.00E+00
0 1 2 3 4 5 6 7 8 9 10
SNRdB
BER MIMO interference free
MIMO interfered
Throughput enh. and error red. via interference cancellation
• The transmitter avoids to transmit for TL slots so the receiver can estimate the statistical feature of V (multi-user interference) and the linear estimator is given by
†1
0v v r h
V AY
A K K I RN
interferencenoise Reference signal
And channel
Throughput enh. and error red. via interference cancellation
• The performances in terms of estimation error variance can be evaluated in the following way
†12
0
†1
0
( )
2
v v v v r h
v v r h v
Tra Tra
Tra
K K K K I R
K K I R K
N
N
Throughput enh. and error red. via interference cancellation
• The error variance reduces itself to
for high noise and/or high level of reference signal
for low noise and low
level of reference signal
2 ( )vTra K
2 0
Throughput enhancement and error via interference cancellation
• The parameter that influences the performance is the SIR after cancellation
0 0( )
0 020 0
s scE E
SIR SIR
N N
Throughput enhancement and error via interference cancellation
• To increase network rate means
0
( )0 1
0
1 †0 0
1 † 10
,..., max log det
( ( ) )
2( ( ) ) ] ]
tots sn n
n U
n n
n
Utot
max rE E n
h r v v v v r h
v v r h v
H H I
R I K K K K N I R
K K N I R K
mT
N
Throughput enhancement and error via interference cancellation
• to reduce the error probability means to minimize
0
20
2
4 ( )
1 2
r
E qS
tP
E
Nq takes into account for the cardinality of modulation format
Performances
performances in terms of BEP for different q (modulation formats)
Performances
Net throughput for t=2, and different values of r
How interference Suppression can aid MAC?
• Approaches as CSMA usually tries to avoid collisions
i.s.
i.s.
The packet is dropped (SISO)
The packet is NOT dropped (MIMO)
Impact on MAC
• For sure the effect of interference suppression SIMPLIFIES the MAC procedures and the architecture• The MAC may operate in severe interference suppression conditions that means when interference is comparable with the main signal
Impact on Routing
• Preliminary results show that MIMO system allows power saving strategies (SISO same performance with low power emission) and interference suppression allows us to consider quasi-orthogonal transmission• This suggests that multi-hop approach is unnecessary in this operating conditions
Conclusions
• Interference suppression allow the system to e more simple at upper layers• Without decreasing transmission rate (due to multi-user interference) we are able to assure good performances in terms of BER• This does not require severe hardware complexity.
Fast Downloading of Large Files via Multi-Channel Wireless
Mesh Networks
Enzo Baccarelli, Mauro Biagi, Nicola Cordeschi Cristian Pelizzoni,
WOmEn Project Third Publication
First International Workshop on "Wireless mesh: moving towards applications" ( WiMeshNets 2006 ), (Co-located with QShine 2006, Waterloo, Ontario, Canada) August 10, 2006
Outline
• System Model and Problem Setup
• The constrained Minimization Problem, the fundamental trade-off: Time-minimization, Budget Constraints, and QoS client Requirements
• The possible strategies: Single-channel v.s. Multi-channel, Optimal and Sub-optimal approach: the On-Off policy
• Optimal energy-allocation policy and Throughput performance of the system: the general framework and two specific rate-functions of practical interest
• An application example. The Broadcast MIMO systems: the Dirty Paper strategy
• Conclusions and Works in Progress
• The last-hop of Wireless Mesh Networks,.
• Broadcast-type multi-channel Wireless Application scenarios: fading affected-links
• Energy-limited (battery powered) Wireless Mesh Router
• Multi-flows and Multimedia applications: the large (and increasing) size of multimedia objects.
• Clients demand for fast-downloading: Guaranteed QoS and maximum allowed Download-Time.
• Conventional current proxies inadeguate to solve the problem.
Problem Setup
Target:
Design of the Optimal Energy-allocation-Policy: ‘’How Much’’ and ‘’How Distribute’’
The Minimization of Download-Time of huge-size data.
