Simulation Result
Chapter 5 Simulation Result
n this section, we show the performance of the MUMO scheme and the comparisons with other grooming schemes in the same configuration. We also reveal the performance of the MUMO scheme with respect to various parameters. In the network we simulated, there are two fibers between any two adjacent nodes and each fiber contains 20 wavelengths. The capacity of each wavelength is OC-48 (2.5G/s). The MUMO scheme was compared with other schemes in different cases. First, we compare the performance of the MUMO and other schemes as the traffic load increases. Then, we also compare the performance as the ratio of multicast calls increases. Finally, we compare the performance as the maximum number of destinations of a multicast call increases. The simulation results of the comparisons in these cases are presented and discussed.
I
After discussing simulation results of the comparisons, we also reveal the performance evaluation of various network parameters. Four kinds of network parameters are used to compare the performance. We evaluate the performance with various mean service rate
μ
m. The difference between different mean service rates will be presented and discussed. Then, we evaluate the performance with various number of network nodes N. We will discuss the effect of number of network nodes. We also evaluate the performance with various ratio of the probability among difference type of traffic. Finally, the performance withdifferent
k
max is simulated.Before illustrating the comparison among the MUMO and other conventional schemes, we give a brief introduction of the three schemes. The MUMO scheme is compared with the two schemes: Single-Hop (SH) and Hybrid Multi-Hop (HYMH). When a new call arrivals at the ring network, SH will decide the route by MST routing scheme. Then, the SH scheme checks if any existing lightpath with available bandwidth has the same source and destinations as the source and destinations of the new call. If such a lightpath is found, the new call will be groomed with the lightpath; else it tries to set up new lightpaths for the new call. If the scheme even cannot setup new lightpath, the new call will be blocked.
The HYMH will also decide the route by MST, but in the scheme, a new call can be served by combination of multiple provisioned and un-provisioned lightpaths. The scheme searches for a lightpath that has the same destinations as the destinations of the new call.
The lightpath is called “to destination lightpath＂ (TDLP) . If TDLP is found, the scheme will search the other lightpaths or set up new lightpaths from the source of new call to the source of TDLP. The lightpath is called “from-source lightpath” (FSLP). If TDLP is not found, the scheme will setup new lightpaths for the new call.
Two performance measurements for MUMO, SH and HYMH are revealed. The first is about the new call blocking rate of the three schemes. The second is about the utilization efficiency of the used wavelength of the three schemes. According to these issues, there are two kinds of graphs in the simulation.
In the following figures from Fig. 5.1 to Fig. 5.6, the network parameters: W=20, N=20, C=48,
k
max=10,μ
m=0.05, R= {1, 3, 12, 16}, ( ,p p1 2,p3,p4)= (0.25, 0.25, 0.25, 0.25) is considered. As for the multicast ratio, the arrival rate and the max number of destination of a multicast call, we will give these parameters for each different figure. Fig. 5.1 and Fig.5.2 show the blocking probability and the utilization efficiency of the used wavelength
0 2 4 6 8 10 12 14 16 18 20 0
0.1 0.2 0.3 0.4 0.5 0.6 0.7
Arrival Rate(call/sec)
Blocking Probability
HYMH (20,20,10) SH (20,20,10) MUMO (20,20,10)
Figure 5.1 The blocking probability versus traffic loads
0 2 4 6 8 10 12 14 16 18 20
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Arrival Rate(call/sec)
Utilization
HYMH (20,20,10) SH (20,20,10) MUMO (20,20,10)
Figure 5.2 The utilization of used wavelength versus traffic load
versus traffic load, respectively, where the ratio of multicast call is 0.5 and the max number of destination of a multicast call is 19. It can be seen that the blocking probability of the MUMO scheme is much less than that of the other two schemes and the utilization efficiency of the MUMO scheme is much higher that of the others. It is because the MUMO scheme can partially groom the multicast traffic. It is easier to find lightpaths to groom a multicast call by the help of partially grooming. In addition, the other two schemes must find a lightpath which has the same destinations as the destinations of the new call while the MUMO scheme does not need. Moreover, since the routing approach of MUMO can always find a route with minimum hops according to the link state, the MUMO scheme can avoid choose the route that cannot satisfy the new call request. The blocking probability can be reduced by the routing approach of the MUMO scheme. On the other hand, the routing scheme of the other two schemes cannot adaptively choose the route, and the blocking probability is then larger.
