Accounting for Uplink Traffic in the Design of Schedulers for Multiple Traffic Classes in OFDMA Networks by editorijettcs


More Info
									   International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)
       Web Site: Email:,
Volume 1, Issue 2, July – August 2012                                          ISSN 2278-6856

      Accounting for Uplink Traffic in the Design
      of Schedulers for Multiple Traffic Classes in
                  OFDMA Networks
                                                        A Kazmierczak
                                                  Computer Information Systems
                                              Northwest Arkansas Community College
                                                         One College Dr
                                                       Bentonville, AR 72712

Abstract: The research literature in wireless networks           structure using Orthogonal Frequency Division Multiple
contains many examples of proposed MAC protocols. Every          Access (OFDMA). In this structure, multiple
such proposed protocol thoroughly addresses downlink traffic     communication channels exist. As a way to efficiently
from the Base Station (BS) to the Mobile Station (MS). Few,      and reliably use these channels, the IEEE 802.16 standard
if any, do more than mention uplink traffic. In this paper, we   includes the band Adaptive Modulation and Coding
look at two MAC protocols, Multiclass MLWDF and Joint            (ACM) scheme. The possibility of assigning multiple
Scheduler, from the perspective of uplink traffic in addition    frequency channels gives a scheduler the opportunity to
to downlink traffic.                                             exploit frequency diversity as well as multiuser diversity
Keywords: MAC protocol, downlink traffic, traffic                to maximize system performance. UEPS was proposed for
scheduler                                                        OFDMA systems, but works well in a single channel
1. INTRODUCTION                                                  There are two examples of single channel schedulers that
In recent years that has been a great deal of activity in the    can exploit channel variations and support multiple
area of research and standardization of mobile wireless          transmission rates; i) maximum channel to interference
networks. Some examples include the 3rd Generation               ratio (max C/I) scheduler [7], and ii) proportional fair
Partnership Project (3GPP) High Speed Downlink Packet            (PF) scheduler [8]. The max C/I scheduler always chooses
Access (HSPA) [2] and systems based on IEEE 802.16               the user whose channel rate is largest. Though it achieves
[3], such as Worldwide Interoperability for Microwave            maximum system throughput, many users whose channel
Access (WiMAX) [4] and Wireless Broadband (WiBro)                state is not good may face starvation. The PF scheduler
[5]. Fourth generation wireless devices are being                uses the ratio of current channel rate to average allocated
introduced into the market even now.                             rate for each user. It does provide proportional fairness
Downloading multimedia files such as music and video or          among users. As good as these schedulers are, they do not
browsing the Internet through a mobile device is no              support specific QoS parameters like minimum allowable
longer a rare occasion. In fact, it is becoming                  delay and maximum throughput.
commonplace. With 4th generation wireless systems being          There have been schedulers proposed that do support
deployed, there will be a growing number of applications         specific QoS parameters. The MLWDF scheduler [9] and
will be far more diverse and the demand on data rates will       exponential rule scheduler [10] consider maximum
skyrocket. Current mobile phones support only one                allowable delay and instantaneous channel rate,
application at a time. However, with the latest generation,      respectively. Both schedulers are throughput optimal and
multitasking will become more popular. Thus, mobile              keep queues stable. MLWDF uses the head of line packets
devices will need to support multiple connections with           waiting time or the total queue length as scheduling
multiple traffic classes for each user.                          metrics.
As part of supporting Quality of Service (QoS),                  In [1], the author’s extend the MLWDF scheduler to
scheduling is a fundamental function. In designing QoS           multiple channels and multiple classes of traffic. Most
schedulers, various QoS parameters for multiple traffic          schedulers that have been proposed in the literature have
classes need to be considered. In [6], Urgency and               limited themselves with handling downlink packet flow.
Efficiency based Packet Scheduling (UEPS) was proposed           Uplink packet flow is either passed over all together or is
to support both non real time (NRT) traffic and real time        given only cursory mention. If users are going to continue
(RT) traffic. UEPS served NRT packets until RT packets           to flock to the wireless environment, they are going to
approached their deadline when RT packets were                   expect two way communications. In this paper, we take
scheduled with higher priority. It is not always effective       the work in [1] on the MLWDF scheduler and their
to give NRT packets priority RT packets                          proposed Joint Scheduler and expand both schedulers to
One of the promising technologies for next generation            enable uplink communication.
wireless systems is the wideband multicarrier frame              This paper is organized as follows. The proposed
                                                                 schedulers are resented in section II. The system model is

