Performance analysis of an innovative
algorithm of Connection Admission Control
for IEEE 802.16 systems
E. Baccarelli, M.Biagi, C.Pelizzoni, N.Cordeschi
This work has been partially supported by Italian National
project: Wireless8O2.16 Multi-antenna mEsh Networks
(WOMEN)under grant number 2005093248.
Outline
 Innovative Contributions
 Description of the IEEE 802.16 standard
 The proposed algorithm of Connection Admission Control
 Test model and simulation results
 Conclusions
Innovative Contributes
 Definition of a CAC strategy optimized for IEEE 802.16 systems
and supporting multi-service traffic
 Integration of the proposed CAC into a fixed radio access system
employing a Largest Weighted Delay First (LWDF) scheduler
 Performance analysis of the proposed CAC e comparisons with
the “Peak Rate Allocation” algorithm
IEEE 802.16 systems
 Frequency: 2-11 GHz
SS1
 Non line of Sight (OFDMA)
SS2
d1
d2
 Centralized access control
 Connection oriented MAC
BS
dn
d3
SS3
DOWNLINK
UPLINK
SSn
MAC Common Part Sublayer (CPS)
 Radio Access Control
 Traffic transport with variable length
Convergence Sublayer (CS)
MAC PDUs (fragmentation, packing)
 Mapping of the MAC SDUs onto the
IEEE 802.16 service classes
PHYsical layer
 Adaptive modulation and coding
Supporting of the QoS
UGS
rtPS
nrtPS
BE
Unsolicited Grant Service
real time Polling Service
non real time Polling Service
Best Effort
 Always on
Sources
 Peridically
guaranteed
bandwidth
E.g.: Voice over IP
 On/Off Sources
 Fixed bandwidth
and garanteed
on demand
E.g.: Video
Conferences
 Sorgenti On/Off
 Minimum
guaranteed
bandwidth on
demand
E.g.:TCP, Telnet
 On/Off Source
 Non Guaranteed
Bandwidth
E.g.: E-mail
Connection Admission Control
Criteria for admission
Tecnica di Controllo preventiva
• Parameter-Based
Admission
Control
Criteri
di
Ammissione
Obiettivi
generali: Control
• Measurement-Based
Admission
• Probe-Based Admission Control
 Massimizzare l’utilizzazione delle risorse
Traffic Descriptors
Test procedure
Garantire
un certo livello
di QoS
•Peak
rate
Admission
• Checking of the Current
Processo
Descrittore
Control
• Mean rate
dihold
Misura
Traffico
Una di
nuova
connessioneUnit
viene accettata
se e solo se:
• Burstiness
• Adopting of the temporal
 Sono
... disponibili risorse sufficienti ad allocarla
window
 E’ possibile garantire la QoS da essa richiesta
Decision
fordelle connessioni già
 E’ possibile mantenere
la QoS
Admission
presenti in rete
Connection Admission Control
Criteria
for admission
Control Technique
• Parameter-Based
Admission
Control
Criteri
di
Ammissione
Main Tasks:
• Measurement-Based
Admission Control
• Probe-Based Admission Control
 Maximizing of the resources utilization
Traffic Descriptors
Measure Procedure
Guaranteing
a fixed QoS
level
•Peak
rate
Admission
• Current hold
Processo
Descrittore
Control
• Mean rate
• Misura
Temporal
window
di Traffico
A nuova is accepeted
if and di
only
if:
Unit
• Burstiness
• Exponential mean
 There
... are sufficient resources to be allocated
 The required QoS can be guaranteed
Decision
for connections must be still
 The Qos of the already
accepted
Admission
guaranteed
The proposed algorithm of CAC (1/2)
UGS
p C

Aggregate
Peak
Rate Allocation
1
2
 Active
Fluid-Flow
Approximation
connections
NUGS
p new 
i
i 1
TOT
Traffic
CTOT
ln( 1 /  i )TON 1  i  pi  B2  4 B ln( 1 /  i )TON i 1  i  pi
S Allowed 2 ln( 1 /  i )TON 1   i 
 C    C
Connection
EB , EB g  CTOT
EBmin  min
Parameter-Based
Traffic  f 
N
ln( 1 /  i )TON 1   i  pi  B 
N
New
EBi 
connection
rtPS
pnew
pnew
TOT
EB( f )   EBi
Estimator
i 1
N
 Gaussian
Approximation
Refused Connection

m, C  
1
 1 
EB g   m   2 ln    ln  

 
 2 
pnew  m 
nrtPS Measurement-Based
( Hoeffding Bound)
ln( 1  )i 1 p 2 i
N
2
 CTOT
System Parameters of the proposed CAC

pi , TON


: peak rate [b / s ] , connection mean burst period
TON
TON  TOFF
: Connection duty cicle