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Multicast Ratio
Blocking Probability
HYMH (20,20,10) SH (20,20,10) MUMO (20,20,10)
Figure 5.3 The blocking probability versus multicast ratio
Fig. 5.3 and Fig. 5.4 show the blocking probability and utilization of used wavelength versus the ratio of multicast calls, respectively, where, The arrival rate is 10 (call/sec) and the max number of destinations of a multicast call is 19. From Fig. 5.3, it can be seen that the blocking probabilities increase as the multicast ratio increases, but that of the MUMO scheme increases much slower than that of the other two schemes. It is because that the SH and the HYMH cannot groom partially, they cannot deal with the multicast traffic efficiently. Beside, as mentioned above, the MUMO scheme only needs to find the lightpath that has the same terminating node; it helps a multicast call finding lightpaths to groom. From Fig. 5.3, it also can be found that the HYMH can have good performance when all traffics are unicast. As the multicast ratio increases, the blocking probability of HYMH increases rapidly. When the multicast ratio up to 0.8, the blocking probability of HYMH even higher than the blocking probability of SH. It results from that the HYMH may serve a new call with provisioned and un-provisioned lightpaths and when a TDLP is
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65
Multicast Ratio
Utilization
HYMH (20,20,10) SH (20,20,10) MUMO (20,20,10)
Figure 5.4 The utilization of used wavelength versus multicast ratio
found, the HYMH scheme may setup a new lightpath from the source to the source of TDLP. There must be only one destination on the new lightpath. Since the HYMH scheme need to find a lightpath has the same destinations as the destinations of the new call, these lightpaths may be groomed hardly when the multicast ratio is high. These lightpaths will waste the capacity and even result in that a lightpath only used by a single call, and the blocking probability of the HYMH scheme is then greater than that of the SH scheme when multicast ratio is high. From Fig. 5.4, similar condition can be seen that the utilization of HYMH decreases rapidly and finally becomes similar to the utilization of SH.
As for the MUMO scheme, it has good performance no matter in which multicast ratio.
Fig. 5.5 and Fig. 5.6 reveal that the blocking probability and the utilization of used wavelength versus the maximum number of destination of a multicast call, respectively, where arrival rate is 10 (call/sec) and the ratio of multicast calls is 0.5. As the max number of destinations of a multicast call increases, it becomes hard for the SH scheme and
2 4 6 8 10 12 14 16 18 20
0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
Max Destination Number of a Multicast Call
Blocking Probability
HYMH (20,20,10) SH (20,20,10) MUMO (20,20,10)
Figure 5.5 The blocking probability versus number of max destinations
HYMH scheme to find a lightpath that has the same destinations as the destinations of new call. Then, the blocking probabilities of the two schemes increase slowly. As for the MUMO scheme, from Fig.5.5, when the number of destinations is less than 11, the blocking probability of the MUMO scheme also increases. But after the number of destinations is larger than 11, the blocking probability of the MUMO scheme starts to decrease. It is because when the number of destinations large enough, the new call may be usually divided into two sub-calls; it helps the new call finding lightpaths to groom.
Beside, when the number of destinations large enough, it is easier for the MUMO scheme to search a partial-destination lightpath. It is also the reason make the blocking probability of the MUMO scheme decreases as the max number of destination of a multicast call is large. The utilization of used wavelength is shown in Fig. 5.6. Since the blocking probability of the HYMH scheme increases as number of destination increases, the utilization of the HYMH scheme decreases. The utilization of MUMO increases as the.
2 4 6 8 10 12 14 16 18 20
0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65
Max Destination Number of a Multicast Call
Utilization
HYMH (20,20,10) SH (20,20,10) MUMO (20,20,10)
Figure 5.6 The utilization of used wavelength versus max number of destinations
destination large enough since the blocking probability decreases. Beside, because of partially grooming, a new call may be served by multiple lightpaths. Thus, a new call may increase the utilization of multiple lightpaths. For the MUMO scheme, the utilization of large number of destinations is even higher than the utilization of small number of destinations.
After we discuss the performance comparisons of the three schemes in the fixed environment, the performance evaluations with various network parameters setting are then discussed and presented. In the following figures, most of the network parameters are the same as above except for the parameter which used to evaluate the difference of performance.
Fig. 5.7 and Fig. 5.8 show the blocking probability and the utilization of used wavelength with different mean service rate, respectively. Two mean service rate: 0.05 and 0.02 are simulated for the HYMH scheme and the MUMO schemes. It can be found that the blocking probability of low service rate is higher than high service rate. As the service rate is low, the bandwidth is occupied by a call for a long time. Then, the low service rate will result in that it is hard to arrange a new call into the network. Although MUMO always has better performance, the gap of blocking probability between the MUMO scheme and the HYMH scheme decreases as the service rate is low. It is because that the calls, which successfully arranged by the MUMO scheme, are much more than that of the HYMH scheme, the blocking probability of the MUMO scheme will increase more as the service rate is low. About the utilization, since low service rate has long service time, the utilization of used wavelength is higher. However, the gap between the two schemes also decrease as the service rate is low. It is because that the utilization of the MUMO scheme is almost saturated even the service rate is high, so the increased utilization is limited.