System Model – Budget Constraints and QoS Requirements
Multi-channel system constituted by orthogonal sub-channels: a mesh-router serves clients requiring the download
1PM
Information Units (IU) to transfer to the clients
tot][Joule
max
mini
]/[ SlotJoule
P
1mm
tot
maximini
Overall available energy
Upper bound on the allowed peak-energy per slot
Minimum energy to be radiated over i-th sub-channel
][Joule
maxi iT TQoS Client requirements:
The Considered Family of Rate-Fuctions
( ) ( ( ); ( ))i i iIU t R t t ]/[ SlotIU
We assume: is in over non decreasing both for and strictly concave over non decreasing for
))();(( ttR 2C 00
0)0,(),0( RR),( R ),( R
ddRR ),(),( 0
( ) ( )( ) ε ( ); ( ); ( )r ri it t t t
budget state download state
channel state
How to choose?
1 2( ) ( ), ( )..., ( ) , for 1,2,...T
Mt t t t t
( ) ( )( ) ( ( ( ); ( )); )ir r
i i tt t t
)(ti)(tIU i
)(t)(tIU
iσ
totmax
1( ) [ ( ),.., ( )]TMt t t σ
Multi-Channel Optimal DesignSingle-Channel Optimal Design ( )i t
How much Energy
to radiate
How much Energy
and
How distribute it
The optimal allocation of
( ) ( ( ); ( )), 1,...,i i iIU t R t t i M The System Rate-Function:
The Throughput performance of the considered System
( ) ( )( )( ) ( ; ( ); ( ))r ri i ttt t σ
Related Works
• Single Client • No QoS constraints are accounted for• Single channel discrete state link• Linear rate-function
“Energy Allocation and Trasmission Scheduling in Satellite and Wireless Networks”, A. Fu, Phd Thesys, Massachusetts Institute of Technology, January 2003
Our Work: Multiple traffic-flows, multiple Clients QoS Requirements Continuous state multiple-orthogonal sub channels The rate-function: a general framework
, R 1 2, ,..., N I
0 1 2 3 4 5 6 7 8 9 10
0
0.2
0.4
0.6
0.8
1
1.2
Th
eu
The On-Off Policy
?The Optimal Policy
max
min
max ( )
. .
( ) ( )
) )
( )
( (T
i i
IU t
s t
t
t
U t
t
t K I
1
A
1
max
min
min ( , )
. .
( ) , 1
( ) ( ) , 1
( ) , 1
TOT
t
i
T
i i
E
s t
i K t
t t t
t t
1
T
The Optimization Problem and its Restatement
For , we have the following equivalent, simpler form:
( ) ( )( , , ) ; ( )r r A ε ε
(Low of Large Number)
The Optimal Energy-allocation Policy (1/3)
HL KKK
)(
)()(
)(
H
HL
L
10 0
max
0
( ) ( ; ) ,1 K
R
0
1
sup ( 0, )R
HK K 0max
1
( )M
ii
• •
Form of the Optimal Policy strongly depends on the Average available energy for the download of a single IU
•
totK
][ IUJoule
totmax
Total energy not sufficient to meet QoSLK Kmax
max
0
( , ) ( )R p d
“Maximal
energy” policy
Single channel Optimal Policy
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 20
1
2
3
4
5
6
7
8
9
0()(
Jou
le)
K=4K=3K=2K=1 )log();( 1R
)(
)()(
)(
H
HL
L
max
11
0
L0
The Logarithmic rate-function - 1
(Shannon capacity)
)()( 0N0eNp
(Rayleigh channel)
8max
1N0 34.4HK
0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.210
2
104
106
108
1010
1012
1014
K
Tem
po
me
dio
(s
lot)
Politica OttimaPolitica Euristica
0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.20
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
K
Th
rou
gh
pu
t (n
ats/
slo
t)
Politica OttimaPolitica Euristica
The Logarithmic rate-function - 2
On-Off policy only sub-optimal:very poor performance in strongly energy-limited application scenarios
2max191KH .
1N0
0 1 2 3 4 5 6 7 8 9 10
0
0.2
0.4
0.6
0.8
1
1.2
Th
eu
)(Th
Ave
rag
e d
ow
nlo
ad t
ime
(slo
t)
Optimal Policy
On-Off Policy
Optimal Policy
On-Off Policy
The Optimal Energy-allocation Policy (2/3)tot
max
)(ti)(tIU i
)(t)(tIU
iσ
min
0 1 1( ; ) ;
MAX
i
i i iR K
min
0 1( ; ) ( ; (( ) ))i
i i iRx x σσ
0max
1
| ( ) ( ) ( )M
ii
x
σ σ σ σB
Full Allocation Sub-Region
Not-Full Allocation Sub-Region
0max
1
1| ( ) ( )
M
ii
x K
σ σ σA
Case of the Logarithmic Rate-function
( , ) log(1 )i i i iR
00 )(:AA 0 σ max)(:B σ0
A
2max
2M 51K .max
50K . 80K . 21K .
