4.4 Call Admission Control using HWMP
4.4.2 Applicability of the Proposed Scheme for Four Class Service System 80
IEEE 802.11 defines four classes of services - voice (VO), video (VD), background (BK) and best effort (BE). The proposed scheme can be extended to support these four classes of service. The specific service requirements fo these four classes are as follows;
Voice (VO): Highest priority traffic, requires minimum bandwidth guarantee, minimum delay and negligible packet loss.
Video (VD): Next highest priority, delay should be as small as possible, jitter should be negligible, can tolerate limited packet loss.
Background (BK):Priority is less than video traffic, however loss rate should be as less as possible.
Best-Effort (BE):No specific QoS requirement.
Following criteria should be satisfied to support these four class services over the proposed scheme;
℘V O> ℘V D> ℘BK > ℘BE (4.28) Equation (4.28) denotes that the priority of VO traffic is maximum, and the traffic priority decreases in the order of VD to BK and BE traffic. Further following minimum traffic guarantee should be ensured.
λBE ≥0
λV O> λV D≥λBK > λBE (4.29) The specific service parameters can be set by the service provider. It can be noted that the end-to-end delay guarantee can be ensured by the timeout interval of HWMP PREP reception during flow admission. The above considerations are sufficient to support QoS provisioning for four class service system. The proposed scheme allocates channel share to the traffic from every class proportionally to their class priority, maintaining minimum bandwidth demand. The minimum bandwidth demand also ensures delay and jitter guarantee. Further, the α value ensures that no traffic class can overuse channel share that may affect BE services. The effectiveness of (α, ℘) proportional fairness over max-min fairness and proportional fairness is analyzed in the next section using simulation results.
4.5 Simulation Results
The proposed scheme is implemented using Qualnet 5.0.1 [71] network simulator frame- work. The performance of the proposed scheme is analyzed for service-differentiation and fairness, and compared with IEEE 802.11s standard. Further the effectiveness of the proposed scheme for (α, ℘) fairness is compared with max-min fairness and proportional fairness. Standard IEEE 802.11 four class service is considered for performance measure- ment in various scenarios.
4.5.1 Simulation Set-up
In the simulation scenario, 40 mesh STAs have been deployed in the simulation arena uniformly using mean connectivity 4. This indicates that every mesh STA has on an
average 4 mesh STAs in its communication range. Out of these mesh STAs, 3 mesh STAs have been selected as mesh gate. The mesh gates are selected using degree based Greedy Dominating Tree Set Partitioning (degree based GDTSP) algorithm proposed in [207]. Degree based GDTSP algorithm optimizes connectivity among the mesh gates and other mesh STAs in the network. Every mesh STA is equipped with 3 interfaces.
Every interface is connected with switched beam smart antenna with main lobe gain as 15dB and side lobe gain as −20dB. IEEE 802.11g 54M bps physical layer technology is used for communication. Capture effect is enabled at physical layer with capture threshold of 4dBwith transmit power 16dBm. These physical layer settings are according to CISCO 1500 series MAPs [208].
MCCA is used as the MAC layer standard. DTIM interval is kept at 600µs. The beacon transmission time is set as 200µs. HWMP is used for mesh path selection. HWMP route timeout is kept as 3sec, and number of PREQ attempts is set to 2. That means a mesh STA can try to find out the the path for a flow at most 2 times. If no PREP message arrives for 2 consecutive times, the flow is dropped.
Four classes of traffic are considered for simulation purpose. The traffic are generated using Qualnet traffic generator framework. The traffic priority, minimum bandwidth demand and average flow duration for each service class is given in Table 4.1. It can be noted that the properties shown in Table 4.1 may vary in real system. For example, the average flow durations are taken smaller compared to their actual durations, to avoid lengthy simulation process. However, for the comparative analysis purpose, the variance in the data matters, not their actual values. For example, the average duration of a VO flow should be less compared to a VD flow. In real system, the results can be scaled up or scaled down based on the actual parameters used. Here we are only interested in a comparative performance analysis. The parameters shown in Table 4.1 are sufficient for this purpose.
The flow durations are chosen using log-normal distribution with mean values given in Table 4.1. This captures real time traffic distribution characteristics in a community network. Every experiment is executed for 10 different times with 10 different seed values.
The average values are taken to plot the graphs. However, the confidence intervals are also shown in every graph.
