An Enhanced EDD QoS Scheduler for IEEE 802.11e WLAN - Advances in Computational Science and Engineering - page 52

since using QoS Control field provides exact queue length information, the delay
accumulated by bursty trac can be decreased.

2nd Step : SI allocation
The SI allocation is as follows. As observed in figure 3, for QST A i , data arriving
within the interval [ t i ( x ), t i ( x + 1)] can be sent no earlier than TXOP i ( x +2).
Hence, in order to keep up with delay bound, The worst case is assumed that
theSIandMSIisthesameand TXOP i ( x +2) = MTD i , Hence the MSI of
ARROW is as follows.

D i − MTD i
2

D i ≥ 2 MSI i + MTD i ⇐⇒ MSI i ≥ (7)

3 Proposed Scheduling Algorithm

In this section, we present a new algorithm that makes delay, jitter and packet
loss probability decrease than the existing algorithms using the characteristics
of trac.
As aforementioned,the reference scheduler presented in IEEE 802.11e is easy
to implement and simple. It has low eciency like considerable packet loss prob-
ability in VBR since SI and TXOP have fixed values. SETT-EDD, however, has
a performance superior to reference scheduler in VBR by using variable TXOP
and SI. But there exists accumulated packet by bursty trac since TXOP al-
location of SETT-EDD considers only average TXOP transmission and do not
consider exactly bursty trac that the characteristics cause delay increase. Also
SETT-EDD has a problem which remaining packets in current SI waits unnec-
essarily because of fixed mSI and MSI. In order to solve the problem caused by
bursty trac, ARROW was proposed to consider queue size because the packets
accumulated by bursty trac wait in queue of QSTA. However, it is dicult for
ARROW to empty queue because the queue size information is about previous
transmission, and it does not overcome unnecessary wait in SETT-EDD.
Basically, the proposed algorithm is based on EDD. The proposed algorithm
also uses the queue length information like ARROW. In addition the algorithm
estimates the number of packets arrived after the end of the previous transmis-
sion. Then the algorithm calculates TXOP which is just large enough to clear
the queue when the current transmission completes. To reduce the average delay,
when the buffer is not empty after the current transmission completes, the next
SI begins earlier, which can be achieved by changing the value of mSI and MSI.
The operation of proposed algorithm can be divided into TXOP and SI allo-
cation, too.

3.1 TXOP Allocation

TXOP allocation algorithm considers the packets generated from t pre when
QSTA was polled at previous SI to the current time and the packets remaining in
the queue at previous SI. To transmit the packets generated during SI, it needs
TXOP avg which is calculated as follows.