7 - Results and Analysis of the Second Scenario vs the
Third Scenario
7.1 - Overview
In this chapter a comparison between the second and the third scenario will be made,
trying to show the differences between them and the improvements brought by the last
scenario. Like in the previous scenario, in the third scenario a DiffServ treatment will be applied to the flows together with Constraint based Routing to obtain a dynamic traffic
engineering, but the difference is in the use of a dynamic bandwidth reservation
(semi-dynamic as regards UDP) depending on the traffic pattern, as has been explained in
section 5.6.
The following sections will analyze in the same order the same variables analyzed in the previous chapter. The examination of the various data will require a constant glance to the
ones in the chapter 6, not represented here to avoid an excessive redundancy.
In section 7.2, three simulations will be analyzed, overloading at first the Gold UDP traffic,
then the Silver UDP and at last both Gold UDP and Silver TCP traffic.
In section 7.3, many simulations will be used to analyze trough a series of graphs the TCP traffic pattern.
Also for the third scenario, it will be shown that the objectives specified by the SLA are
fully respected.
It will also be shown that the dynamic bandwidth will offer levels of quality of service, for
different load patterns, more consistent with the importance of the traffic classes and to what clients are expected to pay.
7.2 - UDP Varying
7.2.1 – Simulation 1 - Gold class overloaded, Silver class underloaded
.In the third scenario, as explained in chapter 5, through the Admission Control, the Gold
UDP exceeding bandwidth is dynamically reassigned till filling the Silver availability (if
there is availability). Looking at Tab 6.1, Tab 6.2 and Tab 7.1, it can be noticed the
different behaviors of the two scenarios.
A good example for a better understanding of the dynamic reassignment mechanism is
still the aggregated flow between nodes 13 and 17 (remembering that the normal load
assignment is 300 Kbit/s for each Gold aggregated flow, while it is 150 Kbit/s for each
Silver one). From Table 6.1, there is 450 Kbit/s for this Gold aggregated flow. As a
consequence there is an exceeding bandwidth for this flow of 150 Kbit/s, whereas the bandwidth that can be used from the corresponding Silver aggregated flow is just
100 Kbit/s. In the third scenario, the ingress Label Edge Router of the MPLS domain
reserves a bandwidth of 400 Kbit/s to the Gold aggregated flow, leaving 50 Kbit/s to the
Silver and declassing the remaining 50 Kbit/s to the best effort class.
Traffic Rate Full Duplex (CBR Sources) Gold Traffic Kbit/sec
From / To Node 17 Node 16 Node 15 Total Rate
Node 14 350 350 300 1000
Node 13 400 450 350 1200
Node 12 300 400 300 1000
Total Rate 1050 1200 950 3200
Silver Traffic Kbit/sec
From / To Node 17 Node 16 Node 15 Total Rate
Node 14 100 100 50 250
Node 13 50 0 100 150
Total Rate 150 100 200 450 Bronze Traffic Kbit/sec
From / To Node 17 Node 16 Node 15 Total Rate
Node 14 100 50 50 200
Node 13 100 0 100 200
Node 12 0 0 50 50
Total Rate 200 50 200 450
Tab 7.1 – Traffic effective with the third scenario
We obtain an average rate of 356 Kbit/s for Gold, 50 Kbit/s for Silver and 50 Kbit/s for
Bronze. The total rate is 4100 Kbit/s.
Percentage of Packets Lost for single service class
Gold Silver Bronze
0% 0% 17%
Percentage of Packets Received for single service class
Gold Silver Bronze
100% 100% 83%
Tab 7.2 – UDP Packet Loss scenario 3
Goodput Gold
3200 Kbit/s
Goodput Silver Total rate
450 Kbit/s 4023,5 Kbit/s
Goodput Bronze
373,5 Kbit/s
Tab 7.3 – UDP Goodput scenario 3
Average delay ms
Gold Flows Silver Flows Bronze Flows
37,16 36,90 56,47
Standard Deviation ms
14,00 10,52 18,39
Tab 7.4 – UDP Delay scenario 3
Average jitter ms
Gold Flows Silver Flows Bronze Flows
2,35 5,76 10,87
Standard Deviation ms
0,76 4,78 10,31
From Tab 7.2, 7.3, 6.5, 6.6, it is shown that the percentage of packets lost is quite similar between the two scenarios. However, as regards the goodput, the Gold goodput is
increased in the third scenario, as less flows are downgraded to a lower traffic class,
increasing the Gold load. For the same reason, the Silver goodput is smaller in the third
scenario, as less traffic is downgraded to Silver, resulting in a lower load for this class.