};{)(};;{)(:B maxmax 00 02
01 σσ
1N0
†
;
;
pre - subtractor
y Hs n
H RQ
s = Q s
R
1)
2)
3)
An Application Example: the Broadcast MIMO System – (1/3)
)(ti)(tIU i
)(t)(tIU
iσ
Mesh Clients Mono-Antenna
Mesh Router Multi-Antenna
Dirty-Paper Strategy:
1) Channel Matrix QR-factorization,
2) Orthogonal Pre-coding,
3) Iterative Interference pre-subtraction
Interference-free
Space-Orthogonal Sub-Channels
,, 1, ,i i ii i
y R s n i M
( , ) log(1 )i i i iR
†
,
; ;
, 1, ,i i ii iy R s n i M
y Hs n H RQ s = Q s
An Application Example: the Broadcast MIMO System – (2/3)
42K .31K .
8max
2M
22K .max
1N0
0.8 0.85 0.9 0.95 1 1.05 1.1 1.15 1.2 1.25 1.3140
150
160
170
180
190
200
210
K
Tem
po
med
io (
Slo
t)
An Application Example: the Broadcast MIMO System – (3/3)
2max
4M 1K
8max
4M
191K .max
1N0
1N0
Ave
rag
e d
ow
nlo
ad t
ime
(slo
t)
Conclusions and Works in Progress
• The Multi-channel Optimal Energy-allocation Policy has been derived for the general framework we considered
• An efficient Algorithm for its computation has been derived, and comparisons with sub-optimal approaches has been carried out
• Performance evaluations in single-link systems and broadcast system application scenarios have been carried out
• How can the Considered Strategy take into account for real-time multimedia applications (not-elastic traffic)?
Thinking about….
WOMEN project:providing QoS with channel state dependent scheduling in WMNs
University of Rome La Sapienza
Presented by Tiziano Inzerilli8/9/2006
Contents
• Models and basic assumptions• Approach for QoS provision: traffic control & scheduler design• Statistics
Contents
• Models and basic assumptions• Approach for QoS provision: channel state dependent
scheduling• Statistics
MODELS & BASIC ASSUMPTIONS
Network model
MR
BS
MC
• MR (mobile routers) & BS (base stations) :– quasi static
nodes– Point-to-
multipoint transmission
– IEEE802.16/ IEEE802.11
• MC (mobile nodes)– Fast moving– Point-to-point
transmission– IEEE802.11
link
channel
Link model
Boff-1
Trans.node
Rec.Node 1
Brec-1
Rec.Node 2
Rec.Node N
Channel 1
Channel 2
Channel N
Boff-2
Boff-
N
Brec-2
Brec-
N
C1,, BER1
C2,, BER2
CN,, BERN
k
totk CtC )(
totkkeff CtCtC )()(
Bandwidth allocation
•
•
Varibility due to• channel impairments• MAC
Errors in differentChannels statistically independent
Channel model
Gilbert channelmodel
Traffic control portion
Boff Btrans
Transmittingnode
Receivingnode
Brec
sink sink
Bqueue-loss Bchannel-loss
Channel model: Gilbert Channel
• Every Tslot a transition occurs.
Good Channel
BadChannel
r
q1-r1-q
rq
qP
rq
rP BG
,
• Rayleigh fading channel: average fade duration (AFD) and average non-fade duration (ANFD) vs. the fade margin M and the Doppler spread fd.
