High Efficiency QoS Guarantee, Channel Aware scheduling scheme For Polling Services in WiMAX

Document Sample
High Efficiency QoS Guarantee, Channel Aware scheduling scheme For Polling Services in WiMAX Powered By Docstoc
					                                                               (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                                               Vol. 9, No. 2, February 2011




            High Efficiency QoS Guarantee, Channel Aware
          scheduling scheme For Polling Services in WiMAX
 

                                  Reza Hashemi, Mohammad Ali Pourmina, Farbod Razzazi
                                     Department of Electronics and Communication Engineering
                                 Islamic Azad University, Science and Research Branch, Tehran, Iran
                                reza_hashemi@ieee.org, pourmina@srbiau.ac.ir, razzazi@srbiau.ac.ir
       
   Abstract—This Paper offers an efficient channel aware                     transmission media, in wireless system, for its extreme time
scheduling scheme for IEEE 802.16e WiMAX Mobile, real-time                   varying nature, we need to consider channel condition to
and non-real-time polling service. Compared to a similar                     prevent waste of resource.
scheduling approach, our considered scheduler can guarantee
                                                                                 Mobile-WiMAX uses TDD mode that makes channel
and achieve lower delay with a good average throughput. In
order to achieve this object, we introduce a scheduling scheme               estimation easier, also operates in 2-11GHz Frequency Range.
with four different segments in a decision making process. The               Both access technologies, OFDM and OFDMA can be used in
first part, a time dependent function that considers the time when           WiMAX. Our used technology is OFDMA that increases
packets wait in queues and in a jitter area to prevent packet                Bandwidth utilization but makes scheduling problem more
deadline. Buffer utility function, as the second part, considers             difficult. Scheduler in OFDM decides for OFDM symbol and
buffer size in scheduling to prevent overflow, specifically in nrtPS
                                                                             all subcarriers are allocated to one user, nevertheless each
class with large size packets. The third part, retrieved from
proportional fairness algorithms, which in normal conditions                 subcarrier can select different Modulation and coding scheme
gives a fair share to users. Channel SNR and service class weight            which makes it difficult to estimate average rate in a forward
are also involved in this part. The final section of scheduling              frame. Decision making for OFDMA is much more difficult
relationship, channel condition, is defined more accurately by               than for a time scheduler, because it must select subchannel in
RSSI and CINR parameters. The simulation results in OPNET                    a frequency domain.
show that our proposed scheme has a very good delay and packet
loss ratio accompanied by a high throughput. In another                         The smallest allocation unit by Scheduler is called slot. In
scenario, with different number of users and limit resources, we             OFDMA, slot is a combination of some subcarriers and some
show relationship between admission control and scheduling.
                                                                             OFDM symbols. Depending on permutation process, slot may
                                                                             have different definitions. For instance, for uplink in PUSC,
  Keywords-component; IEEE 802.16e; WiMAX; Scheduling;                       each slot consists of 3 OFDM symbols and 16 subcarriers. The
QoS; Resource Allocation; OPNET
                                                                             other issues are burst in a downlink that must be rectangular
                                                                             Fig. 1. which makes it difficult for a downlink scheduler to
                       I. INTRODUCTION                                       select the slot in such a manner that best fits to the user
                                                                             allotment [3], [4], [5]. Packing and Fragmentation are other
I  N recent years, bandwidth hungry applications, such as
   video and music streaming, large file downloads, etc have
been significantly used. Wireless and vehicular accession to
                                                                             options that can be used by WiMAX equipment to fit MAC
                                                                             SDU in Mac PDU.
such contents lead companies and standard organizations like,
3GPP and IEEE, to develop BWA technology. IEEE 802.16
standard families with long distance and QoS mechanism
support are among the important and active technologies for
these Issues also, counted as a strong 4G candidate under the
development of 802.16m standard version. In our survey,
IEEE 802.16e mobile-WiMAX standard [1] has been studied
for its special features like power management and handover
capability rather than for its fixed version. Two PHY and
MAC layers are defined by standard, which in Medium
Access Control layer are responsible for QoS mechanism such
as call admission control and scheduling. For resource
allocation, channel aware scheduling [2] are cross layer
                                                                                            Fig. 1. OFDMA Frame in 802.16e
processes which use some physical layer parameters like SNR,
CINR and RSSI for decision making procedure. Unlike
channel unaware scheduling which assumes error free