The total goodput is similar in the second and third scenarios.
Also as regards the delay (Tab 7.4 and 6.8) and the jitter (Tab 7.5 and 6.10), the value for
the Gold delay is worse in the third scenario as compared to the second and the value for
the Silver delay is better because of the load difference, as explained in the previous
paragraph. The remaining values are similar, even if just a little bit worse.
As explained in the last part of section 5.6, the path assignment can be different between the second and the third scenario, leading to differences in the treatments and in the
results. For this reason, the results as regards Packet Loss, Delay and Jitter are (and will
also be in the next sections) sometimes a little bit better and sometimes a little bit worse
for the third scenario.
What is important is that this new scenario reduces the number of downgraded clients for
the UDP traffic, resulting in more satisfied customers.
7.2.1.1 - TCP Flows Rate
Total average rate Kbit/s total rate Kbit/s
Gold 51,57 3324
Silver 24,98 1019
Bronze 3,12 281
Total 4624
Tab 7.6 – TCP Flows Rate Scenario 3
tot reserved Kbit/s
Gold 2983
Silver 1067
In Tab 7.7 is represented the total bandwidth that is reserved to Gold TCP traffic and Silver TCP traffic. Tab 7.6 shows the simulation values for the total average rates in each
traffic class, that are close to those reserved. The average number of Gold and Silver TCP
active sources per aggregated flow are respectively 6.56 and 4.78, this leads to 331 Kbit/s
and 119 Kbit/s of average bandwidth reservation, multiplying by nine (i.e. the number of
TCP aggregated flows per single TCP class) we arrive to the values in Tab 7.7.
Looking at Tab 6.12 and at tab 7.6, it can be noticed that the service obtained is better for
the Gold TCP flows. In particular, the average rate obtained for a single Gold flow is
about the double than the average rate obtained for a single Silver flow. The dynamic
redistribution of bandwidth mechanism from this third scenario increased the Gold
bandwidth and reduced the Silver bandwidth so that the average rate for Gold flows is approximately double of the average rate for Silver flows, respecting the SLA.
7.2.2 – Simulation 2 - Gold class under loaded, Silver class overloaded.
In this second simulation, the Silver traffic is exceeding the reservation (Tab 6.13). This
implies no reassignment of the bandwidth for UDP traffic. The behavior is the same as in
the second scenario, that is: the admission control declasses the Silver exceeding traffic
into best effort traffic (Bronze).
For this reason Tab 7.8 and Tab 6.14 are identical.
Traffic Rate Full Duplex (CB R Sources) Gold Traffic Kbit/sec
From / To Node 17 Node 16 Node 15 Total Rate
Node 14 0 250 150 400
Node 13 250 150 150 550
Node 12 200 250 50 500
Total Rate 450 650 350 1450
From / To Node 17 Node 16 Node 15 Total Rate
Node 14 150 150 150 450
Node 13 150 150 150 450
Node 12 150 150 150 450
Total Rate 450 450 450 1350
Bronze Traffic Kbit/sec
From / To Node 17 Node 16 Node 15 Total Rate
Node 14 100 50 150 300
Node 13 200 50 50 300
Node 12 50 50 100 200
Total Rate 350 150 300 800
Tab 7.8 – Traffic effective with the third scenario
We obtain an average rate of 161 Kbit/s for Gold 150 Kbit/s for Silver and 89 Kbit/s for Bronze. The Total rate is 3600 Kbit/s. All values exactly equal to second scenario.
