Figures 6.4 and 6.5 illustrate the average end-to-end delay of the proposed protocol (indicated by Gamma in these figures) compared to the IEEE 802.15.4 standard under different network configurations.
Figure 6.4 Average end-to-end delay of Gamma CSMA-CA and IEEE 802.15.4 ; Gamma ; IEEE 802.15.4 ; IEEE 802.15.4 ; Gamma ; IEEE 802.15.4 ; IEEE 802.15.4 ; Gamma ; Gamma ; IEEE 802.15.4 ; Gamma ; Gamma ; IEEE 802.15.4 A ver age end -to -en d delay (m il li secon ds ) Number of hops, h
M. Baz, PhD Thesis, University of York 2014
Figure 6.5 Average end-to-end delay of Gamma CSMA-CA and IEEE 802.15.4 under heavy tailed inter-arrival distributions
The results of figures 6.4 and 6.5 demonstrate that, under all cases, the proposed protocol provides the minimal average end-to-end delay readings compared to IEEE 802.15.4. These characteristics are attributed to the ability of the proposed protocol to manage the contention parameters of each node according to its traffic demands which in turn facilitates delivering timely packets. Another main reason for the significant reduction in the average end-to-end delay of the proposed protocol is its collision resolution mechanism that enables a node to resolve the underlying causes of collisions and thereby this mechanism improves the successful traffic flow across the network. Conversely in the Binary Exponential Back-off based protocols (i.e., the IEEE 802.15.4 standards), each node generates its back-off intervals according to a uniform distribution that does not reflect the traffic patterns of the multihop networks. Moreover, this protocol doubles the back-off windows whenever a node fails to access the channel which leads to defer transmission for long periods. Another important reason for the substantial increase in the average end-to-end delay of the IEEE 802.15.4 CSMA-CA protocol over number of hops is its collision resolution mechanism. This protocol requires colliding nodes to reset their back-off windows to the same initial value (which is typically small) and then precede their contention to access the channel. This mechanism magnifies the end-to-end delay as reducing the back-off window of colliding nodes increases the probability that two or more members of these nodes commence their transmission simultaneously which leads to increase the collision probabilities and the end-to-end delay.
; Gamma ; IEEE 802.15.4 ; IEEE 802.15.4 ; Gamma ; IEEE 802.15.4 ; IEEE 802.15.4 ; Gamma ; Gamma ; IEEE 802.15.4 ; Gamma ; Gamma ; IEEE 802.15.4 Number of hops, h A ver age end -to -en d delay (m il li secon ds )
M. Baz, PhD Thesis, University of York 2014
Deep insights into the characteristics of the end-to-end delay over multihop networks can be obtained by measuring its Squared Coefficient of Variation (SCV), i.e., the ratio between the variance and the squared average of the end-to-end delay. The SCV metric is used to quantify the variability of the end-to-end delay, the higher SCV the greater dispersion of the end-to-end delay around its average value and vice versa. Figure 6.6 shows the characteristics of this metric for both the proposed and the IEEE 802.15.4 CSMA-CA protocols under different network configurations.
Figure 6.6 Squared coefficient of variation of end-to end delay of Gamma CSMA-CA and IEEE 802.15.4
Results of figure 6.6 demonstrate that the IEEE 802.15.4 CSMA-CA protocol provides the highest level of variations under all scenarios. For instance, when there is on average two mutually independent paths between non-neighbouring nodes, the average number of nodes per channel is 10 each of which generates its inter-arrival times between packets according to Constant Bit Rate (CBR) with low rate (0.001packet/slot) and these packets are distributed evenly to other nodes, i.e., the case ⟨2,10, CBR, 1 × 10−3, U⟩. Under these conditions, the SCV of the end-to-end delay provided by the IEEE 802.15.4 CSMA-CA protocol increases dramatically
Number of hops, h Sq uar ed Coef fici ent of V ariat ion of end -to -en d delay ; Gamma ; IEEE 802.15.4 ; IEEE 802.15.4 ; Gamma ; IEEE 802.15.4 ; IEEE 802.15.4 ; Gamma ; Gamma ; IEEE 802.15.4 ; Gamma ; Gamma ; IEEE 802.15.4
M. Baz, PhD Thesis, University of York 2014
from about 2 over single hop to more than 7 over 16 hops. Conversely, the squared coefficient of variation of the proposed protocol of this scenario does not exceed 3 over 16 hops. An explanation for this characteristic can be acquired by considering the fact that the BEB mechanism specified in IEEE 802.15.4 CSMA-CA is a power law process which suffers from high variation due to its weak ergodicity [209]. Moreover, this behaviour is the underlying causes of being the medium access distribution of IEEE 802.15.4 CSMA-CA is heavy tailed. On the other hand, the proposed protocol is based on a Gamma process which has a very low level of variation [200]. For the same reason it can be seen that using the EXP or WBL distributions to generate the inter-arrival times between packets leads to great increases in SCV of the IEEE 802.15.4 CSMA-CA protocol compared to slight increases in SCV of the proposed protocol. This characteristic is attributed to the ability of the proposed protocol to mitigate the variation of different inter-arrival distributions, as the Gamma distribution is the conjugate prior for these inter-arrival distributions.