                                                                       233                             http://sites.google.com/site/ijcsis/
                                                                                                       ISSN 1947-5500
                                                             (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                                             Vol. 9, No. 2, February 2011




         Fig. 2. Admission control and scheduling for user                             Fig. 3. Centralized scheduling in WiMAX

Both of the two procedures utilize transmission frame but                  system real capability.
increase overhead and must be considered in scheduling. The
most important parts in QoS are scheduling and call admission                           II. POLLING AND RELATED WORKS
control Fig. 2. which open spaces in standard and have been
                                                                             A. Polling Service
done by MAC layer. Medium access control layer consists of
3 sublayers; a) convergence sublayer, b) common part                         Bandwidth request in WiMAX is categorized as either
sublayer and c) security sublayer. In convergence sublayer,                implicit or explicit method. In WiMAX mobile, there exists as
packet classification and getting proper CID and SFID have                 a whole, 11 different ways for bandwidth request. Unsolicited
been done. According to the requirements and the specific                  request, bandwidth stealing, poll-me bit, piggybacking,
parameters, packets are classified into 5 QoS service classes.             codeword over CQICH, CDMA code-based and contention
1) UGS: this class gets fix bandwidth without any overhead                 region based are as implicit method. Polling based methods
and can guarantee QoS, but wastes resource when traffic                    like unicast polling, multicast polling, broadcast polling and
changes. 2) ertPS: this service is suitable for VoIP traffic with          group polling are categorized as explicit method. Guarantee
silent suppression and needs a polling mechanism to inform
                                                                           QoS, required information about the user queues such as
end of the silence, similar to UGS, QoS parameters are
                                                                           buffer size and head of line packet states. There is a need for
maximum latency tolerance, maximum sustained rate and
tolerated jitter. 3, 4) rtPS, nrtPS: These two service classes are         delay and throughput guarantee in an uplink for a service
for real-time and non-real-time variable rate traffic. For                 class. Its suitable and possible bandwidth request way is
varying nature of packet size like video streaming in rtPS and             polling based method. In this way, station polls users for
FTP download in nrtPS. Polling mechanism is needed to                      requesting slot in a periodic interval to transmit their packets.
specify what amount of resource must be granted to the users.
                                                                             In polling mechanism, at first, user bandwidth must be
For nrtPS, there exists no delay guarantee but minimum
throughput is guaranteed. 5) BE: most of the traffic is                    admitted by admission control unit. According to QoS
classified to this QoS class of services. After all other classes          parameters defined for queues base station, poll users in a
being allocated, there will be no QoS guarantee and queue to               periodic interval to request for a bandwidth. These polling
use the remaining resources when other class was allocated.                intervals may be addressed to individual SSs (unicast polling)
  In point to point configuration, WiMAX uses a centralized                or to groups of SSs (broadcast or multicast polling). Polling-
scheduling Fig. 3. It means, Base station makes decision for               based service scheduler uplink traffic makes decision for
uplink and downlink traffic. Even the grant for a user is based            queues and then grants bandwidth, in accord to the available
on GPSS (grant per subscriber station), we’ll need another                 resources and the number of users. Users by decoding UL-
scheduler in mobile station. Another grant method, GPC                     MAP in uplink can be informed about their grants.
(grant per connection) was outdated in IEEE 802.16e. Before a
                                                                           Choosing suitable approach and polling mechanism delay are
packet be classified and a scheduler make a decision, the call
                                                                           the two problems for this method [6]. Unicast polling prevents
admission control unit accepts or rejects the new connection,
                                                                           the request collisions and can guarantee the delay, but by an
according to the estimation of a system capacity. Clearly,
                                                                           increase in the number of stations, tremendous bandwidth for
inappropriate capacity estimation by CAC unit degrades
                                                                           polling are required which decrease bandwidth for the grant.
scheduling performance, especially when it accepts more than




                                                                     234                              http://sites.google.com/site/ijcsis/
                                                                                                      ISSN 1947-5500
                                                             (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                                             Vol. 9, No. 2, February 2011

                                                                           Where,       is an achievable data rate of user i and is an
                                                                           average data rate in a given time window T t .         t is
                                                                           computed by an exponential averaging in past Ti windows
                                                                           expressed in Eq.2 :

