Under Four-Way Handshake Procedure

In Chapter 5, we elaborated the extended hidden terminal (EHT) problem and proposed our solution: the four-way handshake procedure. In the following, we first present the Reduced Ratio of Collisions (RRC) of three different topologies and show the great reductions in our scheme. Second, we display the throughput results by AATFT.

Reduced Ratio of Collisions:

0 200 400 600 800 1000

0 100 200 300 400 500 600 700 800 900 1000

AATFT (Kbyte/sec)

Connection queue size (number of packets)

Legends fresh clear timer 5sec fresh clear timer 2.5sec fresh clear timer 2sec fresh clear timer 1.5sec fresh clear timer 1sec

Figure 6.7: The throughput of butterfly network under different fresh clear timers

0 500 1000 1500 2000 2500

0 100 200 300 400 500 600 700 800 900 1000

AATFT (Kbyte/sec)

Connection queue size (number of packets)

Legends fresh clear timer 5sec fresh clear timer 2.5sec fresh clear timer 2sec fresh clear timer 1.5sec fresh clear timer 1sec

Figure 6.8: The throughput of 7-node chain network under different fresh clear timers

0 200 400 600 800 1000

0 100 200 300 400 500 600 700 800 900 1000

AATFT (Kbyte/sec)

Connection queue size (number of packets)

Legends fresh clear timer 5sec fresh clear timer 2.5sec fresh clear timer 2sec fresh clear timer 1.5sec fresh clear timer 1sec

Figure 6.9: The throughput of 9-node grid network under different fresh clear timers

Table 6.3: The average reduced ratio of collisions in the four-way handshake proce-dure

Without Data Schedule With Data Schedule Holdoff Algorithm Holdoff Algorithm

7-node chain network 29.86% 17.5%

9-node grid network 1.49% 0.02%

25-node grid network 1.82% 1.16%

The RRC denotes the ratio of collision preventions in our four-way handshake procedure. We compare the four-way handshake procedure with and without the data schedule holdoff algorithm proposed in Section 5.4. The results are shown in Table 6.3. Obviously, our four-way handshake procedure highly reduces the collision occurrences and the Data Schedule Holdoff Algorithm is more effective than the original four-way handshake scheme.

Aggregate Average Traffic Flow Throughput:

After discussing the reduction of collision occurrences, we then show the through-put results comparing with the original WiMAX, network coding scheme, four-way handshake procedure and four-way handshake with data schedule holdoff. Fig. 6.10 shows the AATFT values versus connection queue size under different schemes in the 7-node chain network. The original WiMAX throughputs are much lower than the other network coding schemes. In addition, we have better results when using four-way handshake scheme. This result indicates the collision damage is solved by our design.

In Fig. 6.11, the EHT problem frequently occurs in such a network topology.

We can find out that the throughput with network coding is sometimes close to the original WiMAX scheme. This result shows that the collision actually affects the goodput generated by network coding scheme because the throughput highly increases when using the four-way handshake procedure to prevent the collision occurrences. The result in Fig. 6.12 also displays the improvements by using the

0

Connection queue size (number of packets) Legends original

rule-based NC rule-based NC+4way

rule-based NC+4way+algorithm

Figure 6.10: The throughput of 7-node chain network

four-way handshake procedure. The coding opportunities in such a large network may be less than the coding opportunities in a small network like a 9-node grid network. That is, the EHT problem does not always occur in this case. Although the improvements are not as great as in Fig. 6.11, we still prove our four-way handshake procedure is also suitable for the large networks.

In conclusion, the results in Fig. 6.11, Fig. 6.10 and Fig. 6.12 show the network coding performance gain can be elevated by reducing the collision occurrences. Our design solves the EHT problem in both small and large networks.

Average End-to-end Delay:

We consider the delay issue in this section. The network coding scheme can reduce the average end-to-end delay in a network because packets may be drained out quickly through the encoding procedure. Table 6.1 and Table 6.2 show this condition where the average end-to-end delays of our network coding scheme are

0 200 400 600 800 1000 1200 1400

0 100 200 300 400 500 600 700 800 900 1000

AATFT (Kbyte/sec)

Connection queue size (number of packets) Legends original

rule-based NC rule-based NC+4way

rule-based NC+4way+algorithm

Figure 6.11: The throughput of 9-node grid network

0 200 400 600 800 1000 1200 1400 1600

0 100 200 300 400 500 600 700 800 900 1000

AATFT (Kbyte/sec)

Connection queue size (number of packets) Legends original

rule-based NC rule-based NC+4way

rule-based NC+4way+algorithm

Figure 6.12: The throughput of 25-node grid network

Table 6.4: Average end-to-end delay under different connection queue sizes in 7-node chain networks

original rule-based nc rule-based nc+4way rule-based nc+4way+algorithm

connection queue AETED Sdv AETED Sdv AETED Sdv AETED Sdv

size (packets) (sec) (sec) (sec) (sec)

50 0.317 0.004 0.202 0.015 0.233 0.008 0.223 0.004

100 0.550 0.007 0.374 0.001 0.427 0.004 0.409 0.008

200 1.171 0.091 0.679 0.015 0.784 0.028 0.813 0.004

300 1.908 0.166 1.007 0.036 1.203 0.098 1.057 0.096

400 2.598 0.188 1.309 0.129 1.413 0.098 1.354 0.133

500 3.228 0.249 1.401 0.186 1.916 0.233 1.596 0.001

600 3.867 0.306 1.800 0.019 2.064 0.272 2.217 0.005

700 4.248 0.022 2.019 0.128 2.342 0.125 2.587 0.060

800 4.632 0.092 1.961 0.237 2.508 0.121 2.965 0.092

900 5.031 0.200 2.383 0.314 2.921 0.246 3.300 0.079

1000 5.398 0.303 2.558 0.178 3.150 0.344 3.589 0.406

less than of original WiMAX scheme.

Another delay topic is related to the effects of the four-way handshake procedure and the data schedule holdoff algorithm. The four-way handshake procedure will cost some time to forward the extended confirm messages and lengthen the end-to-end delays. Furthermore, the data schedule holdoff algorithm will postpone the data schedule to prevent collision to occur. This behavior will also increase the end delays. In Table 6.4, Table 6.5 and Table 6.6, we list the average end-to-end delay and the standard deviation under different connection queue sizes. The results correspond with our expectation explained above, the average end-to-end delay under four-way handshake procedure is larger than under our network coding scheme. Besides, the average end-to-end delay using data schedule holdoff algorithm is larger than only using four-way handshake procedure.