![ACEEE Int. J. on Control System and Instrumentation, Vol. 02, No. 01, Feb 2011
Performance Comparison of Rerouting Schemes of
Multi Protocol Label Switching Network.
Dr. Yogesh P Kosta 1, Minesh P Thaker 1,Bhaumik Nagar 2
1
VTP Dept of E&C Engg, Charotar Institute of Technology-Changa Gujarat, India
Email: mineshthaker.ec@ecchanga.ac.in ypkosta@yahoo.com
2
IT Centre, National Institute of Design, Paldi ,Ahmedabad, Gujarat, India
Email : bhaumik_n@nid.edu
Abstract—In this paper, we attempt to present a comparison
through rigorous studies (existing conventional models)
following software based modeling and verifications through
simulations in terms of various traffic parameters such as
packet loss, Recovery Time (Latency), reordering of packets
including recovery time for various widely used path recovery
models for the purpose of end-to-end recovery of LSPs in
MPLS domains using NS2 simulator.
Index Terms—MPLS .path -recovery, software modeling
and simulations ,traffic engineering, LSP (Label Switch
Path), end-to-end delay, recovery time.
I.INTRODUCTION
Traffic engineering is basically the process of Figure 1 Makam’s Model
optimizing the network(s) to maximize its overall performance
[10] that ultimately leads to improved efficiency. Multi Protocol B . HASKIN’s Model(Global Path repair with RNT)
Label Switching (MPLS) being one such technology for This scheme was introduced by Haskin
increasing the efficacy. This technology of forwarding data in the draft [2] and When global recovery is used as in
packets (of a fixed size labels) is based upon a pre-determined Makam’s model, the PSL has to be informed about a
path that the data packet needs to traverse, where thencoming failure in the working path before traffic can be switched
packet labels are examined to determine the next hop (or the over to the recovery path.
next part of the journey), the old label is then replaced with a
new one (label) and once again it is forwarded to the next hop,
and the journey continues till it reaches its destination.
The rest of this paper is organized as follows. Under
section 2, we briefly describe various path recovery models
that we have evaluated. Under section 3, we discuss the
important issues relating to previous studies on the subject
and explaining our work. Section 4 presents the Simulation
environment used for evaluation of the said protocols.
Section 5 helps consolidate our simulation results and our
specific observation. Finally, section 6 concludes the paper.
II. MPLS PATH RECOVERY AND REPAIRING MODELS
In this section we have briefly described the path
recovery schemes that we investigated. A detailed discussion
and comparison of path recovery schemes are availed at [ 1].
A different concern for Haskin’s model is the less
This work is being submitted in requirements against partial fulfillment
for the degree of ME in CSE of Gujarat University efficient use of resources, as the total length of the recovery
path gets longer than the original working path.
A. MAKAM’s Models (Global Path Repairing)
One of the first proposed models for MPLS recovery was C. Fast Reroute Model
presented by Makam in the draft [2].The model provides As shown in the fig 3 The fast reroute [4],[5],[6] shall
end- to-end protection for a LSP by setting up aglobal be used end-to-end then recovery paths needs to be pre-setup
recovery path between the ingress and egress LSR.
© 2011 ACEEE 17
DOI: 01.IJCSI.02.01.49](https://image.slidesharecdn.com/49-120911231202-phpapp01/85/Performance-Comparison-of-Rerouting-Schemes-of-Multi-Protocol-Label-Switching-Network-1-320.jpg)
![ACEEE Int. J. on Control System and Instrumentation, Vol. 02, No. 01, Feb 2011
for each link or node in the working path. This is one-to-one IV.SIMULATION ENVIRONMENT
backup model without merging for local path protection. In
this technique switch over time can be increased but more We have used network simulator ns2 for simulation,
resources are required. most widely used network simulator and freely
downloadable. In order to compare the results we use
D. Fast reroute one-to-one backup Model (Local Repair) CBR traffic flow and a UDP agent with the following
As shown in fig 4 the fast reroute [7],[8] one-to- characteristics: packet size = 1500 bits, source rate=
one technique a separate backup LSP, called a detour 512Kbps, burst time=0 and idle time =0, Tprop = 10msec,
LSP is computed for each LSR in a protected path. and BW(lsp)=BW(back)= BW( alt) = 10Mbps..