Percentage of Packets Lost for single service class
Gold Silver Bronze
0% 0% 34%
Percentage of Packets Received for single service class
Gold Silver Bronze
100% 100% 66%
Tab 7.9 – UDP Packet Loss Scenario 3
Goodput Gold
1450 Kbit/s
Goodput Silver Total rate
1350 Kbit/s 3328 Kbit/s
Goodput Bronze
528 Kbit/s
Tab 7.10 – UDP Goodput Scenario 3
Average delay ms
Gold Flows Silver Flows Bronze Flows
30,2 37,4 51,8
Standard Deviation ms
9,2 7,1 21,8
Tab 7.11 – UDP Delay Scenario 3
Average jitter ms
Gold Flows Silver Flows Bronze Flows
Standard Deviation ms
2,31 1,62 9,26
Tab 7.12 – UDP Jitter Scenario 3
As regards Tab 7.9 and Tab 7.10, the values are exactly the same as in the second scenario
(Tab 6.17 and 6.18). Also in this case the values obtained as regards the Delay and the
Jitter are similar for the two scenarios (although more Delay and Jitter values are a little
bit better for this third scenario).
7.2.2.1 - TCP Flows Rate
Total average rate Kbit/s Total rate Kbit/s
Gold 52,51 3149
Silver 25,58 1185
Bronze 3,86 347
Total 4681
Tab 7.13 - TCP Flows Rate Scenario 3
Total rate reserved Kbit/s
Gold 2983
Silver 1067
Tab 7.14 - TCP total bandwidth reserved to the first two classes
The total rate values are better than those reserved.
The service obtained is what we expected, that is an average rate per single Gold TCP
flow double (a little more than double in this case) than the average rate per Silver TCP flow.
7.2.3 – Simulation 3 - Gold and Silver classes overloaded.
Tab 6.25 shows the effective traffic for this third scenario, where both the Gold and the
Silver traffic are exceeding the reservation. As a consequence there is no reassignment of
the UDP bandwidth also in this case, and, also in this case, the behavior is the same as in the second scenario (Tab 6.26 and 7.21), that is: the admission control declasses the
exceeding Gold traffic and the exceeding Silver traffic to best effort traffic (no Silver bandwidth is available).
For this reason Tab 7.21 and Tab 6.26 are identical.
We obtain an average rate of 300 Kbit/s for Gold, 150 Kbit/s for Silver and 144 Kbit/s for
Bronze. The total rate is 5350 Kbit/s.
Traffic Rate Full Duplex (CBR Sources) Gold Traffic Kbit/sec
From / To Node 17 Node 16 Node 15 Total Rate
Node 14 300 300 300 900
Node 13 300 300 300 900
Node 12 300 300 300 900
Total Rate 900 900 900 2700
Silver Traffic Kbit/sec
From / To Node 17 Node 16 Node 15 Total Rate
Node 14 150 150 150 450
Node 13 150 150 150 450
Node 12 150 150 150 450
Total Rate 450 450 450 1350
Bronze Traffic Kbit/sec
From / To Node 17 Node 16 Node 15 Total Rate
Node 14 150 100 50 300
Node 13 250 250 250 750
Node 12 50 150 50 250
Total Rate 450 500 350 1300
Tab 7.15 – Traffic effective with the third scenario
Percentage of Packets Lost for single service class
Gold Silver Bronze
0% 0% 60%
Percentage of Packets Received for single service class
Gold Silver Bronze
100% 100% 40%
Goodput Gold
2700 Kbit/s
Goodput Silver Total rate
1350 Kbit/s 4570 Kbit/s
Goodput Bronze
520 Kbit/s
Tab 7.17 – UDP Goodput Scenario 3
Average delay ms
Gold Flows Silver Flows Bronze Flows
37,8 43,1 60,9
Standard Deviation ms
9,8 12,3 23,1
Tab 7.18 – UDP Delay Scenario 3
Average jitter ms
2,78 4,83 16,70
Standard Deviation ms
0,62 2,00 9,99
Tab 7.19 – UDP Jitter Scenario 3
In this simulation of the third scenario, the results for the UDP traffic (Tab 7.16, 7.17,
7.18, 7.19) are a little bit worse than in the second scenario (Tab 6.29, 6.30, 6.32, 6.34);
but the difference is too small to speak about worse results, moreover the results are still
very good if compared with those obtained in the first scenario.