                                                                                     1          R t         1                             q t     0
                                                                              t           T                            T                              (2)
                                                                                    R t       1                                             q t   0
        Fig. 4. Polling based bandwidth request mechanism
                                                                             T t has impact on throughput, but its accurate selection is
Multicast or broadcast polling mechanism, according to                     difficult. Proportional Fairness algorithms do not guarantee
contention-based area, is a better approach for a large number             delay or throughput, also short time fairness are not satisfied.
of users, but decreases throughput of the system.                          It cannot be a proper method for a delay sensitive traffic and
  According to Fig. 4. polling mechanism in WiMAX is a 3                   for an application minimum throughput requirement, and it
way handshaking process which increases delays in the                      needs to be modified differently.
queues. A user waiting for poll intervals also must wait for
                                                                             QoS guarantee algorithms provide delay and throughput
the response of scheduler for grants. In the best situation, after
                                                                           requirements for each service class that needs QoS. M-LWDF
2 frames, request for queues, can access to the channel.
                                                                           families [12] are the most important algorithms in this
  B. Common Channel Aware QoS Scheduling Algorithms                        category that try to modify LWDF, in which throughput, are
  Fairness, delay, throughput, Energy Consumption, Power                   optimal. As one of these approaches in [13] queue i that
Control, Complexity and Scalability are the important                      maximizes Eq.3 in subchannel k, can get permission to
parameters in scheduler design, a metric evaluation and                    transfer its packet.
comparison. Scheduling in WiMAX can be classified in two
main parts, channel aware and channel unaware methods. In                           _           ,                  _             _                    (3)
which these method can be used for intra-class and inter-class
scheduling. Most of the channel unaware scheduling comes
                                                                            In this equation       is throughput in coming frames and
from router and CPU fundamentals that extended for WiMAX
                                                                                is average throughput in past specific time window.
[7], [8], [9], [10]. This series of algorithms assume an error
                                                                           Channel gain is the normalized ratio of the square of noise at
free idle channel for each user, and share resources according
                                                                           the receiver and the variance of Additive White Gaussian
to their capacity and QoS parameters. In wireless transmission
                                                                           Noise. HOL_delay is a waiting time in buffer for packet in
channel, for each user, conditions differ and degrade in
                                                                           head of the queues. This algorithm using some buffer state
frequency and time domain. So channel state must be
                                                                           information in decision making, are useful for QoS guarantee.
considered in resource allocation decision making process.
WiMAX uses CQI Channel to inform base station about
channel conditions. CQICH information primarily is used by
adaptive modulation and coding module to select the best
scheme in transmission. Channel aware scheme also can use
these channel parameters in RSSI and CINR decision
makings. For the algorithms that use channel state, four main
categories can be named as: Proportional Fairness based, QoS
guarantee based, power constraint and System throughput
maximization.
                                                                                                 Fig. 5. Time Utility Function
The goal, in PF-based scheme [11], is to achieve the long-term               Another QoS-based approach is UEPS scheme [14] that uses
fairness between the queues, especially in BE service class                time utility function to make an urgency when, packets in
which offers no guarantee for quality of service. In PF-based              queues, enter jitter area or next to deadline.         is time
scheme each user who can maximize Eq.1 gets an opportunity                 utility function for delay. According to Fig. 5. when packets
for transmission.                                                          enter to a jitter area, first derivative of      increases and
                                                                           Eq.4 gets more weight and consequently, high probability to
                                                              (1)          access the resources.