Figure 6.6 also shows that reducing the connectivity degree or sending all packets to the same final destination (indicated by “A” in the last element of tuple) magnifies SCV of IEEE 802.15.4 greatly whereas it keeps the SCV of the proposed protocol about the same. This is attributed mainly to that reducing the connectivity degree or using the "A" routing policy lessens the routing redundancy which in turn increases the collision probability of the IEEE 802.15.4 protocol and variability of the end-to- end delay. By contrast, the proposed protocol is able to mitigate the collision probability over multihop links and hence it can reduce the readings of the SCV metric. Considering the relationships between SCV and probability of collision and noting that increasing the traffic intensities or average number of nodes per channel increases the probability collision justify the high variations of the IEEE 802.15.4 compared to the proposed protocol.
6-7-2 Throughput
Figure 6.7 illustrates the throughput of the proposed and IEEE 802.15.4 CSMA-CA protocols versus number of hops for different scenarios. In general, it can be seen that the characteristics of throughput follow the same trends as the average end-to- end delay in terms of the effects of the connectivity degrees, average number of nodes per channel, traffic intensity and routing policies. It can be seen also from the results of this figure that the throughput of the IEEE 802.15.4 CSMA-CA protocol decreases sharply with increasing number of hops. The key reasons for this characteristic are the increasing of the collision probability and of the variation of the IEEE 802.15.4 protocol over multihop; such increasing minimises the channel utilisation and yields low throughput.
M. Baz, PhD Thesis, University of York 2014 Figure 6.7 Throughput of Gamma CSMA-CA and IEEE 802.15.4
6-7-3 Energy
An assessment for the average energy consumed per packet of both IEEE 802.15.4 and the proposed protocols is presented in figure 6.8. This figure shows that the IEEE 802.15.4 protocol consumes the highest level of energy compared to the proposed protocol. This is attributed to that the service times of IEEE 802.15.4 are much higher than their peers in the proposed protocol which in turn requires a node to contend to access for the channel for longer periods. A further support for the high amount of the energy consumed in the contention procedures of the IEEE 802.15.4 protocol has been introduced in [251]. This study shows that nodes operating in a saturated single hop network consume around 25 % of their energy just in contention to access the channel. The increasing in energy per packet over multihop links is due to the increasing in the collision probability over these links, hence considering the
Number of hops, h Thr ou ghp ut ; Gamma ; IEEE 802.15.4 ; IEEE 802.15.4 ; Gamma ; IEEE 802.15.4 ; IEEE 802.15.4 ; Gamma ; Gamma ; IEEE 802.15.4 ; Gamma ; Gamma ; IEEE 802.15.4
M. Baz, PhD Thesis, University of York 2014
ability of the proposed protocol to alleviate such probability explains the low energy expenditure of the proposed protocol compared to IEEE 802.15.4.
Figure 6.8 Average energy per packet of Gamma CSMA-CA and IEEE 802.15.4
6-8 Conclusion
This chapter has introduced the first main objective of this thesis which is to design an effective MAC protocol for multihop WSNs that can increase throughput, reduce end-to-end delay and save energy consumed during contention. The design methodology used in developing the proposed protocol is to provide a flexible contention scheme which reflects the dynamic characteristics of multihop networks, so each node can adjust its contention parameters to achieve the aforementioned goals. The new protocol replaces the uniform distribution that is used in the existing binary exponential back-off based protocol with a Gamma distribution which is more flexible and replaces the blind doubling of back-off windows by a more intelligent collision resolution algorithm. The simulation outcomes demonstrate the substantial benefits of the proposed protocol with respect to the IEEE 802.15.4 protocol. It is worth noting that although the proposed MAC protocol can improve the three
A ver age ener gy per pack et ( m ic ro joule s) Number of hops, h ; Gamma ; IEEE 802.15.4 ; IEEE 802.15.4 ; Gamma ; IEEE 802.15.4 ; IEEE 802.15.4 ; Gamma ; Gamma ; IEEE 802.15.4 ; Gamma ; Gamma ; IEEE 802.15.4
M. Baz, PhD Thesis, University of York 2014
aforementioned performance metrics compared to one of the state-of-the-art MAC protocol, this does not mean that the proposed protocol is always able to optimise all three metrics simultaneously. The protocol exploits the fact that it enables nodes to reduce the time required to deliver successful packets, can save contention energy (which prolongs lifetimes) ,and can improve the channel occupancy (which enhances the throughput).
M. Baz, PhD Thesis, University of York 2014