2010lnternationalCoperenceOnComputerDesignndppliations(lCCDA2010)

APcwLallAdmssun LunlrulAlgurlhmlur11112.1b Pclwurks

Samaneh Rashidi.B
Department of Computer
Engineering
Azad university, branch of
Arak, Iran
Samane.rashidi@gmail.com
Hasan Taheri
Department of Electrical
Engineering
Amirkabir university
Tehran, Iran
htaheri@aut.ac.ir
Abstract- In this paper, we present a new Call Admission
Control (CAC) for IEEE bZ.1 broadband wireless networks
considering Quality of Service (QoS) constraint. With the
assumption that there is a scheduling algorithm to allocate total
bandwidths, our proposed CAC algorithm accepts new request
for some slot times temporary. In this algorithm, during
certain time slots, departure rate for ongoing connections will
be decreased and for new connection will increase. Afer this
period, if QoS for ongoing connections and new request is
satisfed, new request will be accepted permanently and
otherwise it will be rejected. We study the results in two phases,
frst we show that our algorithm works well with some routine
scheduling algorithms and second we show the performance of
our algorithm in comparison to some other CAC algorithms.
We simulate our algorithm with Matlab tools, and the results
show proposed algorithm is quite efcient.
cy0r05-C0h00m550n C0nlr0l; 5Chc0uhng; LLL0J.1
LINTRODUCTION
IEEE 802.16 standard has been proposed to provide high
speed broadband wireless connectivit with QoS guarantee
for connections. This standard has a base station (BS) and
one or more subscriber station (SS). BS provides a gateway
between SSs and Interet, and SSs provide access to BS for
local users.
Providing QoS is very important in IEEE802.16, and
because of the different trafc classes in the network, IEEE
802.16 introduces four types of services: UGS for real time
data stream with fxed size data packet on a periodic basis
(like VOIP). rtPS for real time data stream with variable
sized packets on a periodic basis (like video MPEG), nrtPS
for non real time polling service (like FTP), BE for data
steam without any QoS guarantee (like HTTP). This
standard determine signaling for providing QoS, but some
main blocks like call admission control (CAC) and
scheduling was lef for researchers.
The aim of CAC is to achieve a certain QoS for ongoing
and accepted requests with limiting connections in the
system. So, when there is a new request for system, CAC
must evaluate system QoS, departure rate for non real-time
connections and delay for real-time connections, and make a
decision to accept or reject the request. If new request is
accepted, QoS must be guarantee for new connection as wel1.
[I ].
There is some research in CAC, and some algorithms
were presented. A good CAC algorithm must have four traits:
Masoud Sabaghi
Department of Electrical
Engineering
Azad University branch of
Ashteyan, Iran
msabaghi@gmail.com
Mohammad Fathi
Department of Electrical
Engineering
Amirkabir university
Iran, Tehran
mfathi@aut.ac.ir
1) minimize handof failed; 2) maximize accepted request; 3)
provide bandwidth efciency and 4) reduce delay [2].
In [I], a CAC algorithm based on a conservative
approach was presented called conservative and quality of
service (CAQoS). In [2], a joint strategy of bandwidth
allocation and admission control for elastic users competing
for a downlink data channel in cellular networks was
considered. In [3], a multiclass CAC based on reservation
pool for handoff requests was presented. In [4], a dynamic
CAC algorithm for UGS connections was proposed.
Most of these algorithms satisf one or two parameters