                                                                     235                                     http://sites.google.com/site/ijcsis/
                                                                                                             ISSN 1947-5500
                                                              (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                                              Vol. 9, No. 2, February 2011
|      |                                                      (4)                                         Table I
                                                                                         Si selection for different polling service
                                                                               Class of service                         Si
Where,        is similar to     and     , is a channel capacity
                                                                                     rtPS           Number of frame needed to transfer packets
for queue i.        is equally calculated by Eq.2 and          ,                                                        ×A
time utility function, is the absolute value of the first                            nrtPS            Number of frame needed to transfer packets
derivative of      . UEPS do not consider buffer size that may                                                          ×B
overflow when the packet size is big.
                                                                                        1         U t         1                          q t      0
In throughput maximization algorithms [14], user with better                   t                                        S                             (5)
channel state, in a heuristic approach, gets more resources                                 t   1                                           q t   0
without QoS guarantee and fairness. Two important factors in
                                                                            Si is time window that measures average user data rate. In
channel quality are RSSI and CINR, where CINR has more
                                                                            proposed scheme Si, relationships have been exerted according
weight. One of these approaches is MAX CINR, in which                       to the number of required frames to transfer user packets, and
user with the highest CINR in its channel, gets more                        also weight of classes is inserted in this part. Si selection is
permission to transfer in channel. MAX CINR cannot                          done according to Table I.
guarantee QoS and degrade fairness. To choose user with the
best channel condition, maximize throughput.                                  Where A and B are rtPS and nrtPS weight in resource
                                                                            allocation. Prior to this, weight of class is used to make a
  Another channel aware algorithms category is power                        decision in inter-class scheduling. When number of frames,
constraint based approach which considers sleep and Idle                    needed to transfer user packet, increases, users have fewer
                                                                            chance to get resources. This is what we need for rtPS with a
mode and also battery limited [15]. Using power management
                                                                            small packet but it must be transferred as soon as possible.
method increases delay, because mobile station goes to sleep
for a specific interval till base station wants to be aware or                B. Deadline
unaware periodically. Packets in buffer remain until mobile                   rtPS is a real-time service class where, more often, traffics
station is aware.                                                           sensitive to delay are classified in this category. Packet size
                                                                            for rtPS is not big and if it does not transmit in a specific time
III. SOFT TRACKING VARIATION OF THE PARAMETERS OF QOS                       interval, deadline, system throws the packet away. Deadline is
                                                                            counted relative to the waiting time in the buffer, and if it rises
  Scheduling decision consists of 4 main parts the goal of
                                                                            from determined threshold, related to its application, it’ll be
defining each algorithm is to meet QoS by tracking important                thrown away by system. We implement an emergency when
parameters more accurately. For polling services rtPS and                   packets are close to deadline to prevent loss of packet.         is
nrtPS we do not need two schedulers for intra-class and for                 time utility function in Fig. 6. which has 2 different graphs,
inter-class scheduling. For simplicity the weight of two polling            T1 and T2. In our proposed scheme we use the first
classes is inserted into main equation. In spite of the                     derivatives of       .
simplification of scheduling to remove 2 steps scheduling into
1 step, if parameters are not assigned properly, algorithms
create cross point that degrade performance of scheduling.

  A. Average Rate Updating
  At First part of proposed scheduling scheme, we use
Proportional fairness relationship which provides long-term
fairness among users. In normal conditions when no packet is
near to deadline or buffer is not shortly to overflow, it is better
to keep fairness among users in achieving resources. We make
some changes in PF formula,           to adapt them to rtPS and
                                                                                      Fig. 6. Time Utility Function in Proposed scheme
nrtPS service classes. According to user channel and his SNR
in calculated sub-channel,          is achievable data rate in
                                                                               In normal condition when packet is not near its deadline,
coming frame. After that adaptive and modulation coding
                                                                            first derivative of      is very small and there is no need for
decides what types of MCs are suitable for user subcarriers,
                                                                            urgent transmission. When packet stays for a long time in
scheduler add all of them to find out user data rate in coming
                                                                            buffer and closer to its deadline, emergency occurs. For two
frame.       is calculated by Eq.5.
                                                                            polling services, we select time utility function by Table II.




                                                                      236                                  http://sites.google.com/site/ijcsis/
                                                                                                           ISSN 1947-5500
                                                             (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                                             Vol. 9, No. 2, February 2011
                             Table II                                                                         Table III
                 Ti for different polling service                                                    for different polling service
               Class of service             Ti                                                   Class of service
                     rtPS                   T2                                                         rtPS                 B1
                    nrtPS                   T1                                                        nrtPS                 B2

