Performance Evaluation
5.2 MAC-layer Performance
5.2.1 Performance Metrics
The Average Transmission Opportunity Utilization of Nodes (ATOUN) The utilization of a node’s control-plane bandwidth is an important metric used to evaluate the efficiency of DEDA. A node’s transmission opportunity utilization is defined as the aggregate transmission opportunity use within a node’s extended neighborhood. It indicates how well its extended neighborhood utilize the net-work’s transmission opportunities. The ATOUN metric is the average across all node’s transmission opportunity utilization in a network case. The detailed defi-nition of the ATOUN metric is given in [16].
The Average Three-way-Handshake Procedure Time (ATHPT)
The three-way handshake procedure time shows the efficiency of data schedule establishment. The average three-way handshake procedure time metric is defined as the average time required by the three-way handshake procedure to establish a data schedule across all network nodes in a network case. The detailed definition of the ATHPT metric is given in [16].
300m 300m
Figure 5.1: The simulation network topology.
The Average Number of Established Data Schedules (ANEDS)
The number of established data schedules (NEDS) of a node, defined by Equa-tion (5.1), is used to measure the achieved spatial reuse degree of a node. We then use Equation (5.2) to compute a network case’s ANEDS value, which is the average of all network nodes’ NEDS values. ANEDS is important to evaluate the overall spatial reuse degree of a directional-antenna network.
N EDS(i) = X
j∈N BR1i
nij (5.1)
where nij is the number of data schedules established from N odei to N odej N BR1i is the set of N odei’s one-hop neighbors
AN EDS = PN
i=1N EDS(i)
N (5.2)
where N is the number of nodes in a network case.
5.2.2 Simulation Environment
We use a 5x5 grid network comprising 25 nodes for simulation. As shown in Fig 5.1, each node is spaced 300 meters apart from its vertical and horizontal neighbors.
All nodes except boundary ones have 8 surrounding one-hop neighbors.
To shed light on the performances of our proposed design, we first conduct simulations with all combinations of the parameters listed in Section 5.1, including the α value, antenna beam width, . . . . As a result, in total 56 different cases are generated using the 5x5 grid network topology. The β value discussed in Section 4.5.3 is fixed to be 7 in all simulation cases. Each simulation case is run five times, each time using a different random number seed.
The simulated time for each run is set to 310 seconds. During simulation, each node starts a MAC-layer pseudo data scheduler at the 150th second to periodically establish data schedules with its neighboring nodes in a round-robin manner. The frequency is chosen to be one data schedule every 100 milliseconds. With this frequency setting, the pseudo data scheduler can generate a heavy traffic load which approaches the maximum control-plane utilization.
5.2.3 Simulation Results
As shown in Table 5.1 and Table 5.2, the ATOUN results show that the directional-antenna network has higher transmission opportunity utilization than the omni-direction-antenna network, if DEDA-ITHP is adopted. The directional-antenna network, however, has larger ATHPT values than omni-direction-antenna net-works. There are two reasons for explaining this phenomenon.
First, since our design employs only a steerable directional antenna for a net-work node, each node is allowed to transmit control messages to one of its antenna domains at a time (i.e., it cannot transmit control messages to multiple beams si-multaneously.). As such, one node transmits control messages to all of its antenna domains in a rough round-robin manner. To complete a three-way handshake procedure, a requesting node should transmit two MSH-DSCH messages, one of which carries the request IE and the other carries the confirm IE. Thus, a request-ing node typically requires two rounds (a round can be roughly defined as the minimum of the required time for a node to transmit its control messages to all of its antenna domains.) to complete a three-way handshake procedure. In contrast,
Table 5.1: MAC-layer results using antenna with beam width π2
(a) With DEDA.
ATOUN ATHPT (ms) ANEDS
Avg. Std. dev. Avg. Std. dev. Avg. Std. dev.