mentioned for a good CAC algorithm. The aim of this paper
is to introduce a new CAC algorithm that satisf all four
parameters, provide suitable QoS for service classes (reduced
delay for real time service classes and increased departure
rate for non-real time service classes) and reduce blocking
for new requests and hand off connections.
To achieve this goal, we make decision during T slot time.
With this strategy, we make a better decision than other CAC
algorithms and there are fewer faults.
Usually a CAC algorithm is evaluated with a scheduling
algorithm, so we evaluate our proposed algorithm with three
scheduling algorithms and will show that it works well with
them. Also, we will compare our proposed algorithm with
three other CAC algorithms and will prove that our algorithm
out performs them.
The rest of this article is organized as follow. In section 2
we present CAC algorithms, at frst subsection we introduce
our proposed algorithm, then we have an overview of some
scheduling algorithms at second subsection and we describe
different CAC algorithms at third subsection. In section 3, we
provide the result of simulation of proposed CAC algorithm
with diferent scheduling algorithms at frst subsection and
compare diferent CAC algorithms at the second subsection.
II.TH CAC ALGORITHMS
In this section, we frst introduce proposed CAC
algorithm in the frst subsection, and then, in the second
subsection, we introduce scheduling algorithms that we use
in our simulations. In third subsection, we introduce some
other CAC algorithms.
A. PROPOSED ALGORITHM
In this subsection, we introduce our proposed algorithm
for making decision to accept or reject the requests. Table I
shows parameters that are used in the pseudo code.
J-1-4Z44-14-3l$Z.00 | Z010III N4-J1 NOumC 4
2010lnternationalCoperenceOnComputerDesignndppliations(lCCDA 2010)
In our algorithm, handof and UGS connections have
highest priority in the system. Therefore, in each slot the
algorithm calculates the amount of bandwidth required for
handof request; if it is less than the available bandwidth, it
will be accepted and otherwise it will be rejected. Then, we
survey UGS requests, If there is enough bandwidth available
and ongoing connection are in a good condition, the request
will be accepted and otherwise it will be rejected.
We assume that the remaining bandwidth
C - C - C
(
'olol handof UO5 _
is allocated to the ongoing
connections in the rtPS, nrtPS and BE class of services. Our
CAC algorithm makes decision as follow. For response to the
new requests, we study rtPS, nrtPS and BE classes
respectively. Let us assume that NC is the new connection
request. Making decision to accept or reject the new request
is shown in the algorithm I. We assume that departure rate at
the beginning is more than the threshold amount. This
algorithm tries to decrease departure rate of other
connections and increase departure rate for the new
connection. If departure rate for ongoing connections and the
new request stay above the threshold, it will be accepted and
otherwise it will be rejected. In this algorithm
__
are
constant amounts that must be small, until the algorithm
synchronies during T slot time.
. SCHDUG ALGORITHMS
In this subsection we review the scheduling algorithms
that CAC algorithm use.
I) Round Robin (RR) algorithm: This algorithm is a
simple algorithm that divides bandwidth equally between
queues.
ALGORITHM I PROPOSED CAC ALGORITHM
1. Upon arrival of any new connection NC
2. for t=1 to T
3.
f;,ps = '|- f;,ps )
4.
frtPs = ':
(
f
-s - f,ps )
5.

