In this section, we evaluate the performance of the proposed scheme with a set of simulation experiments. We consider a single-cell IEEE 802.16e network with a BS and a group of MSs dropped in the cell uniformly. All the MSs move with a walk speed, and the detected SNRs would be the basis of using corresponding MCS. We partition this section into three parts, the first part is showing the performance of two different prediction algorithm and fixed frame-ratio with the same mapping scheme, the second part is showing the performance of different mapping schemes with the same prediction algorithm and sorting algorithm and the third part is showing the performance of schemes with and without sorting algorithm. The values of parameters we used are reported in Table 2.
Table 2.All the parameters used in the simulation experiments
25
The Figure 5 shows the total profit we get when using different prediction algorithm and the corresponding values is shown below the figure. We can clearly observe that the algorithm 1 is better than the algorithm 2, and the two prediction algorithms are better than the fixed one.
The reason is that we allocate more resource to the side which needs more so that we can transmit more requests than the fixed frame-ratio. And the first algorithm is a little bit better than the second because we allocate the resources to the requests with high priority and profit in the first algorithm, it may cause the problem of allocating all the resources to one side.
There is no that problem in the algorithm 2, but causes a little bit worse efficiency.
Figure 5.The total profits with two prediction algorithms and fixed frame-ratio
26
In Figure 6, we show the average delay of two real-time services in three ways. It’s clearly that the ert-VR service has less delay than the rt-VR service no matter which kind of method we use because the ert-VR service has the higher basic priority and profit values than the rt-VR service, the scheduler prefer to serve the ert-VR service first. For one specific service, no matter which kind of the two, the average delay in algorithm 1 is the least, it indicates that the frame utilization in the prediction algorithm 1 is the best.
Figure 6.The average delays with two prediction algorithms and fixed frame-ratio
27
The numbers of dropped requests in the three methods is shown in figure 7. Because of the higher priority and profit, the scheduler prefer to serve the ert-VR service first, it causes that the rt-VR requests are easier to be dropped than the ert-VR requests. For one specific service, the performance is the best in the algorithm 1; it shows that the frame utilization in algorithm 1 is the best in the three once again.
Figure 7.The numbers of dropped requests in three methods
28
The next two figures show the other two QoS requirements performance in the three methods and the results met our expectations. It shows the variance of ert-VR service when the number of MSs is forty in figure 8 and the guaranteed rate of three kinds of QoS services in figure 9. The algorithm 1 has better performance than the other two including not only lower variance of ert-VR service but also higher guaranteed rate of nrt-VR service. It supports our point of view: the prediction algorithm 1 is better than the algorithm 2, and the fixed frame-ratio has the worst performance in every part.
Figure 8.The variance of ert-VR service when the number of MSs is 40
29
Figure 9.The guaranteed rate of three types of QoS services when the number of MSs is 40
30
The second part we show the performance with different mapping schemes including
“eOCSA”, “AHBM” and our proposed “eAHBM”. In figure 10, we can clearly observe that the eOCSA has the worst performance and its total profit start to decrease when the number of MSs is 35 because of its inefficient mapping scheme, the data mapper can’t serve the requests with higher profit but with higher priority when traffic load is heavy. Meanwhile, our proposed mapping scheme “eAHBM” is a little bit better than AHBM when the traffic load become heavier.
Figure 10.The total profits with three different mapping schemes
31
The results which are shown in figure 11, 12, 13 and 14 supports our view: eOCSA has the worst performance in every part and our proposed mapping algorithm is a little bit better than AHBM. In figure 11, the average delays of requests in eOCSA are much higher than the other two schemes, so does the number of dropped requests and the variance which are shown in figure 12 and 13. And in the figure 14, the guaranteed rate of the three services in eOCSA is much lower than the values we set before the simulation, indicates that the eOCSA is not a good mapping scheme for protecting QoS requirements.
Figure 11.The average delays with three different mapping schemes
32
Figure 12.The numbers of dropped requests in three mapping schemes
Figure 13.The variance of ert-VR service when the number of MSs is 40
33
Figure 14.The guaranteed rate of three types of QoS services when the number of MSs is 40
34
In the third part, we show the performance with and without sorting algorithm to indicate that deciding the serving order is an important part in our proposed scheme. The figure 15 shows the result we expected, it has bad performance if we serve the biggest request first.
When traffic load become heavier, the biggest request, which may be due to the bad channel conditions caused lower profit, is to be served first so that the total profits start to decrease when the number of MSs is more than twenty-five. In figure 16 and 17, we can see that the data mapper without our sorting algorithm can’t protect the requests with higher priority, it has almost no difference of the two QoS indices, average delay and number of dropped requests, between the ert-VR service and rt-VR service.
