Simulation Result and Conclusion

In the simulation, we compare the proposed MFL algorithm with FCFS and EDF. Figure 2-8 and Figure 2-9 show the service failure rate versus the mean data size if using these three algorithms, where the maximum tolerable delay T is equal to 60s in Figure 2-8 and the maximum tolerable delay T is equal to 180s in Figure 2-9. Mean data size denotes the average volume of the burst data that an OBU requests information service which is calculated according to equation (2.6).

We can see that MFL has smaller service failure rate than FCFS and EDF. The service failure rate is zero for three methods when the mean data size is below about 900k bytes. It is because that the system utilization is still light, thus almost all OBUs can be completely served within the maximum tolerable delay bound. The service failure rate for three methods starts to increase from zero when the mean data size is larger than around 900k bytes. It can be seen that MFL has the lowest service failure rate while FCFS has the largest. There are two reasons for this result. The first one is that MFL is the best policy to achieve the minimum system handoff rate. OBUs will have lower probability to become service failure.

Figure 2-8. Service Failure Rate for T =60s

Figure 2-9. Service Failure Rate for T =180s

This result will be showed later. And the other one is that MFL will let OBUs, which have

longer queueing delay and are closer to the maximum tolerable delay, have higher probability to be served. The method is by virtually decreasing the remaining transmission time of OBU.

On the other hand, EDF and FCFS do not have any reaction when OBUs have longer queueing delay which is close to the maximum tolerable delay.

Our goal is to design a downlink scheduling algorithm for DSRC networks in order to minimize the system handoff rate under the maximum tolerable delay requirement. Therefore the service failure rate must be limited in a reasonable range. If the required service failure rate is set at 0.1, the curves of three methods that meet this requirement are meaningful.

Figure 2-10. System Utilization for T =60s

Figure 2-10 and Figure 2-11 show the system utilization versus the mean data size for the three algorithms with the maximum tolerable delay is equal to 60s and 180s, respectively.

It can be seen that the three methods can achieve the same system utilization. RSU can transmit data for OBUs if the transaction queues are not empty. And three methods use the same acceptation rule for OBUs. Therefore three methods can achieve the same system

utilization.

Figure 2-11. System Utilization for T =180s

According to the figure of the system utilization, we divide each following figures into three regions to observe and discuss. The first region is the range of the mean data size from 400 kbytes to the size of 800 kbytes at which the system is at light load. The second region is the range of the mean data size from 800 kbytes to the size of 1000 kbytes at which the system is at medium load. The third region is the other region of the mean data size from 1000 kbytes to 1200 kbytes at which the system load is heavy.

Figure 2-12 and Figure 2-13 show the system handoff rate versus the mean data size for the three algorithms with the maximum tolerable delay equal to 60s and 180s, respectively. It can be seen that in the first region of Figure 2-12, MFL has smaller in the system handoff rate than FCFS but is almost the same as compared to EDF. In the second and third regions, MFL has the system handoff rate better than EDF. And the difference between MFL and EDF turns

Figure 2-12. System Handoff Rate for T =60s

Figure 2-13. System Handoff Rate for T =180s

In Figure 2-13, the behavior of the first and the second regions is the same as Figure 2-12. In the third region, although MFL is still better than FCFS and EDF in the system

handoff rate, the difference between MFL and EDF becomes smaller.

We compare with Figure 2-12 and Figure 2-13. It can be seen that when we increase the maximum tolerable delay from 60s to 180s, the system handoff rate of three methods for all mean data size points becomes larger. Then we explain why the increasing rate of EDF in the service failure rate is slower than FCFS in Figure 2-8 and Figure 2-9. First, both FCFS and EDF do not have a reaction to the queueing delay and think about the maximum tolerable delay. And in Figure 2-12 and Figure 2-13, we can see that EDF is better than FCFS in the system handoff rate. It means that EDF has better ability to determinate the service order for OBUs than FCFS. Therefore EDF has lower probability to let OBUs become failure than FCFS. So the increasing speed of EDF in the service failure rate is slower than FCFS.

As we mentioned before, we divide the OBUs into two kinds, new OBU and handoff OBU. We can individually observe to the new OBU and handoff OBU in the handoff rate.

And the result of the system handoff rate comes from combining the new OBU handoff rate and the handoff OBU handoff rate according to the number of new OBUs and handoff OBUs.