Volume 1, Issue 2 July-August 2012                                                                               Page 238
   International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)
       Web Site: Email:,
Volume 1, Issue 2, July – August 2012                                          ISSN 2278-6856

described briefly in section III. Section 4 should contain      Originally MLWDF is a single channel scheduler that
simulation results, while section IV concludes the paper.       satisfies stability and throughput optimality [8] and is
                                                                well known for satisfying the delay requirements of QoS
                                                                users. Multiclass MLWDF works fine for downlink
2. PROPOSED SCHEDULERS                                          traffic. Since MLWDF was originally a single channel
                                                                scheduler, we propose to use the original MLWDF
Two schedulers are considered: i) one for multiple traffic
                                                                algorithm for scheduling uplink traffic. Since the
classes, and II) one for multiple frequency channels.
                                                                multiclass MLWDF is used strictly for downlink traffic, it
   2.1 Scheduler Structures                                     needs to be slightly modified to account for uplink traffic
In our architecture, the Base Station (BS)contains the          also. Figure 1 shows the multiclass MLWDF algorithm
status information of all queues and performs the actual        modified to account for multiple classes of traffic from the
scheduling. Each Mobile Station (MS) sends the BS its           MS to the BS.
channel quality information through the feedback                At each scheduling instance
channel.       Since    this    paper   addresses     uplink      {
                                                                      For j = 1 to N
communication, we propose that the MS also use the
                                                                      { update CR(j), QLB(j) and WT(j)
feedback channel to send additional information to
                                                                      AR(j) = w * AR(j) + (1-w) * SR(j)
include: i) MS identification, ii) queue information of BE            }
traffic, and iii) bandwidth request allocation for specific           QoS_schedule = 0
QoS class.                                                            For j = 1 to N
Let M be the overall number of traffic classes supported              { if (QLQ(j) > 0 QoS_schedule = 1}
by the system. Let N be the number of traffic classes that            If ( QoS_schedule> 0 )
can be generated by the mobile devices. We assume that                { SM = argmaxj (CR(j)/AR(j)) * (WT(j)/MD(j)) }
N < M. For downlink communication, the BS has a                       Else
separate queue for each of the traffic classes and each               { SM = argmaxj (CR(j)/AR(j)) * QLB(j) }
user. For uplink communication, the BS has a separate                 For j = 1 to N
queue for each traffic class and each user.                           { if( SM = j)
                                                                          { DAAR(j) = WD * DAAR(j) + (1-WD) * CR(j)
For downlink, traffic classes are prioritized so that class I
                                                                          UAAR(j) – WU * UAAR(j) }
has a higher priority than class j ( 1<= I <= j <= M ). QoS           }
parameters are defined for each traffic class that has its            Else {
own scheduler. In addition, each class has an indicator to                DAAR(j) = WD * DAAR(j)
represent the urgency of a packet. At each scheduling                     UAAR(j) = WU * UAAR(j) } }
instance, the scheduler checks the class priority and the
urgency of each packet. Within a class, an urgent packet        Variables
will be transmitted first. The lowest priority is Best Effort   DAAR(j): moving average of MS j’s allocated downlink
(BE) traffic. Barring higher priority traffic, BE traffic is    channel rate (bits)
scheduled by default.                                           UAAR(j): moving average of MS j’s allocated uplink
For uplink traffic, the same operation applies as for           channel rate (bits)
downlink.                                                       DAAR(j) = KD * AAR(j): allocated rate downlink
In a system with multiple frequency channels available          UAAR(j) = KU * AAR(j): allocated rate uplink
for downlink transmission, each channel can be                  AR(j): moving average of MS j’s CR(j) (bits)
considered separately and run a single channel scheduler.       CR(j): MS j’s current channel rate (bits)
Since the BS has channel state information, there is the        MD(j): Max allowed delay of MS j’s QoS class
possibility of using multiple frequency channels,               connection
frequency diversity, to achieve performance gains. In a         N: number of MS’s having QoS class connection
multiple frequency channel environment, each MS claims          QLB(j): MS j’s BE class queue length (bits)
a channel whose channel rate for the MS is highest. The         SM: index of the selected MS
BS selects a user and channel which the user claims             WT(j): waiting time of head of line packet in MS j’s class
according to the single channel scheduler.                      queue (ms)
For the uplink transmission, there are no multiple              WD: weighting factor for downlink
frequencies. Uplink from a BS usually is to a wired             WU: weighting factor for uplink
network with a channel rate much higher than the
wireless channel rate. The BS will checkthe class priority      Figure 1 Algorithm for Multiclass MLWDF for downlink
and urgency of a packet. In the same class, urgent packets                          and uplink traffic
are transmitted first. Barring higher priority traffic, BE      The QoS scheduler uses the head of line packets waiting
traffic is selected by default.                                 time as a scheduling metric whereas BE traffic scheduler
   2.2 MLWDF                                                    uses the queue length information for each user. As long
In [1], the authors propose to extend MLWDF for                 as a QoS class packet exists, the QoS traffic, there is a
multiple traffic classes and name it multiclass MLWDF.          low possibility of exploiting multiuser diversity which
Volume 1, Issue 2 July-August 2012                                                                               Page 239
      International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)
       Web Site: Email:,
Volume 1, Issue 2, July – August 2012                                          ISSN 2278-6856