Omni-direction-antenna 0.575 0.000 35.275 0.304 366.872 5.873
Using next txopp of the covered
beam only
α = 0 0.301 0.000 263.967 3.643 521.472 21.429 α = 1 0.376 0.000 219.514 4.338 602.416 22.813 α = 2 0.391 0.000 211.961 3.197 615.024 20.303 α = 3 0.397 0.000 210.843 3.598 613.504 20.669 α = 4 0.400 0.000 210.043 3.323 621.624 17.666 α = 5 0.401 0.000 209.860 3.572 624.568 15.375 α = 6 0.401 0.000 209.119 3.377 624.992 16.423
Using next txopps of all beams
α = 0 0.323 0.000 260.384 1.282 551.664 17.337 α = 1 0.377 0.000 219.946 4.113 603.984 19.772 α = 2 0.391 0.000 213.957 0.709 625.656 2.583 α = 3 0.397 0.000 212.887 0.765 631.208 1.978 α = 4 0.400 0.000 212.045 0.918 634.392 10.288 α = 5 0.401 0.000 211.184 0.756 630.384 7.284 α = 6 0.402 0.000 212.229 1.057 629.864 5.150
(b) With DEDA-ITHP.
ATOUN ATHPT (ms) ANEDS
Avg. Std. dev. Avg. Std. dev. Avg. Std. dev.
Omni-direction-antenna 0.575 0.000 35.275 0.304 366.872 5.873
Using next txopp of the covered
beam only
α = 0 0.301 0.000 263.967 3.643 521.472 21.429 α = 1 0.552 0.000 155.311 1.193 744.680 3.436 α = 2 0.617 0.000 142.506 1.120 785.776 1.942 α = 3 0.652 0.001 138.470 0.494 793.312 3.640 α = 4 0.681 0.001 138.712 0.720 807.200 5.448 α = 5 0.701 0.000 136.008 4.477 795.888 38.682 α = 6 0.709 0.001 135.385 6.179 790.096 62.781
Using next txopps of all beams
α = 0 0.323 0.000 260.384 1.282 551.664 17.337 α = 1 0.557 0.001 152.286 3.354 730.048 23.838 α = 2 0.621 0.001 141.695 0.895 781.752 6.596 α = 3 0.657 0.000 138.123 0.688 800.464 7.798 α = 4 0.686 0.001 138.187 0.626 802.224 8.980 α = 5 0.705 0.000 137.351 0.927 805.744 9.814 α = 6 0.714 0.001 137.369 0.356 812.912 7.712
Table 5.2: MAC-layer results using antenna with beam width π3
(a) With DEDA.
ATOUN ATHPT (ms) ANEDS
Avg. Std. dev. Avg. Std. dev. Avg. Std. dev.
Omni-direction-antenna 0.575 0.000 35.275 0.304 366.872 5.873
Using next txopp of the covered
beam only
α = 0 0.301 0.000 397.102 21.905 384.376 32.487 α = 1 0.512 0.000 243.006 1.004 607.936 5.622 α = 2 0.540 0.000 231.830 0.478 631.560 9.062 α = 3 0.555 0.000 227.656 1.067 630.704 8.988 α = 4 0.567 0.000 225.551 1.210 640.576 5.898 α = 5 0.574 0.000 224.659 1.032 633.312 10.199 α = 6 0.578 0.000 223.063 1.069 635.560 6.080
Using next txopps of all beams
α = 0 0.317 0.000 410.736 1.903 409.744 8.867 α = 1 0.513 0.000 242.636 1.212 609.888 9.287 α = 2 0.541 0.000 232.718 1.103 638.120 10.549 α = 3 0.556 0.000 226.700 1.344 629.320 10.470 α = 4 0.568 0.000 225.115 0.568 639.352 3.796 α = 5 0.575 0.000 223.731 0.935 635.200 8.699 α = 6 0.579 0.000 222.394 1.068 632.944 9.892
(b) With DEDA-ITHP.
ATOUN ATHPT (ms) ANEDS
Avg. Std. dev. Avg. Std. dev. Avg. Std. dev.