Therefore in the following, we observe and discuss about the new OBU handoff rate and the handoff OBU handoff rate in order to explain the system handoff rate.

Figure 2-14 and Figure 2-15 show the new OBU handoff rate versus the mean data size for the three algorithms with the maximum tolerable delay equal to 60s and 180s, respectively. In the first region of Figure 2-14 and Figure 2-15, it can be seen that the new OBU handoff rate of FCFS increases faster than EDF and MFL. It is because that FCFS does not consider either the remaining SCH dwell time or the remaining transmission time. Also, MFL and EDF have almost the same new OBU handoff rate. It is because that the mean data size is still small in the first region, and thus the remaining transmission time is much less than the remaining SCH dwell time. As we mentioned before, MFL schedule OBUs according to the scheduling index I . If the mean data size is small, the scheduling index _i

will become almost equal to the remaining SCH dwell time. Therefore the result of MFL and EDF will become almost the same.

Figure 2-14. New OBU Handoff Rate for T =60s

Figure 2-15. New OBU Handoff Rate for T =180s

In the second and third regions, the increasing rate of the new OBU handoff rate in MFL is slower than EDF. It is because the effect of the remaining transmission time considered in MFL becomes obvious. And the increasing speed of FCFS is still faster than EDF. The reason is mentioned before.

Then we compare Figure 2-14 and Figure 2-15. It can be found that the performance results for three methods are almost the same individually in the first region. It is because that in the first region the system is at light load and the service failure rate is zero for the three methods. When we increase the maximum tolerable delay from 60s to 180s, the number of handoff OBU only has small increment and the new OBU handoff rate will not be affected at all.

But in the second and third regions, the system is at medium and heavy load and the service failure rate starts to increase from zero. There is more and more handoff OBUs in the system. If the maximum tolerable delay is increased, OBUs can tolerate longer waiting time before service failure. Therefore the number of handoff OBU will become larger and the new OBU handoff rate will become higher.

Figure 2-16 and Figure 2-17 show the handoff OBU handoff rate versus the mean data size for the three algorithms with the maximum tolerable delay equal to 60s and 180s, respectively. In both figures, it can be seen that the handoff OBU handoff rate of FCFS is almost close to zero in the first region. It is because that the system is still at light load and the handoff OBUs have the highest priority in FCFS. As we mentioned before, when a handoff OBU moves to tRSU, it must register to the tRSU first. After it completes the handoff procedure, its remaining data will be forwarded from sRSU to tRSU. And then it can be scheduled when the RSU sends the next RST. Therefore the handoff OBU data will always come to the RSU before the new OBU data in a round.

Figure 2-16. Handoff OBU Handoff Rate for T =60s

Figure 2-17. Handoff OBU Handoff Rate for T =180s

We compare Figure 2-16 and Figure 2-17. It can be found that when the maximum tolerable delay is increased from 60s to 180s, the handoff OBU handoff rate for three

methods will all become higher. The reason is that when we increase the maximum tolerable delay, there is more handoff OBUs staying in the system but the time of RSU for transmission data is the same. Besides, it can be seen that the increasing degree of handoff OBU handoff rate in MFL is larger than other two methods. The reason is that the priority for OBUs that have longer queueing delay is increased according to the ratio of the queueing delay over the maximum tolerable delay. Thus as the maximum tolerable delay is increased, the degree of priority increment for handoff OBUs will decrease. We observe the figures for the new OBU handoff rate and the handoff OBU handoff rate. It can be found that MFL is better than EDF and EDF is better than FCFS in new OBU handoff rate. But the order is reverse in handoff OBU handoff rate. It is because that the total time that RSU can transmit data for OBUs is fixed. Consequently, the new OBU and handoff OBU can not be kept in mind at the same time.

Finally, we come back to see the performance of system handoff rate. For three regions in Figure 2-12, the performance behaviors for the three methods are dominated by the new OBU handoff rate. However, for the first region in Figure 2-13, the performance behaviors are dominated by the new OBU handoff rate. But for the second and the third regions in Figure 2-13, the performance behaviors are dominated by the handoff OBU handoff rate. It is because that when we increase the maximum tolerable delay from 60s to 180s, the number of the handoff OBU will increase significantly.