    Packet waiting time previous to enter to jitter area, first                 Overflow for nrtPS with large packet size is more possible
derivative of rtPS increases, and is shown in T2. Jitter does not              than for rtPS. For this reason, we choose B2 for nrtPS in
define QoS parameters for nrtPS service class but in this part                 which the impact of incline and consequently weight of
of scheme, according to application, we defined threshold and                  buffer is softer and increases before being close to overflow.
jitter area for non real-time traffic, when get permission to                  According to our experiments for rtPS, we don’t need to
transmit while waiting too much time in queues. According to                   consider buffer size because of packet size, until 85 percent
T1, in nrtPS, when packet enters the predefined jitter area, the               of overall buffer size is filled. When size of the buffer
incline of T1 increases more rapidly and causes more weights                   reaches to 85 percent relative to overall size, something like
in scheduling decision. In this way we can guarantee delay for                 emergency has occurred and slope of B1 increases which
real-time traffic. Important point is that deadline in nrtPS is                causes to more weight in scheduling decision.
very large about 3 seconds, for example, but in rtPS is about
                                                                             D. CINR and RSSI
milliseconds.
                                                                             In average rate updating formula and       t , to consider
  C. Buffer overflow                                                       channel state we only use SNR, which is in WiMAX, power of
   Buffer size and number of packets in a buffer are the                   noise are calculated by Eq. 6.
important parameters which must be considered in making                          174       10   log
                                                                                                                       
                                                                                                                                                     (6)
                                                                                                                   
decision for bandwidth allocation, especially when the size of
packet is big. When packets do not get permission to transmit                 -174 is normal thermal noise in dBm/Hz for base station
and with a finite buffer size, overflow can occur. Delay for a             environment, BW is overall system bandwidth and n is the
thrown away packet, because of overflow, considered as                     number of OFDM symbols. In SNR we cannot accurately
deadline time and this time is added to average delay and                  determine channel state. CINR and RSSI are two important
increases it. Most of the time, buffer overflow degrades                   parameters in channel quality which are considered to make a
system performance and causes very bad damage.                             better channel estimation for scheduling decision. CINR is
In this scheme we introduce a buffer utility function alike a              already and usually used in throughput maximizing scheduling
time utility function to manage buffers and involve their states           algorithms, and RSSI almost used in power constraint scheme
in scheduling decision.     t , with two graphs B1 and B2, is a            to consider power in each subcarrier.
function which depends to buffer size and is illustrated in Fig.              We introduced           ,       defined by Eq.7 for a better
7.                                                                         consideration of channel state in making a decision.

                                                                                       ,              |log                    |                      (7)

                                                                           CINR are more important parameter than RSSI. For example,
                                                                           a condition those two Base Stations are close to each other, in
                                                                           this situation when user is near the base station, RSSI is high
                                                                           but CINR is low which means Signal strength is high but
                                                                           channel condition, is not really good because of interference.
                                                                           High CINR shows better channel with low packet loss
                                                                           probability. According to the ranges of CINR and RSSI we
                                                                           offer Eq.7 where CINR has more weight in equation.
                                                                             E. Final relationship
                Fig. 7. Buffer State Utility Function
                                                                              Final relationship in our proposed scheme for scheduling
                                                                           decision making is represented by Eq.8 for real-time and non-
   In nrtPS service class, size of the packet is big and it is not
                                                                           real-time class of service.
 so much sensitive to delay, but in rtPS, size of the packet is
 not big and the possibility of overflow in finite buffer is low.
 According to Table III, we select       t and use the absolute            |       |       |      |                       ,                          (8)
 value of the first derivative of    t in our proposed scheme
 to consider different polling classes and their impacts.                    And according to the previous section, |                | is the
                                                                           absolute value of the first derivative of time utility function,




                                                                     237                                      http://sites.google.com/site/ijcsis/
                                                                                                              ISSN 1947-5500
                                                           (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                                           Vol. 9, No. 2, February 2011
                                                                         achieveing bandwidth. There are two possible ways for
At each scheduling instance {                                            resource limitation, one way is to increase number of users,
for j=1 to N { update         ,      , RSSI(j) , CINR(j) , and           then output analysis would be much more difficult, and the
                                                                         other one is limitation by Admission control unit. In our
                                                                         simulation the latter is used, by Admission control we create
                                                                         competition for bandwidth .
if ( Qj(t) ≠ 0 ) {Uj(t) = * Rj(t) + (1- )* Uj(t-1) }
if (Qj(t) = 0) {Uj(t) = Uj(t-1)) }                                         B. Schedulers for Multiple Traffic Classes
}                                                                           At first we design a simple scenario to compare our scheme
QoS_schedule = 0                                                         with some main QoS guarantee and Fairness channel aware
for j=1 to N { if (Qj(t)) > 0) { QoS_schedule = 1 } }                    algorithms.
if (QoS_schedule > 0) { IS= arg max j
(      *        *M(RSSI(j),CINR(j))*           ) }