These detour LSPs are set up to use node recovery if
possible otherwise link recovery. To fully protect an LSP V. SIMULATION RESULTS AND OBSERVATION
that traverses N nodes, there could be as many as N-1
detours. A. . Packet Loss:
The graph in figure 5 shows the number of dropped
packets for the different models and depending on which
link that breaks. In both Haskins model and the fast reroute
model, traffic is switched onto a pre-setup backup path by
the LSR that detects the failure, so for both of those models
the only packets that are dropped are the ones dropped during
the failure detection time.
For Makam’s and the best effort model, the number
of dropped packages increases the further away from the
ingress LSR the failure occurs. This is because the FIS has
to be sent back to the ingress before traffic can be switched
to the backup path. In Makam’s model the number of
dropped packages is the same as for Haskin and fast reroute
only if the failure occurs on the link out from the ingress,
this is because in this case no FIS has to be sent before
traffic can be forwarded on the backup path. For the best
effort model, the number of dropped packages is large
because packages are dropped during the failure detection
time, the time for the FIS to be sent to the ingress router.
B.Full Restoration Time
The graph in fig 6 shows the Full Restoration Time,
measured from the last packet that was sent over the link
before it breaks is received by node 10, until the first packet
that is using the backup path is received by this node. The
time for Haskin’s and Makam’s models are the same, this
time is larger than fast reroute in all cases where the failure
occurs on a different link then on the ingress node. The
time increases for these models the further away from the
ingress node the failure occurs, this is because the FIS or
reversed traffic has to be sent upstream to the ingress before
it can be switched over on the global recovery path. For the
best effort model, the Full Restoration time is further
increased by the time to calculate and setup the backup path.
For the reroute model the Full Restoration time decreases
the closer to the egress LSR the failure occurs. This is partly
because the closer to the egress the failure occurs, the shorter
III. PREVIOUS WORK the new setup backup path can be.
In this section we analyze the most relevant C. Pre-reserved backup resources
previous studies concerning. Path recovery schemes for The chart in figure 7 shows the number of resources
MPLS network performance comparisons. Most of the reserved for backup traffic in the network before the failure
previous work is limited on performing simulations for occurs. Both the best effort and rerouting model setup the
MPLS networks with Packet loss and end-to-end delay.In backup path on demand after the failure has occurred, and
this paper, we observed and commented on the behavior therefore no backup resources are reserved before the
of each path repairing schemes. failure in those models. Makam’s models holds five
© 2011 ACEEE 18
DOI: 01.IJCSI.02.01.49](https://image.slidesharecdn.com/49-120911231202-phpapp01/85/Performance-Comparison-of-Rerouting-Schemes-of-Multi-Protocol-Label-Switching-Network-2-320.jpg)
![ACEEE Int. J. on Control System and Instrumentation, Vol. 02, No. 01, Feb 2011
resources reserved for the global backup path. Having compared the performance of above four rerouting
Haskin’smodel holds five resources for the global backup techniques Haskin model has got fast path recovery than
path and three for the reversed path, a total of eight. The Makams’s model. But uses more resources ,also Haskin has
fast reroute holds eight resources reserved low packet loss ratio, less number of disordered packets than
for the backup path. Mkam’s model. As far as two Fast reroute with one-to-one
techniques are concerned Fast reroute without merging was
D. Packets Disordered: much faster as far as restoration time is concerned with
Figure 8 presents the packet disorder result for respect to Fat reroute with merging. But as far as resources
the these schemes. Note that the packet disorder that we utilizations are concerned Fast reroute with merging uses
consider here is the disorder produced during the restoration less resources than Fast reroute without merging.
period which does not include the disorder produced by the If we compare global Path repair and Local Path Repair
retransmission of lost packets by a high level protocol (i.e., techniques the local path repair is good at fast rerouting where
TCP).Makam’s and other schemes do not introduce more as Global path is poor at fast rerouting or at full restoration
packet disorder but cause more packet losses. It is evident time.Similarly As far as Packet disordering is concern Haskin
from the figure that Fast reroute and Fast reroute 1to1 has is very good .
low packet disorder than Haskin.