f
f = '
(f
Z -
f
;,ps
)
6
1+1 ,
\
t
fr,ps = frtps - ffrtps
7
1+1 ,

'

fnrtps = Jnrlps - fnrtps
6

f
= f
E -
f
l
,+1 ,
\
t
\
t
\
t
fNC = fNC ffrtps ffnrtps
10. end
11 'f (,,1
+
I
>
f
'
H
d
f
1
+1
>
f
'H d
f
T+
I
>
f
1
H
'` \rrps rtps
an
nrtps nrtps
an
bl bl
and
f
;
1
>
f
)
12.admit the arriving connection NC.
13.else
14. block NC.
15. end
2) Modied Largest Weighted Delay (W
algorithm: This algorithm is one of the best algorithms that
exist. ML WDF guarantee delay and throughput. In this
algorithm, at each time slot, queue j will be served as
follows
p/t)
=
yjw/t)r/t)
(1)
Where (t) is the bandwidth allocated to queue j, W
J
(t) is
the head-of-the-line packet delay for queue j, l (t) is the
channel capacity with respect to fow j, and are arbitrary
positive constants. For better result we can replace
W_ (t) with the length of the queue for some services [5].
3) f ' WFQ ' hybrid algorithm: This
algorithm is a combination of Earliest Deadline First (EDF)
for rtPS class of services, Waited Fair Queue (WFQ) for
nrtPS class of services and FIFO for BE class of services.
In EDF algorithm, packets are sort according to their
deadline. So, packets with earliest deadline will be given
service sooner than others. In WFQ algorithm each
connection has a weight, and packets are sort according to
their weight and packets with highest weight will be given
service sooner than others [6].
C CAC ALGORIT
We chose ML WDF as a scheduling algorithm for the
following CAC algorithms.
1) ALGORITHM 1
This algorithm is based on token bucket and acts as
follows:
Step 1: Compute remaining bandwidths according to:
B . = C -
t
-
t
-
t
-
t
remaIn total /UO5 frtps J.nrtps /bl
(2)
Step2: Compares remaining bandwidths with new
request. If there are enough bandwidths, the new request
will be accepted, else follow to step 3.
Step 3: if lower class fows (BE and nrtPS) have
bandwidth more than their threshold
( f
,
f
:
s
), allocate
less bandwidth to them, otherwise follow to step 4.
Step4: if higher class fow has bandwidth more than its
threshold (,), decrease departure rate of it to r _rtPS, else
go to step 5.
Step J. reject the new request [7].
2) ALGORITHM 2
This algorithm is proposed in [6]. The algorithm assumes
that:
N4-JZ NOumC 4
2010lnternationalCoperenceOnComputerDesignndppliations(lCC 2010)

There are N rtPS connections with r token bucket

(packet per slot) and b token bucket size (packet) and
delay [d )
TABLE! lAkAVL!LkU8LO! !PLl8LUOOtOOL
ParamcIcr UcscripIicn
'tt.
Total available bandwidth
'h
Request bandwidth for hand off connection
':.:
Request bandwidth for UGS service
/
fnps
Departure rate for rtPS queue at slot t
/
fnnps
Departure rate for nrtPS queue at slot t

C
npS =
d

m
=
" that D must be integer and f is session
J
t
duration

To guarantee delay in connection i with rtPS fow,

they use
C
. ..
|,
[(m, -i).
i
)-i
(3)
fnps
And the algorithm policy is as follow
StepI: for rtPS connection:
a) If there 1b enough bandwidth
' 'omad
C
;;j f;
S
/,
), accept new request,
else reject it.
b) Check delay guarantee (equation 2), if it satisfes
guarantee delay for new connection and accept it,
else rej ect it.
c) Check delay for ongoing connections (equation 2).
If it does not satisf for even one connection, reject
new request.
d) Update rtPS connection.
Step z. for nrtPS connection, if there is enough
bandwidth ( = C -

-f
'
-f
'
) accept it
~mnd tt [JGS ups nnps
and update nrtPS connections.
Step 3: accept 1request without any guarantee.
3) ALGORITHM 3
In [4, the author focuses on nrtPS service. If there is
enough bandwidth available (there is few connections in
the system), nrtPS fow could have high departure rate.
When amount of request for connecting to the system
increases, it decreases departure rate for ongoing nrtPS
fows until the amount of UGS, rtPS and nrtPS connections
increase in the system. They call this action, degradation.
ParamcIcr UcscripIicn

rtPS delay threshold

/.
Departure rate for BE queue at slot t
H
fnnps
nrtPS departure rate threshold
/
BE departure rate threshold