<Variables>
      : Time utility Function for MS j's
   : Buffer Utility Function for MS j's
RSSI(j): Received signal strength Indicator of MS j's (mw)
CINR(j): Carrier to Interference and Noise Ratio of MS j's
(mw)
N : Number of MS’s having QoS class connection
      : Data Rate for MS j's in coming farme according to
bits/sec
IS : Index of the selected MS
       : Average Data Rate in Si Farme before                                          Fig. 9. First simulation scheme in OPNET
Qj(t): Buffer state for MS j's according to number of                       In Fig. 9. we define 3 user station and one base station that
packets                                                                  is directly connected to the server. Stations have 3 different
                                                                         classes of service, ertPS, rtPS and nrtPS where resources are
              Fig. 8. Algorithms for proposed scheme
                                                                         limited by admission control. For ertPS node we define voice
                                                                         application whose frame size is 57 bytes and with 0.25
|     | is absolute value of the buffer size in dependant
                                                                         seconds inter-arrival time. In second node, we implement rtPS
function,       is user’s data rate in coming frame,          is         with video application in which frame size have chi-square
user’s average data rate in Si time window of past, and finally          distribution with a 32kbytes mean and 12frame/sec. At FTP,
         ,       is channel state function for better measuring          the third node, download traffic application is classified into
channel condition. This scheme with its used parameters is               nrtPS service class. For third node, we define traffic by a
showed in Fig. 8.                                                        variable rate in which packet size has a exponential
                                                                         distribution with 86kbytes mean and Poisson inter-arrival
                                                                         time with 500 milliseconds mean. All traffics transfer in
               IV. PERFORMANCE EVALUATION                                TCP/IP and the header of these protocols must be measured in
  A. Simulation Environment and Parameters                               calculations. Channel model for pathloass is based on Erceg
   For performance evaluation we developed wimax layers in               model and for multipath modeling we use ITU pedestrian
OPNET simulator [16]. Scheduling for polling based service               model. Terrain is mostly flat with light tree densities. At first
and BE service worked in wimax_bs_control process model.                 we evaluate WiMAX average delay for and in our proposed
At first we compared our proposed scheme with UEPS, PF,                  scheme we consider 2 parameters first with variable Si
OFDM frame-based PF [17], M-LWDF and MAX CINR in a                       according to the above mentioned equation and second with a
simple scenario and then, at second case, we studied                     fixed Si equal to 1000. Fig. 10. shows that average delay for
admission control impact on our scheduling scheme. For our               our scheme is lower than the other schemes even with a fixed
simulation, we used TDD mode with 20MHz OFDMA access                     time window. By holding time window Si fixed or variable
technology and PUSC permutation in uplink and downlink.                  average delay does not change, but with a variable time
Frame duration are 5 msec with 48 OFDM symbols in each                   window, at the beginning, causes lower delay. Fig. 11.
farme, 12 for uplink and 36 OFDM symbols for downlink.                   illustrated average throughput of system which consider with
TTG is 106 µsec and RTG is 60 µsec and numbers of data                   and without             ,         . Better measured channel
subcarriers in uplink are 1440 and in downlink are 1120 of               condition causes more average throughput, because error in bit
2048 subcarriers. Transmitted power for Base station and user            rate decreases in transmission occurrence.
station is equal to 0.5 watt. We can evaluate scheduling
scheme when there are few resources, and users compete for




                                                                   238                              http://sites.google.com/site/ijcsis/
                                                                                                    ISSN 1947-5500
                                                        (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                                        Vol. 9, No. 2, February 2011




           Fig. 10. Average delay in WiMAX system                                     Fig. 12. Packet loss ratio in WiMAX


                                                                        A. Impact of Admission Control in scheduling
                                                                        In this scenario we investigate Admission control unit
                                                                      impact on our scheduling scheme. Channel condition, OFDM
                                                                      symbol duration, number of subcarriers, frequency, bandwidth
                                                                      and permutation are the same as in before section. QAM64
                                                                      modulation with ¾ coding rate is used for all users whose
                                                                      distance from base station is the same according to Fig. 13.