4.5.2 Effect of Inter-Class Flow Differentiation and Intra-Class Fairness For this class of experiments, two types of flows are taken - high priority VO flows, and low priority BE flows. The traffic generation rate for VO flows is exponentially distributed with
Table 4.1: Traffic properties for different service classes Traffic
Class Priority
Minimum Demand (kbps)
Mean Flow Duration (min)
VO 7 64 1
VD 5 48 5
BK 3 20 5
BE 1 0 20
mean 100Kbps and variance 10Kbps. Similarly the traffic generation rate for BE flows is Poisson distributed with mean 1M bps and variance 0.1M bps. Service differentiation gets affected just after the network saturation (total traffic demand overshoots network capacity). Therefore number of flows are selected based on saturation point. Jain fairness index [197] is used to check fairness among the flows from similar traffic classes. Jain fairness index is defined as follows;
F(λ) = (P
λ(F))2 n(P
λ(F)2) (4.30)
whereλ(F) is the throughput for flowf, andnis the total number of such flows. For these set of experiments, α value is taken as 20 that provides a trade-off between proportional fairness and max-min fairness.
Figure 4.1 shows the throughput of VO flows with respect to increasing number of BE flows. For this experiment, network saturation occurs with 40 numbers of BE flows.
30 VO flows are distributed uniformly in the network. BE flows are also distributed uniformly so that complete network would get saturated. Without uniform distribution of flows, performance comparison is difficult as some portion of the network gets saturated whereas some portion may still remain in unsaturation. Average throughput for VO flows is considered to plot the graph. It can be observed from the graph that before saturation, both the proposed scheme and standard IEEE 802.11s performs similarly. However after saturation, the performance for VO flows drops exponentially in case of IEEE 802.11s network. On the other hand, the throughput for the VO flows in the proposed scheme is always maintained to its minimum demand, 64Kbps. In case of IEEE 802.11s, the BE flows reserve maximum MCCAOPs in the DTIM interval, which affects the performance of VO flows. In the proposed scheme maximum MCCAOP reservation for BE flows is limited by number of high priority VO flows present in the network.
20 30 40 50 60 70 80 90 100 110 120
20 30 40 50 60 70
Per-Flow VO Throughput (Kbps)
Number of BE Flows Proposed Scheme
IEEE 802.11s
Figure 4.1: Throughput: VO Flows in Presence of BE Flows
0.5 0.6 0.7 0.8 0.9 1 1.1
40 50 60 70 80 90
Per-Flow BE Throughput (Mbps)
Number of VO Flows Proposed Scheme
IEEE 802.11s
Figure 4.2: Throughput: BE Flows in Presence of Voice Flows
Figure 4.2 shows the throughput for BE flows with respect to increasing number of VO flows. For this experiment, 20 numbers of BE flows are considered. Like previous, the BE and VO flows are distributed uniformly in the network. For this scenario, network saturation occurs with 50 numbers of VO flows. It can be noted that there is no minimum traffic demand for BE services. Therefore, as the number of VO flows increase in the network, the throughput for the BE flows drops in the proposed scheme to
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1
20 30 40 50 60 70
Per-Flow BE Throughput (Mbps)
Number of BE Flows Proposed Scheme
IEEE 802.11s
Figure 4.3: Throughput: BE Flows in Presence of BE Flows
accommodate the high priority VO flows. Standard IEEE 802.11s does not make service class differentiation. Therefore, increasing number of VO flows in the network does not affect the performance of BE flows. As the data rate for BE flows is significantly higher than the VO flows, BE flows reserve most of the MCCAOPs in the DTIM interval for IEEE 802.11s, affecting performance of VO flows.
Figure 4.3 shows the performance of BE flows with increasing number of BE flows in the network. For this experiment the network contains only BE flows. Network saturation occurs with 35 numbers of BE flows. In this case, the throughput for BE flows drops for both the cases - the proposed scheme and standard IEEE 802.11s. However, average per-flow throughput for the proposed scheme is higher compared to IEEE 802.11s. The reason behind this is intra-class fairness provisioning used in the proposed scheme. In IEEE 802.11s standard, some flows get starved when traffic demand is very high. Fairness assurance in the proposed scheme solves the problem of starvation, resulting increase in per-flow throughput.
Intra-class fairness for the above three experiments have been shown in Figure 4.4- Figure 4.6. Figure 4.4 shows fairness among VO flows with respect to increasing number of BE flows. Figure 4.5 plots the fairness index for BE flows with increasing number of VO flows. Similarly Figure 4.6 shows fairness among BE flows when number of BE flows are increased in the network. In all the three cases, the proposed scheme provides more intra-class fairness compared to standard IEEE 802.11s.
0.6 0.7 0.8 0.9 1 1.1 1.2
20 30 40 50 60 70
Jain Fairness Index for VO Flows
Number of BE Flows Proposed Scheme
IEEE 802.11s
Figure 4.4: Fairness Index: VO Flows in Presence of BE Flows
0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2
40 50 60 70 80 90
Jain Fairness Index for BE Flows
Number of VO Flows Proposed Scheme
IEEE 802.11s
Figure 4.5: Fairness Index: BE Flows in Presence of Voice Flows
4.5.3 Trade-off Between Proportional Fairness and Max-Min Fairness