Positive constant
This algorithm acts as follow:
Step I: when there is new rtPS request, if current
available bandwidth is enough for ongoing connections and
new request, accept new request, otherwise decrease
departure rate for nrtPS connections step by step. If
degradation receives to maximum amount and there is not
enough bandwidth, reject the new request.
Step z. when there is a new nrtPS request, if there is
enough bandwidth accept it, else decrease departure rate for
ongoing nrtPS connections until it can be accepted. If there
is not enough bandwidth, the new request will be rejected.
Step 3: when there is new 1 request, if another service
do not need bandwidth, 1request could have bandwidth.
IILSIMULA nON
We show the results of simulation in two subsections. In
subsection one we show the results of simulation for our
proposed CAC algorithm with three scheduling algorithms
that we introduced. In the second subsection, we show the
results of simulation for comparing our algorithm with other
CAC algorithms that we mentioned.
We consider a scenario with i:requests, S requests for
each service. Afer time slot 200, new requests reach to the
system one by one afer 100 time slot. Thus rtPS, nrtPS and
1 requests enter to the system respectively.
A. PROOSED ALGORIT WTH DFENT
SCHEDUG ALGOT
Figures i-show throughput and delay for mentioned
algorithms in section 2, subsection 2. As you can see,
departure rate for 1 and nrtPS remain higher than the
threshold amount (, = :aa,, = +aa and delay for rtPS
remain less than rtPS threshold (
f
:= 6), and also in all
of them, all new requests are accepted. The time of request
arrivals are marked, rtPS requests are marked with r, nrtPS
requests are marked with nr and 1 requests are marked
with 1.
N4-JJ NOumC 4
2010lnternationalCoperenceOnComputerDesignndppliations(lCCD2010)
6`

E
r
E

' , r r
_ __

.

` __ .r.p, .|... .. , . . ...
l ''
p
^

|__ .

. - -- - -| - _ -| -|.- _.- .-

j__

..

..

.
-riPS
nRlps
uL

l
_ ,__j_ _,__ _ _, __ _ _,_ _ _, ,__, j_ _, , __ _ _ , __ __ __
SIcl
Figure I. Departure rate with Rscheduling algorithm

....._..._..

aza c+ cc: aa s a a ccatza at + cc |: cc

SIoI
Figure :Delay for rttPS fows with Rscheduling algorithm
t+aa -

_ --
v p; t .

_ tzaa

, ~~
u

` l l
= taaa

+ -I nr~ ~+
C | - . O
_ saa r - ~ ~
:aa
..

.. .

. ...
, l l l l l `
5 +aa

=4 -1 rl-__ - +
l l l l , l l
,
.

zaa
'. . ....... .. .....+.....

l l l l l l l
aRis
u|

:aazaazaa+aazaa :aa /aa saa:aa:aa~ta:iza:iza~+a:iza~:a~/aa

IcI
Figure 3. Departure rate with MLWDF scheduling algorithm

. . _ . . .

. . . _.

. . _. . . . . . _ .
l l l l
l l l l

._.

. .
l
. . . ._
a
_ _ _ _ _ _ _
a zaa +aa :aa saa |aaa :aa
SIoI
Figure 4. Delay for rttPS fows with MLWDF scheduling algorithm

I
'

0 3
0
5IOl
Figure 5. Departure rate with hybrid scheduling algorithm

........

...

_. ..._....

.. . ._. . . ._... _. .._..._ .

|
_ _ _
...
_
....
l
....
_
....
_
........

....

.....
_
l
_ _
l
_ _
__ _ -
bIOl
Figure 6, Delay for rttPS fows with hybrid scheduling algorithm
. COMPARNG CAC ALGORTHMS
To compare these CAC algorithms, we chose ML WDF
scheduling algorithm.
Figures 7-9 show the result of departure rate for nrtPS
and 1 with diferent CAC algorithms and variable bit rate,
fgure 9 shows delay for rtPS algorithm.
The mentioned CAC algorithms in section 3 made
decision on admitting or blocking new request as soon as it
has been arrived, whilst our proposed algorithm take some
times (T slot) to make this decision. (Figure 7 and Figure 6)
Some parameters like arrival trafc and channel state
information in a wireless network are non deterministic, i.e.
vary over time. So making a decision at the moment might
not be the right one, while our algorithm makes decision
during a time period, so it has more comprehensive
information about the system to make a better decision.
Another point that is necessary to consider, is that
admitting a new connection and providing bandwidth for it
at just one moment causes the allocated rate for existing
connections to vary abruptly. This issue causes instability to
the network that needs more signaling overhead to stabilize
the network. In the proposed CAC, we vary the allocated
rates gradually over time thus avoiding rate oscillations.
Table II list amounts of accepted requests in the system
and you can see our proposed algorithm has more
acceptances with QoS guarantee for all services.
N4-J4

