Scenario One

In the first scenario, the traffic of IEEE 802.11 is downlink traffic with different service rate. The traffics of IEEE 802.16e are three rtPS connections. The detail parameters of IEEE 802.16e connections are showed in table 4-2.

Table 4-2 Parameters of IEEE 802.16e connections in scenario one Flow id Max rate(Kbps) Min rate(Kbps) Max latency(frames)

1 400 200 d1

2 450 250 d2

3 550 300 d3

The values of maximum latency in table 4-2 are not defined because we will observe the effect of different latency in the overlapping time of sleep state between IEEE 802.11 and 802.16e. We compare our power saving mechanism with transmission merging mechanism in [18]. And the improvement of overlapping time of sleep state is showed in table 4-3.

Table 4-3 Improvement of overlapping time of sleep state (IEEE 802.11 with 0.1M CBR)

% Listen interval (beacons)

dmin 2 3 4 5 6 7 8 9 10

6 -0.025 3.608 0.009 0.926 2.488 -0.522 0.203 1.413 -0.461 7 0.322 -0.009 0.357 -0.163 0.464 1.663 0.212 -0.152 -0.174 8 3.151 0.852 3.122 1.180 2.201 0.482 1.651 -0.137 1.236 9 -1.146 1.450 2.047 1.051 -0.083 0.251 0.297 1.461 -0.260 10 2.948 2.985 2.921 2.444 2.059 1.803 1.544 1.417 1.156 11 0.084 -0.321 0.102 0.905 0.916 0.426 0.150 -0.482 -0.356 12 -0.815 0.891 1.860 0.864 1.053 1.062 0.163 0.744 0.853 13 -0.170 1.717 -0.132 1.919 -0.036 1.235 -0.094 0.609 0.041 14 2.712 -0.276 2.547 -0.087 0.784 1.962 -0.151 0.732 -0.101

15 3.278 2.993 1.699 2.604 2.855 0.350 1.201 2.067 0.347

Enter sleep mode

S1 W1 W2 S3 W3 W4

Period Period

802.16e

traffic ……

Figure 4-1 Behavior of listening/sleep window in transmission merging mechanism

From table 4-3, we can find two obvious results. First, in some cases the values of improvement are negative. It means using our power saving mechanism can’t increase the overlapping time of sleep state. Before explain this problem, we need to know the behavior of listening/sleep window while using transmission merging mechanism in [18]. An example is showed in figure 4-1. From figure 4-1, we can find that the behaviors of listening/sleep window are periodic. In transmission merging mechanism, if there is no UGS traffic, it will only compute formula (3) in chapter 3 to obtain the value of maximum size of next sleep window, Smax. Then it will set Smax be next sleep window. When we have a long next sleep window, it will make the next two sleep window be zero to satisfy formula (1) in chapter 3. The value of period will equal to dmin. In this situation, the awake/sleep state of IEEE 802.11 and 802.16e traffics are both period. Therefore the awake/sleep state of IEEE 802.11 and 802.16e traffics will overlap periodically even if doing nothing. But when applying our power saving mechanism, it will destroy the period of IEEE 802.16e. So in some cases the summation of overlapping time which using our power saving mechanism might less than transmission merging mechanism.

Then the second result is the improvements are not regular even if under the

Table 4-4 Average length of sleep window (IEEE 802.11 with 0.1M CBR) Minimum latency (dmin)

Listen

interval 6 7 8 9 10 11 12 13 14 15

2 3.72 4.68 6.00 6.02 8.00 8.02 4.59 9.94 10.51 10.69 3 4.00 4.67 5.51 6.58 8.00 8.01 5.43 9.21 9.21 12.00 4 3.99 4.68 6.00 6.89 8.00 8.02 5.09 9.95 10.51 10.53 5 3.89 4.99 5.70 6.97 8.00 8.96 4.83 9.64 10.94 11.44 6 4.00 4.99 6.00 6.59 8.00 8.92 5.20 9.14 10.47 12.00 7 3.99 5.00 5.79 6.76 8.00 8.78 5.05 9.82 11.00 11.11 8 3.93 4.96 6.00 6.90 8.00 8.68 4.91 9.40 10.44 11.65 9 4.00 4.93 5.84 7.00 8.00 8.62 5.13 9.93 10.90 12.00 Ours

10 3.99 4.91 6.00 6.98 8.00 8.96 5.05 9.60 10.44 11.40

Original N/A 4 5 6 7 8 9 5 10 11 12

same listen interval or same value of dmin. The main reason is our power saving mechanism makes the behavior of awake/sleep state of IEEE 802.16e be non-period.

Even through, in most case, our power saving mechanism actually increases the overlapping time of sleep state.

Table 4-4 are the results of average length of sleep window. Original in table 4-4 means transmission merging mechanism. From table 4-4, we can find the average length of sleep window in our method is slightly small than transmission merging mechanism. Because in our method, we will choose small next sleep window to make the awake state of IEEE 802.11 and IEEE 802.16e synchronously. And in table 4-4, we can find when minimum latency is twelve the average length of sleep window is

unusually small. This is because the behavior of listening/sleep window is this case is not like figure 4-1. It doesn’t have zero length of sleep window. It has a long sleep window and a short sleep window interleaved.