VII. FUTURE WORK
In this simulation study, we have not used large
no of nodes and simulation time was 100s. Increasing both
or either of them will increase computational time which
was limited due to various reasons. Thus, in future we will
try to carry out more vigorous simulation so as to
gain better understanding of such networks and
subsequently helps in development of new protocols or
modification in existing protocols.
REFERENCES
[1] S.Makam, V.Sharma, K.Owens, C.Huang “Protection/
Restoration of MPLS Networks” draft-makam-mpls-protection-
00.txt October 1999
[2]D. Haskin, R.Krishnan
“A Method for Setting an Alternative Label Switched Paths to
Handle Fast Reroute” draft-haskin-mpls-fast-reroute-05.txt
November 2000
[3 E. Rosen, A.Vishwanathan, and R. Callon “Multi Protocol Label
Switching architecture.” RFC 3031, January 2001
[4]D. Haskin and R. Krishnan. “A Method for Setting an Alternative
Label Switched Paths to Handle Fast Reroute. Internet draft <draft-
haskin-mpls-fast-reroute-05.txt>, “ November 2000.
[5] Makam, V. Sharma, K. Owens, and C. Huang.” Protection/
Restoration of MPLS Networks. Internet draft <draft-makam-mpls-
protection-00.txt>” October 1999
[6 K. Owens, V. Sharma, S. Makam, and C. Huang. “A path
protection/Restoration Mechanism for MPLS Networks. internet
draft <draft-chang-mpls-protection-03.txt>” July 2001.
[7] S Shew “Fast Restoration of MPLS Label Switched Paths.,
Internet draft <draft-shew-lsp-restoration-00.txt>” October1999.
[8]L. Hundessa and J. Domingo. “Reliable and Fast Rerouting
mechanism for a protected label Switched Path.” Proceedings of
the IEEE GLOBECOM ’02, November 2002
[9] Nortel networks “introduction to multi protocol label switching”
white paper April 2001
Figure 9 Topology used for simulationVI. Conclusion
© 2011 ACEEE 19
DOI: 01.IJCSI.02.01.49](https://image.slidesharecdn.com/49-120911231202-phpapp01/85/Performance-Comparison-of-Rerouting-Schemes-of-Multi-Protocol-Label-Switching-Network-3-320.jpg)
Performance Comparison of Rerouting Schemes of Multi Protocol Label Switching Network
- 1. ACEEE Int. J. on Control System and Instrumentation, Vol. 02, No. 01, Feb 2011 Performance Comparison of Rerouting Schemes of Multi Protocol Label Switching Network. Dr. Yogesh P Kosta 1, Minesh P Thaker 1,Bhaumik Nagar 2 1 VTP Dept of E&C Engg, Charotar Institute of Technology-Changa Gujarat, India Email: mineshthaker.ec@ecchanga.ac.in ypkosta@yahoo.com 2 IT Centre, National Institute of Design, Paldi ,Ahmedabad, Gujarat, India Email : bhaumik_n@nid.edu Abstract—In this paper, we attempt to present a comparison through rigorous studies (existing conventional models) following software based modeling and verifications through simulations in terms of various traffic parameters such as packet loss, Recovery Time (Latency), reordering of packets including recovery time for various widely used path recovery models for the purpose of end-to-end recovery of LSPs in MPLS domains using NS2 simulator. Index Terms—MPLS .path -recovery, software modeling and simulations ,traffic engineering, LSP (Label Switch Path), end-to-end delay, recovery time. I.INTRODUCTION Traffic engineering is basically the process of Figure 1 Makam’s Model optimizing the network(s) to maximize its overall performance [10] that ultimately leads to improved efficiency. Multi Protocol B . HASKIN’s Model(Global Path repair with RNT) Label Switching (MPLS) being one such technology for This scheme was introduced by Haskin increasing the efficacy. This technology of forwarding data in the draft [2] and When global recovery is used as in packets (of a fixed size labels) is based upon a pre-determined Makam’s model, the PSL has to be informed about a path that the data packet needs to traverse, where thencoming failure in the working path before traffic can be switched packet labels are examined to determine the next hop (or the over to the recovery path. next part of the journey), the old label is then replaced with a new one (label) and once again it is forwarded to the next hop, and the journey continues till it reaches its destination. The rest of this paper is organized as follows. Under section 2, we briefly describe various path recovery models that we have evaluated. Under section 3, we discuss the important issues relating to previous studies