Information Technology Reference
In-Depth Information
not restrict the amount of time which a polled node can use, which makes it
dicult to guarantee QoS. Futhermore, beacon intervals can not be precisely
controlled. In HCCA, AP accepts a flow only if it can guarantee QoS of both the
existing and the arriving flows. It then controls the duration that a polled station
can use by assigning a time limit in the poll message. The duration is called
TXOP(Transmission Opportunities) [2]-[9]. In HCCA, HC(Hybrid Coordinator)
located at AP allocates TXOPs through a polling mechanism. The operation of
HCCA is influenced by TSPEC(Trac Specification) of flows. Hence a scheduler
is required to achieve this operation eciently in HCCA.
In the IEEE 802.11e standard, a reference scheduler was proposed [2]. The
reference scheduler is simple to implement, however, it is ecient only for the
CBR(Constant Bit Rate) trac since it assigns the same SI(Service Interval) to
all stations. For VBR(Variable Bit Rate) trac, assigning the same SI for all
flows results in the bad eciency in which the average delay, packet loss and
jitter are increased drastically as the packet sizes are varied more [7]-[9].
Many schedulers have been proposed to improve the QoS performance in
terms of packet loss, delay and jitter for VBR trac [7]-[9]. For example, SETT-
EDD(Scheduling based on Estimated Transmission Time-Earliest Due Date) was
proposed to decrease the packet loss and the average delay by allocating TXOP
and SI adaptively using a token bucket and the EDD algorithm [8]. In the sched-
uler, although the average delay and the packet loss are decreased compared with
the reference scheduler, when bursty trac is applied, it does not handle the traf-
fic properly since the token bucket guarantees the average transmission only: the
amount of accumulated packets in queue becomes larger as the trac is more
bursty [6]. If the scheduler takes the queue length of the flow into consideration,
it may perform better. For example, ARROW(Adaptive Resource Reservation
Over WLANs) allocates TXOP using the queue length information of a flow,
thus it can reduce the accumulated packets in the queue [9]. However, the queue
length information delivered in the packet is the queue length of the previous
frame transmission Thus, it assigns TXOP which is just enough to transmit the
frames remained and the frames arrived after the previous transmission: it is
dicult to empty the queue. Accordingly ARROW has the same problem with
SETT-EDD, too.
In this paper, we propose a new QoS scheduler for IEEE 802.11e WLAN,
which can support not only CBR but also VBR with minimal queueing delay
and jitter. Our scheduler allocates TXOP and SI adaptively based on TSPEC
of the flow. The TXOP allocation of proposed scheduler considers not only the
queue size but also the mean date rate of a flow. Then it estimates TXOP which
is just large enough to empty the queue. The proposed scheduler changes SI
dynamically so that it can reduce the transmission delay also: when the buffer
was not empty after previous transmission, shorter SI will results in reduced
delay.
This paper is organized as follows. Section 2 explains the basic operation of
IEEE 802.11e and some QoS schedulers. In section 3 we present the proposed
Search WWH ::
Custom Search