B
: dimension of the assigned buffer

i
: maximum packet loss rate tollerance

EBi
: effective connection bandwidth [b / s ]

CTOT
: Effective downlink channel capacity

m  i 1 mi
N



C ( )

2

i
i 1
N
: Mean rate of the aggregate traffic
[b]
[b / s ]
[b / s ]
: standard deviation of the aggregate traffic [b / s ]
: equivalent capacity of the aggregate traffic [b / s ]
[s ]
The proposed CAC algorithm (2/2)
Requesting for Admission
Connection
type
UGS
PRA
PBAC
p new
no
no
pnew 
EBmin
NUGS
 pi  CTOT
i 1
CrtPS  CTOT  CUGS
no
no
nrtPS
rtPS
no
MBAC
C ( ), pnew
EBmin  CTOT  CUGS
no
pnew  C    CTOT  CUGS  CrtPS

m
CnrtPS  CTOT  CUGS  EBmin
Traffic
Estimator
CnrtPS  CTOT  CUGS  CrtPS
NUGS
CUGS  pnew   pi
i 1
C rtPS  EBmin
C nrtPS
Base Station DL Scheduler
Available bandwidth
It divides the available band into
different traffic classes by adopting
a strictly hierarchic algorithm
Residual bandwidth
DL Frame
 UGS: continuous granted
 rtPS, nrtPS: MLWDF
Residual bandwidth
j  arg max
 i Wi ( t ) EBi ( t )
i
~
ri ( t )
 BE: MLWDF
j  arg max
i
Wi ( t ) ri ( t )
~
ri ( t )
i 
 log( i )
Ti
Simulation tool
®
BS
Network level
Process level
Node level
CAC 802.16 vs CAC PRA
Task
Comparative analysis among the proposed CAC and the Peak Rate Allocation
Operating conditions




Dowlink transmission
QPSK Modulation  Maximum System Capacity 14.4 Mb/s
Maximum offered traffic 55.7 Mb/s
11 UGS sources, 8 rtPS sources, 30 nrtPS sources and 6 BE sources
Peformance Parameters
 System Throughput
 Number of admitted connections
Throughput
Mean Throughput
Addmitted traffic
Time (s)
Time (s)
 The admitted traffic of the CAC is 16 Mb/s > Channel capacity
 The attained gain is 18.5% higher than
 Mean throughput : 13% higher than
PRA
PRA
admitted rtPS
admitted UGS
Number of admitted connections
Time (s)
Time (s)
admitted nrtPS
802.16
N max
Multiplexing Gain: G 
PRA
N max
 UGS
G = 1.2
 rtpS
G=8
 nrtPS G = 1.1
Time (s)
Proposed optimization of the observation interval
Task
Optimal size of the observation interval
Operating Conditions





Downlink transmission
QPSK modulation  Maximum system capacity 14.4 Mb/s
Maximum offered traffic 41.5 Mb/s
nrtPS ( Ti=10 ms )
Temporal window of 20 frames (0.04 s) – 100 frames (0.2 s)
Performance Analysis
 Number of Admitted connections
 Transfer delay
Number of admitted connections & delays
Time (s)
Frame number
Time (s)
Optimal Length W = 40 frames
Performance analysis for heterogeneous traffic
Task
Checking for the QoS constraints
Operating conditions





Downlink transmission
QPSK modulation  Maximum System capacity 14.4 Mb/s
Maximum Traffic offered 69 Mb/s
11 UGS sources, 8 rtPS sources, 30 nrtPS sources e 6 BE sources
3
2
UGS (Ti=5 ms)
rtPS (  i  10 , Ti=10 ms)
nrtPS (   10 , Ti=10 ms)
Performance Analysis
 Mean Throughput per frame
 Transfer delay and packets loss ratio
i
Throughput
Mean Throughput
Throughput per frame
Time (s)
Time (s)
System Delays & Losses
rtPS
nrtPS
rtPS
nrtPS
Conclusions
The proposed connection admission control algorithm is able :
 to efficiently manage the available frequency band
 to fully meet the
losses
QoS requirements in terms of transfer delay and e packet
 to efficiently process heterogeneous traffic
 to experience a gain of 13% higher than a more conservative strategy such as the
Peak allocation Rate