Chapter 4.
Experimental Result
The PHY and MAC parameters and all related information used in simulations are shown in Table I. Note that the sizes of QoS-ACK and QoS-Poll in the table only include the sizes of MAC header and CRC overhead. We assume that the minimum physical rate is 2Mbps and
t
PLCP is reduced to 96 μs. It is assumed thatSI
m= 10
ms. Two types of TSs, with characteristics and QoS requirements shown in Table II, are considered in simulations. Type I and Type II TSs are lecture room cam and interactive video, respectively. We assume that 90% of the bandwidth is used by HCCA, i.e.,ρ
hcca=0.9.
Chapter 4. Experimental Result
Table II. TSPECs for two different types of traffic flows.
Traffic characteristics and
QoS requirements Type I Type II Maximum Service Interval 20 ms 40 ms
Delay Bound 40 ms 80 ms
Mean Data Rate 42 Kbps 246 Kbps Nominal MSDU size 211 bytes 1232 bytes Scheduled Service Interval 20 ms
Chapter 4. Experimental Result
In the first experiment, we compare the performances of the sample scheduler, the PRO-HCCA scheduler, the SPRO-HCCA scheduler, and the reduced-overhead SPRO-HCCA (RO-SPRO-HCCA) scheduler with real traces [17]. There are three QSTAs.
QSTA
1,QSTA
2, andQSTA
3 that are attached with two Type I TSs, one Type I TS and another Type II TS, and two Type II TSs, respectively. As a result, the scheduled SI is set to 20 ms= 2
SI
m. The polling order isQSTA
1, thenQSTA
2, followed byQSTA
3. In the PRO-HCCA scheme, for the two TSs attached toQSTA
2, the Type I TS is polled before the Type II TS. Because of the 80 ms delay bound, Type II TSs attached toQSTA
3 are polled once every two SIs in the RO-SPRO-HCCA scheme.Figs. 8(a)-8(d) show the cumulative distribution functions (CDFs) of delay for TSs attached to various QSTAs. Note that, as illustrated in Fig.
8(a), the RO-SPRO-HCCA scheme is the same as the SPRO-HCCA scheme for the TSs attached to
QSTA
1. One can see in Fig. 8(a) that packets of TSs attached toQSTA
1 experience roughly the same delay under the SPRO-HCCA scheme and the PRO-HCCA scheme. The curves shown in Fig. 8(b) reveal that packets of Type I TS attached toQSTA
2 experience smaller delay under the SPRO-HCCA scheme than the PRO-HCCA scheme. The reason is that packets of Type I TS have smaller delay bound than packets of Type II TS and, therefore, under the SPRO-HCCA scheme, may use the bandwidth allocated to packets of TypeChapter 4. Experimental Result
II TS that can be kept for more than two SIs. Because of this, packets of the Type II TS experience larger delay under the SPRO-HCCA scheme than the PRO-HCCA scheme, as can be seen in Fig. 8(c). Note that the RO-SPRO-HCCA scheme is different from the SPRO-HCCA scheme for
QSTA
2. There are four entries for the partition list maintained forQSTA
2 under the SPRO-HCCA scheme. However, under the RO-SPRO-HCCA scheme, there are only two entries for the partition list maintained forQSTA
2 and only the data that will violate their delay bounds if not served in the next SI are reported to QAP. Packets experience more delay under the RO-SPRO-HCCA scheme than under the SPRO-HCCA scheme and the PRO-HCCA scheme. However, all packets meet their delay bounds.The sample scheduler obviously cannot meet QoS requirements.
Although the PRO-HCCA, SPRO-HCCA, and RO-SPRO-HCCA schedulers meet QoS requirements, their ratios of overhead transmission time to total transmission time are different. They are 34.92%, 31.40%, and 29.51% for the PRO-HCCA, SPRO-HCCA, and RO-SPRO-HCCA schemes, respectively.
Chapter 4. Experimental Result
Chapter 4. Experimental Result
Fig. 8. Performance comparison of various schedulers for three different QSTAs.
Chapter 4. Experimental Result
In the second experiment, we assume that there are one Type I TS and one Type II TS attached to each QSTA. Simulations are performed to determine the maximum number of QSTAs N that can be supported without violating delay bound requirements under the SPRO-HCCA scheme. The result is N =11. Therefore, we simulate a system which consists of 11 QSTAs under various scheduling schemes. QSTAs are polled one by one. In the PRO-HCCA scheme, Type I TS is polled before Type II TS attached to the same QSTA. Note that, similar to the situation of
QSTA
2 in the first experiment, the RO-SPRO-HCCA scheme is different from the SPRO-HCCA scheme for this experiment. Figs. 9(a) and 9(b) show, respectively, the CDFs of delay for packets of Type I TS and Type II TS attached to the eleventh QSTA. As one can see from the figures, the SPRO-HCCA scheme outperforms the PRO-HCCA scheme for both types of TSs. The phenomenon we observed forQSTA
2 in the first experiment, i.e., the bandwidth allocated to less urgent packets of Type II TS could be used by packets of Type I TS, does not appear in the second experiment.The reason is that almost all bandwidth are allocated to the most urgent packets that will be dropped if not served in the next SI. Some packets violate their delay bounds and are lost under the PRO-HCCA scheme because it suffers from higher overhead than the SPRO-HCCA scheme.
The packet loss probabilities due to violation of delay bounds are summarized in Table III. The low packet loss probabilities make the RO-SPRO-HCCA an attractive scheme for real systems.
Chapter 4. Experimental Result
Fig. 9. Performance comparison of various schedulers for eleven identical QSTAs.
Chapter 4. Experimental Result
Table III. Packet loss probabilities for various schedulers.
TS type of QSTA 11 Scheduler
Type I TS Type II TS
PRO-HCCA 7.11% 0.2%
SPRO-HCCA 0% 0%
RO-SPRO-HCCA 0.15% 0.13%
Sample Scheduler 53.92% 81.04%
Chapter 5. Conclusion
Chapter 5.
Conclusion
We have presented in this paper a per-QSTA based scalable TXOP allocation scheme for HCCA to guarantee QoS for VBR traffic in WLANs.
The scheme is modified to allow different polling periods for different traffic streams. Computer simulations with real traces show that our proposed schemes meet QoS requirements. Besides, according to simulation results, our proposed schemes utilize bandwidth more efficiently than the sample scheduler and the PRO-HCCA scheduler. An advantage of the proposed schemes is that they do not require traffic models. It suffices to know the mean arrival rate of each traffic stream. Simplicity and robustness make the proposed schemes good candidates of the TXOP allocation scheme for HCCA to guarantee QoS for variable bit rate traffic streams.