on the subject and explaining our work. Section 4 presents the Simulation environment used for evaluation of the said protocols. Section 5 helps consolidate our simulation results and our specific observation. Finally, section 6 concludes the paper. II. MPLS PATH RECOVERY AND REPAIRING MODELS In this section we have briefly described the path recovery schemes that we investigated. A detailed discussion and comparison of path recovery schemes are availed at [ 1]. A different concern for Haskin’s model is the less This work is being submitted in requirements against partial fulfillment for the degree of ME in CSE of Gujarat University efficient use of resources, as the total length of the recovery path gets longer than the original working path. A. MAKAM’s Models (Global Path Repairing) One of the first proposed models for MPLS recovery was C. Fast Reroute Model presented by Makam in the draft [2].The model provides As shown in the fig 3 The fast reroute [4],[5],[6] shall end- to-end protection for a LSP by setting up aglobal be used end-to-end then recovery paths needs to be pre-setup recovery path between the ingress and egress LSR. © 2011 ACEEE 17 DOI: 01.IJCSI.02.01.49
- 2. ACEEE Int. J. on Control System and Instrumentation, Vol. 02, No. 01, Feb 2011 for each link or node in the working path. This is one-to-one IV.SIMULATION ENVIRONMENT backup model without merging for local path protection. In this technique switch over time can be increased but more We have used network simulator ns2 for simulation, resources are required. most widely used network simulator and freely downloadable. In order to compare the results we use D. Fast reroute one-to-one backup Model (Local Repair) CBR traffic flow and a UDP agent with the following As shown in fig 4 the fast reroute [7],[8] one-to- characteristics: packet size = 1500 bits, source rate= one technique a separate backup LSP, called a detour 512Kbps, burst time=0 and idle time =0, Tprop = 10msec, LSP is computed for each LSR in a protected path. and BW(lsp)=BW(back)= BW( alt) = 10Mbps.. These detour LSPs are set up to use node recovery if possible otherwise link recovery. To fully protect an LSP V. SIMULATION RESULTS AND OBSERVATION that traverses N nodes, there could be as many as N-1 detours. A. . Packet Loss: The graph in figure 5 shows the number of dropped packets for the different models and depending on which link that breaks. In both Haskins model and the fast reroute model, traffic is switched onto a pre-setup backup path by the LSR that detects the failure, so for both of those models the only packets that are dropped are the ones dropped during the failure detection time. For Makam’s and the best effort model, the number of dropped packages increases the further away from the ingress LSR the failure occurs. This is because the FIS has to be sent back to the ingress before traffic can be switched to the backup path. In Makam’s model the number of dropped packages is the same as for Haskin and fast reroute only if the failure occurs on the link out from the ingress, this is because in this case no FIS has to be sent before traffic can be forwarded on the backup path. For the best effort model, the number of dropped packages is large because packages are dropped during the failure detection time, the time for the FIS to be sent to the ingress router. B.Full Restoration Time The graph in fig 6 shows the Full Restoration Time, measured from the last packet that was sent over the link before it breaks is received by node 10, until the first packet that is using the backup path is received by this node. The time for Haskin’s and Makam’s models are the same, this time is larger than fast reroute in all cases where the failure occurs on a different link then on the ingress node. The time increases for these models the further away from the ingress node the failure occurs, this is because the FIS or reversed traffic has to be sent upstream to the ingress before it can be switched over on the global recovery path. For the best effort model, the Full Restoration time is further increased by the time to calculate and setup the backup path. For the reroute model the Full Restoration time decreases the closer to the egress LSR the failure occurs. This is partly because the closer to the egress the