7.2.3.1 - TCP Flows Rate
Total average rate Kbit/s Total rate Kbit/s
Gold 54,46 3111
Silver 26,78 1082
Bronze 1,21 109
Total 4302
Tab 7.20 - TCP Flows Rate Scenario 3
tot reserved Kbit/s
Gold 2983
Silver 1067
Like in the previous simulation, the values are very close to those reserved, and the service
obtained is better.
The service obtained is still improved, the average rate per single Gold TCP flow is the
double (a little more here too) than the average rate per Silver TCP flow.
If compared with the total TCP rate shown by the second scenario, we can see that it
increases by about 170 Kbit/s in the third scenario.
7.3 - TCP Varying
As in the previous chapter, in this set of simulations we will vary the TCP load leaving the
UDP fixed to heavily loaded values. This case is the ideal to show the differences between
the second and the third scenario, because this last scenario has been thought particularly
to improve the treatment of the Gold TCP traffic. Those differences are shown through the
use of 18 graphs regarding 6 simulations with the same characteristics as in the preceding
chapter.
In the following section, some considerations will be made about the third scenario
simulations in general. In the subsequent sections, the differences between the second
scenario and the third scenario will be analyzed in more detail, first fixing the number of
Silver TCP sources per aggregated flow to 1, 5 and 10, and varying the number of Gold
TCP Sources, then the number of Gold TCP Sources per aggregated flow will be fixed to
those numbers and the Silver will vary.
The values represented in the graphs are still the average rate per single class (Gold, Silver,
7.3.1 – Third scenario simulation results (general)
The fist thing that can be noticed looking at the Graphs from 7.1 till 7.6 in this third scenario,
is that the rates of both Gold values and Silver values decrease, differently from the second
scenario (Graph 6.2, 6.4, 6.6, 6.8, 6.10, 6.12), where only one of the two rates decreases.
But the most important thing shown in each Graph, as expected, is that the Gold rates are
always more or less (many times more, few times less) the double than the Silver rates.
Scenario 3 - 1 Silver TCP sources
0,00 50,00 100,00 150,00 200,00 250,00 300,00 1 2 3 4 5 6 7 8 9 10 Gold TCP sources Rate Kbit/s Gold 3 Silver 3 Bronze 3 Graph 7.1
Scenario 3 - 5 Silver TCP sources 0,00 20,00 40,00 60,00 80,00 100,00 120,00 140,00 160,00 180,00 1 2 3 4 5 6 7 8 9 10 Gold TCP sources Rate Kbit/s Gold 3 Silver 3 Bronze 3 Graph 7.2
Scenario 3 - 10 Silver TCP sources
0,00 20,00 40,00 60,00 80,00 100,00 120,00 1 2 3 4 5 6 7 8 9 10 Gold TCP sources Rate Kbit/s Gold 3 Silver 3 Bronze 3 Graph 7.3
Scenario 3 - 1 Gold TCP sources 0,00 50,00 100,00 150,00 200,00 250,00 300,00 1 2 3 4 5 6 7 8 9 10 Gold TCP sources Rate Kbit/s Gold 3 Silver 3 Bronze 3 Graph 7.4
Scenario 3 - 5 Gold TCP sources
0,00 10,00 20,00 30,00 40,00 50,00 60,00 70,00 80,00 90,00 100,00 1 2 3 4 5 6 7 8 9 10 Gold TCP sources Rate Kbit/s Gold 3 Silver 3 Bronze 3 Graph 7.5
Scenario 3 - 10 Gold TCP sources 0,00 5,00 10,00 15,00 20,00 25,00 30,00 35,00 40,00 45,00 50,00 1 2 3 4 5 6 7 8 9 10 Silver TCP sources Rate Kbit/s Gold 3 Silver 3 Bronze 3 Graph 7.6
7.3.2 – Silver class TCP fixed to 1, Gold class TCP varying
Looking at Graph 7.6, it is clear that in the third scenario the Gold rate is higher than in the
second scenario. In all the subsequent graphs the rate of the first scenario will be always
represented to remember the great improvements brought by both the second and the third
scenarios in comparison to the first.