saa {-:,

AIgcr|Ihm2
AIgcr|Ihm3
Prcpc
a
a:

cc aa
zaa
+ cc z
aa

:aa
,
cc

cc

a
-
a
|a
_ ~~ |
a
_ a
_ z
a
_
+

a
_ za
_
:

a
,
aa
S|Ot
Figure 7. Departure rate for nrtPS with diferent CAC algorithm
NOumC 4
2010lnternationalCoperenceOnComputerDesignndppliations(lCCD2010)
"
,
:
00

0
0
0
b!OI
Figure 8. Departure rate for BE with different CAC algorithms
sc
U
:c
-- Algrithm 1
Algnlhr2
c
' sc
5
zc
"
ic
Figure 9. Delay for rtPS with diferent CAC algorithms
TABLE II.
Algorithms
Algorithm 1
Algorithm2
Algorithm3
Proposed
AVOU!!OIAttLl!kLQUL8!! OIIILkL!!tAt
AL0OkI!BV8
rtPs nrtPS BE
4 request 3 request 3 request
0 request 3 request 0request
4 request 4 request 1 request
5request 5request 5 request
IV.CONCLUSION
In this paper, we introduced a new CAC algorithm, and
we show that our algorithm throughput is good with
different scheduling algorithms, and we showed that our
algorithm works very well in the comparison to other CAC
algorithms. It has less rejects for new requests and provide
QoS for ongoing connections and the new connection. To
achieve this goal, we use temporary accept policy at the
system, then we evaluate system performance and make
decision to accept or reject this request.
REFRENCES
[1] Pia, JVirtamo, JMartinez-Bauset, "Optimal robust policies for
bandwidth allocation and admission control in wireless network,"
Computer Netork, vol. 52, no. 17, pp. 3258-3272, 2008.
[2] Y C. Yee, K. N. Choong, LYLow, S. W. Tan, S. F. Chien, "A
conservative approach to adaptive call admission control for QOS
provisioning in multimedia wireless networks," Computer
Communications, vol. 30, no. 2, pp 249-260, 2007.
[3] F. Hu, N. K. Sharma, "multimedia call admission control in mobile
network: a dynamical reservation-pool approach", Computer
Netorks vol. 4J pp 263-288, 2003.
[4] H. Wang, B. He, D. P. Agrawal, "Above packet level admission
control and bandwidth allocation for IEEE 802.16 wireless MAN,"
Simulation, Modeling, practice and Theory, vo. IJ, pp. 366-382,
2007.
[5] M.Andrews, K.Kumaran, K.Ramanan, A. Stolyar, P.Whiting,
"Providing Quality of Service over a Shared Wireless Link," IEEE
Communications Magazine , vol. 32, no. 9, pp 150-154, 2001.
[6] K. Wongthavarawat, A.Ganaz, "Packet Scheduling for QoS
Support in IEEE802.16 Broadband Wireless Access Systems,"
Interational Joural of Communication Systems, pp. 81-96, 2003.
[7] T. Tzu, H. i. Chi, Y W. Chuang, "CAC and packet scheduling using
token bucket for IEEE802.16 networks", Joural of
Communication, Vol. I, No.2, pp 30-37, 2006.
N4-J3 NOumC 4