                                                                          Three types of services have been defined for users, ertPS,
                                                                      rtPS and nrtPS. All users have constant data rates for ertPS
                                                                      defining 16kbps with 100 bytes packet size. Data rate for rtPS
                                                                      are 56kbps with 400 bytes packet size and in nrtPS data rate is
                                                                      120kbps with 800 bytes packet size. Users randomly get one
               Fig. 11. Average throughput in Mbps                    of the services in a network. After limitation made on
                                                                      resources by admission control, maximum users that can be
   Obviously MAX-CINR have better throughput than                     supported with this configuration are 72 users. We force
proposed scheduling, because user with a better channel               admission control to cross the line and accept 80. We have
condition can get a high priority without any QoS guarantee.          investigated throughput of the system in different points with
Variable time window in average data rate calculation causes          distinctive number of users, as shown in Fig. 14.
better throughput even if we do not consider         ,      .
For nrtPS, buffer management has an important role to prevent
overflow and increase throughput especially when resources
for scheduling are limited and only high priorities can get
permission.
   Packet loss ratio in our scheme is approximately equal to
MAX-CINR as considered in Fig. 12. In Fig. 11 throughput is
lower than MAX-CINR which shows more error bit rate. For
using different coding scheme as CRC and convolution coding
adaptively with channel condition some missed bits
recovered and consequently packet loss ratio has no
expansion. MAX-CINR for selecting users with appropriate
channel states, error correction schemes are not required any
more, but in our scheduling, error correction schemes are
helpful and hinders loss of packets as in MAX-CINR.
                                                                            Fig. 13. Admission control Analysis scheme in OPNET