From Graph 7.8, it can be noticed that the Silver rate of the third scenario is decreasing and
crosses the second one very soon. This fact is due to the increasing number of Gold sources
that are sharing the same bandwidth. When there are more Gold sources, the rate per Gold
source decreases, resulting in a decrease of the Silver rate, according to the specified
3 Scenarios Gold - 1 Silver TCP sources 0,00 50,00 100,00 150,00 200,00 250,00 300,00 1 2 3 4 5 6 7 8 9 10 Gold TCP sources Rate Kbit/s Gold 1 Gold 2 Gold 3 Graph 7.7
3 Scenarios Silver - 1 Silver TCP sources
0,00 20,00 40,00 60,00 80,00 100,00 120,00 140,00 160,00 180,00 1 2 3 4 5 6 7 8 9 10 Gold TCP sources Rate Kbit/s Silver 1 Silver 2 Silver 3 Graph 7.8
7.3.3 – Silver class TCP fixed to 5, Gold class TCP varying
Looking at Graph 7.9, it can be noticed that the Gold rate in the third scenario is higher than
the rate in the second scenario when there are many TCP Gold sources, while it is lower when
there are few Gold sources. The behavior of the two scenarios is more or less the same with 5
Gold sources. This is obvious because the same number of sources implies a total Gold
bandwidth assignment that is exactly the double of the Silver, and, as the total rate must be 450 Kbit/s, the total bandwidth assigned to the Gold traffic is 300 Kbit/s and to the Silver
traffic is 150 Kbit/s, that is the same assignment as in the second scenario.
Graph 7.10 shows that the Silver rate in the third scenario is decreasing and crosses the rate of
the second scenario for around 5 sources. As explained in the previous section, this happens
because the rate per Gold source decreases in the third scenario, and to keep the established proportion, the rate per Silver source decreases also in the third scenario.
The total bandwidth assignment is higher than 300 Kbit/s for the Gold (and lower than 150
Kbit/s for the Silver), if the number of Gold sources is higher than the number of the Silver
sources. For this reason, the values in the graphs are better for the Gold class when there are
more than 5 Gold sources, the opposite happens for the Silver class.
For this reason, for both the following graphs, it makes sense to speak about the crossing of
3 Scenarios Gold - 5 Silver TCP sources 0,00 50,00 100,00 150,00 200,00 250,00 1 2 3 4 5 6 7 8 9 10 Gold TCP sources Rate Kbit/s Gold 1 Gold 2 Gold 3 Graph 7.9
3 Scenarios Silver - 5 Silver TCP sources
0,00 10,00 20,00 30,00 40,00 50,00 60,00 1 2 3 4 5 6 7 8 9 10 Gold TCP sources Rate Kbit/s Silver 1 Silver 2 Silver 3 Graph 7.10
7.3.4 – Silver class TCP fixed to 10, Gold class TCP varying
Looking at Graph 7.11, it can be noticed that the Gold rate in the third scenario is always
lower than the rate in the second scenario independently of the number of TCP Gold sources.
But, even if higher, the behavior of the two scenarios is more or less the same with a high
number of Gold sources; while it is quite the same with 10 sources, because, as explained in
the previous section, the bandwidth assignment is the same in the two scenarios.
Graph 7.12 shows that the Silver rate in third scenario is decreasing but higher than in the
second scenario, this happens because with a number of Gold sources that is inferior to the
number of Silver sources, the proportion leads to a total Silver bandwidth assignment that is
higher than the 150 Kbit/s (and consequently lower than 300 Kbit/s for the Gold) obtained
with the second scenario, obviously similar values are reached with a high number of sources being quite the same with 10 sources.
3 Scenarios Gold - 10 Silver TCP sources
0,00 50,00 100,00 150,00 200,00 250,00 1 2 3 4 5 6 7 8 9 10 Gold TCP sources Rate Kbit/s Gold 1 Gold 2 Gold 3 Graph 7.11
3 Scenarios Silver - 10 Silver TCP sources 0,00 5,00 10,00 15,00 20,00 25,00 30,00 35,00 40,00 1 2 3 4 5 6 7 8 9 10 Gold TCP sources Rate Kbit/s Silver 1 Silver 2 Silver 3 Graph 7.12
7.3.5 – Gold class TCP fixed to 1, Silver class TCP varying
It can be observed in all the following Graphs, that what was the rate of the first and the
second graph in the previous paragraphs, in a certain sense, now is inverted. This happens
because now the number of Gold sources is fixed, while it is the number of Silver sources to
vary. As they are sharing the same bandwidth, when there are more Silver sources, the rate
per Silver source decreases, resulting in a decrease of the Gold rate, according to the specified
proportion between their rates.