Then we increase the traffic load of IEEE 802.11 to 0.3Mbps and 0.5Mbps. And the results are showed in table 4-5.

Table 4-5 Improvement of overlapping time of sleep state (IEEE 802.11 with 0.3M and 0.5M CBR separately)

0.3Mbps CBR

% Listen interval (beacons)

dmin 2 3 4 5 6 7 8 9 10

6 1.590 4.542 -2.149 0.400 2.615 -0.845 -0.347 1.030 -0.066 8 7.214 -0.144 3.423 -0.787 0.879 0.861 1.735 -0.144 0.845 10 6.719 4.531 3.288 2.590 1.342 0.291 1.346 1.490 1.212 12 1.499 2.242 1.180 -0.124 1.057 0.014 -0.282 0.137 1.043 14 6.510 -0.298 1.529 -1.528 -0.559 2.082 -1.560 -0.450 -0.375

0.5Mbps CBR

6 -0.930 3.636 -1.050 -0.635 2.678 -1.040 0.379 1.684 -0.306 8 7.759 -1.686 3.010 0.054 1.706 0.080 1.820 -0.015 1.316 10 7.190 4.811 0.817 2.649 2.291 1.892 0.573 1.243 1.110 12 0.475 2.395 0.304 0.100 1.167 0.889 -0.374 0.173 0.393 14 5.639 -1.921 0.218 -3.105 -0.146 2.063 -0.546 0.487 -0.647

Like the result in table 4-3, in some cases, our power saving mechanism can’t get

better performance and the ratio of bad case increases. But in some cases, the improvement of overlapping time of sleep state is increasing compare with table 4-3.

The main reason is that awake state of IEEE 802.11 will become more longer while traffic load increases. And this makes the summation of overlapping time of sleep state of mobile device decrease. The percentage of improvement of our power saving mechanism becomes bigger naturally.

From table 4-2 and 4-3, we can find one more thing. The percentage of improvement with small listen interval is much bigger than long listen interval.

Before explaining the reason, we need to know one more important thing. From chapter 3, we know our power saving mechanism in IEEE 802.16e is choose a proper next sleep window to make the next awake state of IEEE 802.11 and IEEE 802.16e to be synchronous. If IEEE 802.11 traffic has a small listen interval, IEEE 802.16e has more chances to adjust its next awake window to synchronize with IEEE 802.11 traffic. And the summation of overlapping time of sleep state will be big with high probability.

Then we will add UGS connections into IEEE 802.16e traffic. The parameters of UGS connections are showed in table 4-6. The results of simulation are showed in table 4-6.

Table 4-6 Parameters of UGS connections

Flow id Grant period(frames) Min rate(Kbps) Max latency(frames)

4 14 64 5 5 9 64 4 6 5 64 2

Table 4-7 Improvement of overlapping time of sleep state (IEEE 802.11 with 0.1M CBR and IEEE 802.16e with UGS connections)

% Listen interval (beacons)

dmin 2 3 4 5 6 7 8 9 10

6 -0.450 0.252 0.840 0.458 0.164 -0.298 -0.042 0.340 -0.109 8 -0.516 0.550 0.530 0.356 -0.107 -0.034 -0.125 0.000 0.066 10 0.000 0.064 0.095 -0.144 -0.480 -0.699 -0.511 0.000 -0.256 12 2.058 1.160 1.203 0.626 0.095 0.004 0.000 0.000 0.128 14 2.058 1.160 1.203 0.626 0.095 0.004 0.000 0.000 0.128

Table 4-8 Average length of sleep window (IEEE 802.11 with 0.1M CBR and IEEE 802.16e with UGS connections)

Minimum latency (dmin) Listen

interval 6 8 10 12 14

2 2.23 2.69 3.60 3.99 3.99

3 2.12 2.68 3.63 3.99 3.99

4 2.10 2.70 3.74 3.99 3.99

5 2.06 2.71 3.80 3.99 3.99

6 2.07 2.70 3.79 3.99 3.99

7 2.11 2.71 3.86 3.99 3.99

8 2.06 2.70 3.87 3.99 3.99

9 1.99 2.71 3.86 3.99 3.99

Ours

10 2.04 2.71 3.89 3.99 3.99

Original N/A 2.01 2.72 3.99 3.99 3.99

Comparing with table 4-3 and 4-7, we can find the improvement is decreasing. The main reason is the addition of UGS connections makes IEEE 802.16e traffic have more awake states. So these awake states will reduce the amount of sleep states and decrease the performance of our power saving mechanism.

Table 4-8 are the results of average length of sleep window. Comparing with table 4-4, we can find when IEEE 802.16e has UGS connections the length of sleep window will decrease. The main reason is the one grand period among UGS connections is small. From table 4-6, we can find there is a UGS connection with grand period is five. It means every five frames, the mobile station must wake up and receive/transmit UGS data.