                                                                239                             http://sites.google.com/site/ijcsis/
                                                                                                ISSN 1947-5500
                                                                  (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                                                  Vol. 9, No. 2, February 2011
                                                                                                       REFERENCES
                                                                            [1] IEEE Std 802.16™-2009, “Part 16: Air Interface for Broadband
                                                                                 Wireless Access Systems”, May. 2009, 2082 pp.
                                                                            [2] C. So-In, R. Jain, and A. Al-Tamimi, “Scheduling in
                                                                                 IEEE802.16e Mobile WiMAX Networks: Key Issues and a
                                                                                 Survey,” IEEE Journal on Selected Areas in Communications,
                                                                                 Vol. 27, No. 2, Feb. 2009.
                                                                            [3] T. Ohseki, M. Morita, and T. Inoue, “Burst Construction and
                                                                                 Packet Mapping Scheme for OFDMA Downlinks in IEEE
                                                                                 802.16 Systems,” in Proc. IEEE Global Telecommunications
                                                                                 Conf., Washington, DC, 2007, pp. 4307-4311.
                                                                            [4] Y. Ben-Shimol, I. Kitroser, and Y. Dinitz, “Two-dimensional
                                                                                 mapping for wireless OFDMA systems,” IEEE Trans.
                                                                                 Broadcast., vol. 52, pp.388-396, Sept. 2006.
                                                                            [5] Bacioccola, C. Cicconetti, L. Lenzini, E. A. M. E. Mingozzi, and
                                                                                 A.A. E. A. Erta, “A downlink data region allocation algorithm
                                                                                 for IEEE802.16e OFDMA,” in Proc. 6th Int. Conf. Information,
                                                                                 Communications& Signal Processing., Singapore, 2007, pp. 1-5.
                                                                            [6] F. YIN, and G. Pujolle, “Performance Evaluation of Polling
                                                                                 Mechanisms in IEEE 802.16 Networks” in Proc. IEEE Int. 14th
                                                                                 Asia-Pacific Conf., Tokyo, 2008, pp. 1-5.
                                                                            [7] A. Sayenko, O. Alanen, J. Karhula and T. Hamaainen,
     Fig. 14. Average throughput for different class of service                  “Scheduling solution for the IEEE 802.16 base station,” Int. J.
After increasing 70 users to 80 users, throughput badly                          Computer and Telecommunications Networking, vol. 52, pp. 96-
                                                                                 115, Jan. 2008.
degrades because admission control accepted more than
                                                                            [8] K. Wongthavarawat, and A. Ganz, “Packet scheduling for QoS
system’s supported capacity and scheduler cannot properly                        support in IEEE 802.16 broadband wireless access systems,” in
allocate. It shows a high relationship between scheduler and                     Proc. Int. J. Communication System, vol. 16, pp. 81-96, Feb.
admission control unit. Simulation results are for 15 minutes.                   2003.
                                                                            [9] D. Niyato, and E. Hossain, “Queue-aware uplink bandwidth
                                                                                 allocation for polling services in 802.16 broadband wireless
                                                                                 networks,” in Proc. IEEE Global Telecommunications Conf., St.
        V. CONCLUSIONS AND FUTURE STUDY ISSUES
                                                                                 Louis, MO, 2005, vol. 6, pp. 5-9.
   PF and OFPF are suitable for BE service class but cannot                 [10] J. Chen, W.Jiao and H. Wang, “A service flow management
guarantee QoS. They have high delay and low throughput                           strategy for IEEE 802.16 Broadband Wireless Access Systems
rather than QoS guarantee algorithms. Since M-LWDF and                           in TDD mode,” in Proc. IEEE Int. Conf. Communications.,
                                                                                 Seoul, korea, 2005, vol. 5, pp. 3422-3426.
UEPS do not consider buffer state and channel condition
                                                                            [11] F. Hou, P. Ho, X. Shen, and A.Chen, “A Novel QoS Scheduling
exactly they have lower performance to proposed scheme.                          Scheme in IEEE 802.16 Networks,” in Proc. IEEE Wireless
Original UEPS uses one graph to represent emergency but in                       Communication and Networkin Conf., Hong Kong, 2007, pp.
our scheme to remove inter-class scheduling in delay and                         2457-2462.
buffer state functions we use 2 figures. We have also much                  [12] M. Andrews, K. Kumaran, K. Ramanan, A. Stolyar, P. Whiting,
better considered channel condition by CINR and RSSI, in                         and R. Vijayakumar, “Providing quality of service over a shared
                                                                                 wireless link,” IEEE communications Mag., vol. 39, pp. 150-
making a decision. Simulation results show better delay and
                                                                                 154, Feb. 2001.
high throughput for proposed scheme with the near to same                   [13] P. Parag, S. Bhashyam, and R. Aravind, “A subcarrier allocation
packet loss ratio equal to MAX-CINR which is the highest in                      algorithm for OFDMA using buffer and channel state
throughput. Admission control evaluated as important unit in                     information,” in Proc. IEEE Vehicular Technology Conf.,
good performance of scheduling and almost estimate channel                       Dallas, TX, 2005, vol. 1, pp. 622-625.
capacity more precisely.                                                    [14] V. Singh, and V. Sharma, “Efficient and Fair Scheduling of
                                                                                 Uplink and Downlink in IEEE 802.16 OFDMA Networks,” in
   Power management mechanism which examines Idle and
                                                                                 Proc. IEEE Wireless Communication and Networking Conf., Las
sleep state in increasing delay is not considered in our work.                   Vegas, NV, 2006, vol. 2, pp. 984-990.
In real world power saving and battery consideration is                     [15] Y. J. Zhang, and S. C. Liew, “ Link-adaptive largest-weighted-
necessary. Fairness index is another aspect that can be studied                  throughput packet scheduling for real-time traffics in wireless
in this kind of works to properly compare with PF-Based                          OFDM networks,” in Proc. IEEE Global Telecommunications
scheme. Other great approaches are smart antenna and MIMO-                       Conf., St. Louis, MO, 2005, vol. 5, pp. 5-9.
                                                                            [16] http://www.opnet.com/solutions/network_rd/modeler.html
OFDM which are highly used in 4G wireless system and make
                                                                            [17] N. Ruangchaijatupon, and Y. Ji, “Simple Proporional Fairness
scheduling harder in resource allocation. A lot of matters                       Scheduling for OFDMA Frame-Based Wireless Systems,” in
could be regarded in resource allocation to make it better but                   Proc. IEEE Wireless Communication and Networking Conf., Las
they may also create more complexity.                                            Vegas, NV, 2008, pp. 1593-1597.
                                                                             




                                                                      240                               http://sites.google.com/site/ijcsis/
                                                                                                        ISSN 1947-5500

				
DOCUMENT INFO
Shared By:
Stats:
views:211
posted:3/12/2011
language:English
pages:8