• Transition Probabilities
1
2
1
M
d
ef
MAFD
df
MANFD
2
BPAFDANFD
AFDBLER
1
slotT
AFDr q r
P
PB
B
1
Channel model: types of channels
0 200 400 600 800 1000 1200 1400 1600 1800 20000
1
2
Idea
l cha
nnel
0 200 400 600 800 1000 1200 1400 1600 1800 20000
1
2
GE
cha
nnel
0 200 400 600 800 1000 1200 1400 1600 1800 20000
1
2
CG
E c
hann
el
No MAC contentionNo channel errors
No MAC contentionChannel errors
MAC contentionChannel errors
Channel State Dependent (CSD) & Independent (CSI) Scheduling Model
CSD Packet Scheduling
CSI Packet Scheduling
Error Correction: FEC, ARQ, interleaving
Flo
w 1
Flo
w 2
Flo
w 3
Flo
w 4
Flo
w 5
Flo
w N
Flo
w 1
Flo
w 2
Flo
w 3
Flo
w 4
Flo
w 5
Flo
w N
Cha
nnel
1
Cha
nnel
2
Cha
nnel
3
Cha
nnel
K
Wireless Link Wireless Link
WirelessLink
MonitoringWireless
LinkMonitoring
Contents
• Models and basic assumptions• Approach for QoS provision: channel state dependent
scheduling• Statistics
APPROACH FOR QOS PROVISION:traffic control & scheduler design
Metrics & Constraints
Delay Constraint (only real-time traffic)
maxjj DtD
Bandwidth Allocation Metrics
)(
totj
recj
totj
j Ct
tBC
tBAI
N
tBAI
tBAI jj
2
Data Loss Metrics
1)()(
)(
)(
)(
)(
)()()(
ttB
tB
tB
tB
tB
tBtt tot
k
offk
k
reck
k
transk
k
reck
k
offk
k
transk
channcong
Link Utilization Metric
1)()(
)(00
tot
k
transk
tot
k
reck
linkCtt
tB
Ctt
tBt
Design of traffic control
Regulation
Flow Classification
SchedulingLink
monitoring
C(t)
S(t)
Mean link Capacity overtime
Instantaneous channel state vector
Classification
Offeredtraffic
Real-timeNon-
Real-time
video1
video2
voice1
voice2
Web br.
File transf
With timeconstraints
Without timeconstraints
Other design options
• Assessment for channel estimation– SNR/SIR, Monitoring ACK reception, RTS/CTS frame exchange,
BER/BLER measured at destination, Ad-hoc probing frames, …
• Regulation strategies– Algorithms: Dual leaky buckets, markers, droppers, …– Strategies: never drop, drop after deadline, deadline only for
real-time traffic, …
• Scheduling strategies– Algorithms: Round robin, priority queuing, weighted fair
schedulers, deadline-based schedulers
• SELECTION CRITERIAS– Computational cost– Optimize metrics
Contents
• Models and basic assumptions• Approach for QoS provision: channel state dependent
scheduling• Statistics
STATISTICS
Simulation Scenario
BS
MC1
• Main features– Multiple real-time and non-
real-time flows to MC1, MC2, MC3
– Total capacity: 4Mbps– Link load: 99,1%
• Three subscenarios– Error-prone channel– Bandwidth variability per MC
due to MAC– Comparison with WFS
• Mapping of Flows:– MC1: video, voice– MC2: video, email, FTP– MC3: HTTP, email, FTP
MC2 MC3
Error-prone channel subscenario
10 20 30 40 50 60 70 80 900.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
1.1
Theoretical curveExperimental curve
% data loss
Link
Util
izat
ion
Effi
cien
cy
Variable bandwidth subscenario
0 10 20 30 40 50 60 70 80 90 1000.7
0.75
0.8
0.85
0.9
0.95
1
% of variable bandwidth
Link
Util
isat
ion
Effi
cien
cy
Comp. with WFS subscenario:average BAI statistics
0.00%
5.00%
10.00%
15.00%
20.00%
25.00%
30.00%
35.00%
40.00%
45.00%
50.00%
Correct DLB Setting Partially IncorrectDLB Setting
Totally IncorrectDLB Setting
Static DLB Alg. Dyn. Reg. with DLB Dyn. Reg. with LB WFS
Comp. with WFS subscenario:Link Utilization Efficiency
0,00%
10,00%
20,00%
30,00%
40,00%
50,00%
60,00%
70,00%
80,00%
90,00%
100,00%
Correct DLB Setting Partially IncorrectDLB Setting
Totally IncorrectDLB Setting
DLB Alg. Dyn. Reg. with DLB Dyn. Reg. with LB WFS
Future Ways
• To evaluate the potentials offered by MIMO-UWB at physical layer
• To consider the joint effect of beamforming and interference cancellation in order to aid MAC ad routing
• Performance Analysis of an innovative scheduling algorithm for OFDMA based IEEE 802.16a systems
• Performance analysis of an innovative algorithm of Connection Admission Control for IEEE 802.16 systems
• To take into account for real-time multimedia applications (not-elastic traffic)