As can be seen in Graph 7.13, in the third scenario, the Gold rate is lower than that rate in the
second scenario. The only exception is the simulation with one Gold source and one Silver,
that is the same as in Graph 7.7, where the value of the third scenario is higher.
Graph 7.14 shows that the Silver rate of the third scenario is decreasing and always higher
As explained in the previous section, this happens because with a number of Gold sources that is inferior to the number of Silver sources, the proportion leads to an assignment that is higher
than 150 Kbit/s for the total Silver bandwidth.
3 Scenarios Gold - 1 Gold TCP sources
0,00 50,00 100,00 150,00 200,00 250,00 300,00 1 2 3 4 5 6 7 8 9 10 Silver TCP sources Rate Kbit/s Gold 1 Gold 2 Gold 3 Graph 7.13
3 Scenarios Silver - 1 Gold TCP sources
0,00 20,00 40,00 60,00 80,00 100,00 120,00 140,00 160,00 180,00 1 2 3 4 5 6 7 8 9 10 Silver TCP sources Rate Kbit/s Silver 1 Silver 2 Silver 3 Graph 7.14
7.3.6 – Gold class TCP fixed to 5, Silver class TCP varying
Looking at Graph 7.15, it appears that in the third scenario the Gold rate is higher than in the
second scenario for a low number of Silver sources, while it is lower for a high number. The
intersection is placed near 5 sources, because, as should be well know now, 5 sources implies
the same bandwidth assignment between the third and the second scenario.
In Graph 7.16 the situation is inverted, the Silver rate of the third scenario is higher than in the
second scenario when the number of Silver sources is high, while it is lower when the number
is low. The intersection is still around five.
The reasons for the rate variations in the graphs are exactly the same as explained in section
7.3.3, with the only difference that here the Silver sources are varying, while the Gold sources
are fixed to 5.
3 Scenarios Gold - 5 Gold TCP sources
0,00 10,00 20,00 30,00 40,00 50,00 60,00 70,00 80,00 90,00 100,00 1 2 3 4 5 6 7 8 9 10 Silver TCP sources Rate Kbit/s Gold 1 Gold 2 Gold 3 Graph 7.15
3 Scenarios Silver - 5 Gold TCP sources 0,00 20,00 40,00 60,00 80,00 100,00 120,00 140,00 1 2 3 4 5 6 7 8 9 10 Silver TCP sources Rate Kbit/s Silver 1 Silver 2 Silver 3 Graph 7.16
7.3.7 – Gold class TCP fixed to 10, Silver class TCP varying
As can be noticed in Graph 7.17, thanks to the number of Gold sources that is always higher
than the number of Silver sources, the Gold rate in the third scenario is always higher than in
the second scenario. The values are near just towards 10 Silver TCP sources.
In Graph 7.18 it appears that the Silver rate of the third scenario is decreasing and always
lower than the rate of the second. The values are near for a high number of Silver TCP
sources.
Also in this case the reasons of the proximity of the two graphs for the same number of
sources are due to the same reservation made in both the cases of the second and the third scenario.
3 Scenarios Gold - 10 Gold TCP sources 0,00 5,00 10,00 15,00 20,00 25,00 30,00 35,00 40,00 45,00 50,00 1 2 3 4 5 6 7 8 9 10 Silver TCP sources Rate Kbit/s Gold 1 Gold 2 Gold 3 Graph 7.17
3 Scenarios Silver - 10 Gold TCP sources
0,00 20,00 40,00 60,00 80,00 100,00 120,00 140,00 1 2 3 4 5 6 7 8 9 10 Silver TCP sources Rate Kbit/s Silver 1 Silver 2 Silver 3 Graph 7.18