leads to lower system throughput. Thus, we should               traffic, there simulations deal only with downlink traffic.
reconsider relaxing thus rule and use a Joint Scheduler         Our simulation results should present some interesting
(JS).                                                           results.
   2.3 Joint Scheduler
                                                                4. CONCLUSIONS
Service providers typically operate a wireless network
with a load that is much lower than the maximum                 Most MAC protocols for wireless networks that appear in
capacity. The QoS scheduler rule may be too strict when         the research literature treat mainly the scheduling of
the system load is low. A proposed JS algorithm modified        downlink traffic from the BS to the MS. The possibility of
to account for both uplink and downlink traffic is shown        uplink traffic is glossed over or not mentioned at all. In
in Figure 2. The JS scheduler treats QoS and BE traffic         this paper we modified several proposed wireless MAC
together in the event that the QoS load is low. Only if a       protocols and adjusted the MAC protocol to account for
QoS packet experiences delay longer than some x% of             uplink traffic as well as downlink traffic.
maximum allowable delay does it call specifically for the
QoS scheduler. By relaxing the QoS rule, the scheduler
becomes more flexible in choosing a user in such a way to       REFERENCE
take advantage of multiuser diversity. The BE scheduler           [1] W-H. Park, S. Cho, S. Bahk, “Scheduler Design for
counts the scheduling metric of the queue lengths of QoS              Multiple Traffic Classes in OFDMA Networks,
traffic as well as BE traffic.                                        Proc IEEE, 2006.
                                                                  [2] 3GPP TR 25.308 v5>7.0, “High Speed Downlink
   At each scheduling instance                                        Packet Access: Overall Description; Stage 2
                                                                      (Release 5)” Dec 2004
    For j = 1 to N
                                                                  [3] IEEEStd 802.16-2004, “Air Interface for Fixed
    { update CR(j), QLB(j), QLQ(j), WT(j)
        AR(j) = W * AR(j) + (1-W) * CR(j)
                                                                      Broadband Wireless Access Systems, Oct 2004.
    }                                                             [4] Ghosh, D. R. Wolter, J. G. Andrews, R. Chen,
    QoS_schedule = 0                                                  “Broadband          Wireless        Access      with
    For j = 1 to N                                                    WiMAX/802.16:             Current       Performance
    { if ( WT(j) > x * MD(j) )                                        Benchmarks and Future Potential,” IEEE
        (QoS_scheduler = 1                                            Communication Magazine, vol 43, no 2, pp 129-
    }                                                                 136, Feb 2005
            If(QoS_schedule> 0 )                                  [5] Telecommunications Technology Association 2.3
            { SM(j) = argmaxj(CR(j)/AR(j) ) * (Wt(j)/MD(j) )}         GHz Portable Internet Project Group (PG 302),
            Else {SM = argmaxj(CR(j)/AR(j)) *                         “2.3 GHz Portable Internet (WiBro) Overview”
  (QLQ(j)+QLB(j) ) }
                                                                      Aug                  2004,                 Available
            For j = 1 to N
            { if (SM = j)
                      {DAAR(j) = WD * DAAR(j) + (1-W) *           [6] S. Ryu, B. Ryu, H. Seo, M. Shin, “Urgency and
                      CR(j)                                           Efficiency Based Packet Scheduling Algorithms for
                      UAAR(j) = WU * UAAR(j) + (1-WU) *               OFDMA Wireless Systems,” Proc IEEE ICC, 2005
                      CR(j) }                                     [7] Knopp, P. A. Humblet, “Information Capacity and
               Else {DAAR(j) = WD * DAAR(j)                           Power Control in Single Cell Multiuser
              UAAR(j) = WU * UAAR(j) }                                Communications” Proc IEEE ICC, 1995
                                                                  [8] A. Julali, R. Padorani, R. Pankaj, “Data Throughput
Additional Variables                                                  of CDMA=HDR a High Efficiency High Data Rate
QLQ(j): MS j’s QoS class queue length (bits)                          Personal Communication Wireless System,” Proc
X: threshold parameter                                                IEEE Vehicular Technology Conference, 2000
Figure 2: Algorithm for Joint Scheduler with uplink and           [9] M. Andrews, et al, “Providing Quality of Service
                     downlink traffic                                 Over a Shared Wireless Link,” IEEE
                                                                      Communication Magazine, pp 150-154, Feb 2001
The BE scheduler uses the average allocated rate AAR(j)           [10] S. Shakkottai, A. Stolyer, “Scheduling Algorithms
of MS j as a metric instead of the average channel rate.              for a Mixture of Real Time and Non Real Time
Using this metric avoids the situation where a user with a            Data in HDR,” Proc ITC, pp 793-804, Sep 2001
bad channel or low traffic rate arrival rates could be            [11]The Network Simulator ns-2 (ns-2 1b&a) video
starved                                                               traffic generator based on TES (Transform Expand
                                                                      Sample) Model of MPEG 4 trace files, Available
3. SIMULATION RESULTS                                       
                                                                  [12]Internet2 Netflow Weekly usage reports for the
Simulation of the proposed schedulers is an area of
                                                                      Abilene               network.             Available
ongoing research. Since no other algorithms treat uplink

Volume 1, Issue 2 July-August 2012                                                                              Page 240

To top