failure occurs, the shorter III. PREVIOUS WORK the new setup backup path can be. In this section we analyze the most relevant C. Pre-reserved backup resources previous studies concerning. Path recovery schemes for The chart in figure 7 shows the number of resources MPLS network performance comparisons. Most of the reserved for backup traffic in the network before the failure previous work is limited on performing simulations for occurs. Both the best effort and rerouting model setup the MPLS networks with Packet loss and end-to-end delay.In backup path on demand after the failure has occurred, and this paper, we observed and commented on the behavior therefore no backup resources are reserved before the of each path repairing schemes. failure in those models. Makam’s models holds five © 2011 ACEEE 18 DOI: 01.IJCSI.02.01.49
- 3. ACEEE Int. J. on Control System and Instrumentation, Vol. 02, No. 01, Feb 2011 resources reserved for the global backup path. Having compared the performance of above four rerouting Haskin’smodel holds five resources for the global backup techniques Haskin model has got fast path recovery than path and three for the reversed path, a total of eight. The Makams’s model. But uses more resources ,also Haskin has fast reroute holds eight resources reserved low packet loss ratio, less number of disordered packets than for the backup path. Mkam’s model. As far as two Fast reroute with one-to-one techniques are concerned Fast reroute without merging was D. Packets Disordered: much faster as far as restoration time is concerned with Figure 8 presents the packet disorder result for respect to Fat reroute with merging. But as far as resources the these schemes. Note that the packet disorder that we utilizations are concerned Fast reroute with merging uses consider here is the disorder produced during the restoration less resources than Fast reroute without merging. period which does not include the disorder produced by the If we compare global Path repair and Local Path Repair retransmission of lost packets by a high level protocol (i.e., techniques the local path repair is good at fast rerouting where TCP).Makam’s and other schemes do not introduce more as Global path is poor at fast rerouting or at full restoration packet disorder but cause more packet losses. It is evident time.Similarly As far as Packet disordering is concern Haskin from the figure that Fast reroute and Fast reroute 1to1 has is very good . low packet disorder than Haskin. VII. FUTURE WORK In this simulation study, we have not used large no of nodes and simulation time was 100s. Increasing both or either of them will increase computational time which was limited due to various reasons. Thus, in future we will try to carry out more vigorous simulation so as to gain better understanding of such networks and subsequently helps in development of new protocols or modification in existing protocols. REFERENCES [1] S.Makam, V.Sharma, K.Owens, C.Huang “Protection/ Restoration of MPLS Networks” draft-makam-mpls-protection- 00.txt October 1999 [2]D. Haskin, R.Krishnan “A Method for Setting an Alternative Label Switched Paths to Handle Fast Reroute” draft-haskin-mpls-fast-reroute-05.txt November 2000 [3 E. Rosen, A.Vishwanathan, and R. Callon “Multi Protocol Label Switching architecture.” RFC 3031, January 2001 [4]D. Haskin and R. Krishnan. “A Method for Setting an Alternative Label Switched Paths to Handle Fast Reroute. Internet draft <draft- haskin-mpls-fast-reroute-05.txt>, “ November 2000. [5] Makam, V. Sharma, K. Owens, and C. Huang.” Protection/ Restoration of MPLS Networks. Internet draft <draft-makam-mpls- protection-00.txt>” October 1999 [6 K. Owens, V. Sharma, S. Makam, and C. Huang. “A path protection/Restoration Mechanism for MPLS Networks. internet draft <draft-chang-mpls-protection-03.txt>” July 2001. [7] S Shew “Fast Restoration of MPLS Label Switched Paths., Internet draft <draft-shew-lsp-restoration-00.txt>” October1999. [8]L. Hundessa and J. Domingo. “Reliable and Fast Rerouting mechanism for a protected label Switched Path.” Proceedings of the IEEE GLOBECOM ’02, November 2002 [9] Nortel networks “introduction to multi protocol label switching” white paper April 2001 Figure 9 Topology used for simulationVI. Conclusion © 2011 ACEEE 19 DOI: 01.IJCSI.02.01.49