2 4 6 8 10 12 14 16 18 20
Figure 5.7 The blocking probability with different mean service rate
2 4 6 8 10 12 14 16 18 20 Figure 5.8 The utilization of used wavelength with different mean service rate
In Fig. 5.9 and Fig. 5.10, the blocking probability and the utilization of used wavelength in three different numbers of network nodes are simulated: 16, 20 and 24. As the number of network nodes increases to 24, the blocking probabilities of the two schemes both increase slightly. Since the destinations of a multicast call can be all nodes in the network except for the source, the maximum number of destinations of a multicast will also increases as the number of network nodes increases. As the result, it is hard to find lightpaths to groom for the two schemes, and the blocking probability then increases.
Conversely, the blocking probabilities of the two schemes decrease as the number of network nodes decreases to 16. As for the utilization of used wavelength in various number of network nodes, the utilization of used wavelength decreases in the large number of network nodes since the blocking probability increases. Nevertheless, the gap of utilization between each number of network nodes is very small. The reason is that a new call may be served by more number of lightpaths when the number of network nodes
2 4 6 8 10 12 14 16 18 20
0 0.1 0.2 0.3 0.4 0.5
Arrival Rate(call/sec)
Blocking Probability
MUMO (20,20,10) MUMO (24,20,8) MUMO (16,20,5) HYMH (20,20,10) HYMH (24,20,8) HYMH (16,20,5)
Figure 5.9 The blocking probability in different number of network nodes
2 4 6 8 10 12 14 16 18 20 Figure 5.10 The utilization of used wavelength in different number of network nodes
increases. Therefore, the gap of utilization between each number of network nodes will be reduced.
The Fig. 5.11 and Fig. 5.12 show the blocking probability and utilization of used wavelength with different probability of the type of new request, respectively. Two probabilities of the type of new request are simulated. One is (0.4, 0.3, 0.2, 0.1), which most of the new requests are small and the other one is (0.25, 0.25, 0.25, 0.25), which the probabilities of all type of new request are equal. From Fig. 5.11, the blocking probabilities decrease as most of the new requests are small. Beside, the gap of blocking probability between the MUMO scheme and the HYMH scheme becomes larger as most of the new requests are small. For the MUMO scheme, the blocking probability decreases as most of the new requests are small because the load of the network becomes very low.
However, for the HYMH scheme, the blocking probabilities of the two probabilities
1, 2, 3, 4
( p p p p )
2 4 6 8 10 12 14 16 18 20 Figure 5.11 The blocking probability with different probability of the type of new request
2 4 6 8 10 12 14 16 18 20
Figure 5.12 The utilization of used wavelength with different probability of the type of request
are similar because the HYMH scheme cannot efficiently arrange the new call into the network no matter which probability of the type of new call. As for the utilization of used wavelength, the utilization of the MUMO scheme always performs better than the utilization of the HYMH scheme.
Finally, the simulation result with different are shown in Fig. 5.13 and Fig. 5.14, respectively. We simulated with three different maximum number of destination on a lightpath. In Fig. 5.13, the blocking probability of small decreases for the MUMO scheme, but the blocking probability of small increases for the HYMH scheme. As is small, more lightpaths may be used to serve a new multicast call. Beside, the distance of the lightpath with small may be shorter than the distance of the lightpath with large . For the MUMO scheme, it will be easier to arrange a new call into the network as is small because of partially grooming. However, for the HYMH scheme, it become hard to find lightpath s to groom since it also become hard to find a lightpath that has the same destinations as the destinations of a multicast call. It is the reason which the blocking probability of the two schemes vary conversely as becomes small.
From the Fig. 5.14, we can see the utilization of the MUMO scheme become higher as become small. It verifies that the MUMO scheme arrange the new call more efficiently with small again. Although the blocking probability of the HYMH scheme increases, the utilization of the HYMH also increases as is small. The reason is that the new call will be served by more lightpaths in the HYMH scheme as is small; the average utilization will increase although the blocking probability increases.
2 4 6 8 10 12 14 16 18 20
Figure 5.13 The blocking probability with different kmax
2 4 6 8 10 12 14 16 18 20
Figure 5.14 The utilization of used wavelength with different kmax
From the simulation result, the gap of the blocking ratio between the MUMO and the SH scheme and the HYMH scheme are very large. The improvement of blocking probability is about 70% and that of utilization is about 50%. In some cases, the blocking ratio of other two schemes even twice as which of the MUMO scheme. In addition, the performance of the MUMO scheme does not deteriorate in various environments, including mean service time, different network size, probability of the type of new call request and the max number of destinations on a lightpath. Therefore, the MUMO scheme is an efficient and feasible scheme.