Omni-direction-antenna 0.575 0.000 35.275 0.304 366.872 5.873
Using next txopp of the covered
beam only
α = 0 0.301 0.000 397.102 21.905 384.376 32.487 α = 1 0.581 0.000 213.000 0.843 633.416 5.135 α = 2 0.635 0.000 199.284 1.111 666.336 5.626 α = 3 0.661 0.001 196.341 0.806 676.680 14.175 α = 4 0.684 0.000 194.016 0.156 685.992 11.105 α = 5 0.703 0.001 193.432 0.672 701.952 6.773 α = 6 0.716 0.000 192.495 1.327 720.144 4.936
Using next txopps of all beams
α = 0 0.317 0.000 410.736 1.903 409.744 8.867 α = 1 0.582 0.001 212.646 0.926 625.512 7.455 α = 2 0.638 0.000 198.638 1.170 658.808 8.489 α = 3 0.664 0.001 195.994 1.258 677.928 8.368 α = 4 0.687 0.000 193.061 0.341 695.096 6.481 α = 5 0.707 0.001 192.608 0.875 703.224 14.102 α = 6 0.720 0.000 191.917 1.023 699.608 19.719
in an omni-directional antenna network, a network node can transmit informa-tion elements for different nodes using the same MSH-DSCH message due to the broadcast nature and thus reduce the required time of the three-way handshake procedure.
Second, our proposed DEDA may choose a larger holdoff time exponent for an antenna domain for maintaining network operation (avoid collisions of antenna coverage with neighboring nodes). To address this problem, we also propose the DEDA-ITHP design, extended from DEDA, to improve the efficiency of establish-ing data schedules. The ATHPT results show that DEDA-ITHP shortens ATHPT by a factor of 1.5, when compared with DEDA. Regarding ANEDS, the directional-antenna network can complete more three-way handshake procedures within the same simulated time by factors from 1.05 to 2.216 (at least 1.642 if the holdoff time exponent value is not fixed), when compared with the omni-direction-antenna network. In addition, DEDA-ITHP can on average outperform DEDA in ANEDS by a factor of 1.25.
Fig 5.2 shows that the ATOUN value on average is increased as the α value or the number of antenna domains is increased. Besides, we can see that DEDA-ITHP efficiently increases the ATOUN value. Lastly, the ATOUN value is slightly increased by the improved one-hop neighbors eligibility determination.
Fig 5.3 shows that the ATHPT value is decreased as the α value is increased.
Besides, the results show that DEDA-ITHP can further shorten ATHPT. Also note that if the number of antenna domains decreases, ATHPT can be decreased because in such a condition the required time for a round of control message dissemination can be reduced.
Fig 5.4 shows that the ANEDS value increases as the α value increases. As we can see in this figure, DEDA-ITHP greatly increases ANEDS in all cases. However, increasing the number of antenna domains does not increase the ANEDS value.
The reason for this unexpected result is that for a grid network, using 4 or 6 antenna domains does result in much different spatial reuse degree because for each node, the density of neighboring nodes is quite regular. The effect of the
0 PI/2, DEDA-ITHP, 1 txopp PI/2, DEDA-ITHP, N txopps PI/3, DEDA, 1 txopp PI/3, DEDA, N txopps PI/3, DEDA-ITHP, 1 txopp PI/3, DEDA-ITHP, N txopps
Figure 5.2: ATOUN versus α value.
0 PI/2, DEDA-ITHP, 1 txopp PI/2, DEDA-ITHP, N txopps PI/3, DEDA, 1 txopp PI/3, DEDA, N txopps PI/3, DEDA-ITHP, 1 txopp PI/3, DEDA-ITHP, N txopps
Figure 5.3: ATHPT versus α value.
0 PI/2, DEDA-ITHP, 1 txopp PI/2, DEDA-ITHP, N txopps PI/3, DEDA, 1 txopp PI/3, DEDA, N txopps PI/3, DEDA-ITHP, 1 txopp PI/3, DEDA-ITHP, N txopps
Figure 5.4: ANEDS versus α value.
number of antenna domains can further studied using random network topologies in the future.
In summary, these simulation results show the following observations. First, a directional-antenna network utilizes the control-plane bandwidth more efficiently than an omni-direction-antenna network. Second, DEDA is essential to the directional-antenna network since it provides a higher control-plain utilization, a shorter three-way handshake procedure time, and a larger number of established data schedules, as compared with the static holdoff time exponent setting. Then, DEDA-ITHP provides better performances than DEDA does. Fourth, the improved one-hop neighbors eligibility determination slightly increases the control-plane utilization.
Finally, a larger number of antenna domains slightly increases the control-plane utilization but lengthens the procedure time of three-way handshake.