Figure 15.The total profits with and without sorting algorithm
35
Figure 16.The average delays with and without sorting algorithm
The figure 18 and 19 shows the results as we think. It is worth to be mentioned that the guaranteed rate of nrt-VR is higher than the other two services because the nrt-VR service is insensitive to delays so that the requests could stay in the queue for a long time and wait to be served, but the ert-VR service and rt-VR service have the opposite position. They have delay bound restriction and the data mapper doesn’t take this restriction into consideration to server them first so that they are easily to be dropped, caused lower guaranteed rate.
36
Figure 17.The numbers of dropped requests with and without sorting algorithm
Figure 18.The variance of ert-VR service when the number of MSs is 40
37
Figure 19.The guaranteed rate of three types of QoS services when the number of MSs is 40
38
Chapter 5.
Conclusion
In this thesis, we present a resource allocation algorithm for taking care of both QoS requirements for every request and frame utilization. There are four parts in our algorithm, the first part is to define the priority and profit for every request so that we can decide the serving order based on these two values; then sort all the requests according to the two indices mentioned in part one is in our second part; the third part is to make the decision of the ratio between downlink and uplink sub-frames depend on the queue status, then an improved mapping scheme is in our fourth part.
The simulation results confirm that the prediction algorithm is working for raising the frame utilization because the traffic load in downlink part and uplink change over time so that allocating more resource to the side with more traffic data is quite a reasonable way. The simulation results also show that our algorithm has better performance for protecting the QoS requirements, for example, higher profit, less number of dropped requests, lower average delay and etc.
39
References
[1] IEEE Std 802.16-2009, IEEE Standard for Local and metropolitan area networks – Part 16: Air Interface for Broadband Wireless Access Systems
[2] S. T. Sheu, M. H. Tsai, T. Y. Tsai, and Y. H. Tsai, “Condensed Downlink MAP Structures for IEEE 802.16e Wireless Metropolitan Area Networks (MANs)” in Proc.
IEEE VTC’10, pp1-5, May 2010
[3] C. So-In, R. Jain and A. A. Tamimi, “OCSA: An algorithm for burst mapping in IEEE 802.16e mobile WiMAX networks1,2, ” in Proc. IEEE APCC’09, pp.52-58, Oct. 2009.
[4] C. So-In, R. Jain and A. A. Tamimi, “eOCSA: An algorithm for burst mapping with strict QoS requirements in IEEE 802.16e Mobile WiMAX networks, ” in Proc. IEEE 2nd IFIP’09, pp.1-5, Dec. 2009
[5] C. Desset, E. B. de Lima Filho, and G. Lenoir, “WiMAX Downlink OFDMA Burst Placement for Optimized Receiver Duty-Cycling” in Proc. IEEE ICC’07, pp.5149-5154, Jun. 2007.
[6] A. Erta, C. Cicconetti, and L. Lenzini, “A Downlink Data Region Allocation Algorithm for IEEE 802.16e OFDMA” in Proc. IEEE ICICS’07, pp.1-5, Dec. 2007.
[7] X. Jin, J. H. Zhou, J. L. Hu, J. L. Shi, Y. Sun, and E. Dutkiewicz, “An Efficient Downlink Data Mapping Algorithm for IEEE 802.16e OFDMA Systems” in Proc. IEEE GLOBECOM’08, pp.1-5, Dec. 2008.
[8] C. Cicconetti, L. Lenzini, A. Lodi, S. Martello, E. Mingozzi and M. Monaci,
“Efficient Two-dimensional Data Allocation in IEEE 802.16 OFDMA, ” in Proc. IEEE INFOCOM’10, pp.1-9, Mar. 2010
[9] P. G. Sarigiannidis, G. I. Papadimitriou, P. Nicopolitidis, M. S. Obaidat, and A. S.
Pomportsis, “A Novel Adaptive Mapping Scheme for IEEE 802.16 Mobile Downlink Framing” in Proc. IEEE GLOBECOM’10, pp.1-5, Dec. 2010
[10] P. G. Sarigiannidis, M. D. Louta, and P. Chatzimisions, “A Novel Fair Mapping Scheme for IEEE 802.16 Downlink Sub-Frame” in Proc. IEEE ICUMT’10, pp.763-768, Oct. 2010
[11] Y. Ben-Shimol, I. Kitroser, and Y. Dinitz, “Two-Dimensional Mapping for Wireless OFDMA System” in Proc. IEEE Transactions on’06, pp.388-396, Sep. 2006
[12] S. Kuei-Ping, C. Hung-Chang, C. Chi-Tao, and H. Tsung-Han, “Channel-Aware Subchannel Renumbering and Downlink Burst Allocation for IEEE 802.16 OFDMA Systems” in Proc. IEEE WCNC’10, pp.1-6, Apr. 2010
[13] R. Cohen, and L. Katzir, “Computational Analysis and Efficient Algorithms for Micro and Macro OFDMA Scheduling” in Proc. IEEE INFOCOM’08, pp.511-519, Apr. 2008
100