MPLS TE Over ATM
Deepak Gulla Nishant Tambe
{dsgulla, nptambe1}@wichita.edu
Electrical & Computer Engineering Department
Wichita State University,
1845 Fairmont, Wichita, Kansas, 67208
Abstract
The conventional method for managing the network bandwidth deployed in Internet Service Provider’s
(ISP) is the use of layer-2 technologies such as Asynchronous Transfer Mode (ATM), Frame Relay, etc.
This method uses the concept of virtual circuits to connect to the IP backbone. Although virtual circuits
fulfill this purpose, however, there are certain drawbacks to this approach, which include issues of
scalability, manageability, and costs. A simple link failure can cause many virtual circuits to fail, forcing
IP routing protocols to re-converge. Multi-protocol Label Switching (MPLS) is a more scalable,
manageable and cost effective solution used to forward packets through the network. In this paper, a
solution to the drawbacks observed with the virtual circuits is presented by deploying MPLS Traffic
Engineering (TE) over an ATM network with ATM Traffic Management (TM) capabilities. The real
time and non-real time traffic was taken into consideration for various simulations. Perceptual Speech
Quality Measurement (PSQM) scores were used to evaluate the quality of voice in case of real time
traffic. The results show that upon a link failure, the network switches onto the next available path
within a short period of time and with minimal packet loss; Thus, network connectivity is maintained. It
was also observed that the voice quality does not largely deteriorate, even with the network congested.
1. Introduction:
MPLS performs routing at edges of the network and switching at the core. It is a high performance method
for forwarding packets through the network. TE in general is a method of optimizing the utilization of
network resources. MPLS TE integrates TE capabilities of the layer-2 technologies with those of the layer-
3 technologies. ATM is an ITU-T standard capable of multiplexing different services including voice,
video and data. It is an evolving technology that is being adopted by most of the ISPs to manage their
network bandwidth. The objective of this study was to combine the positive aspects of ATM and MPLS
TE to obtain better network performance. Packet Generator (Pagent), Call Generator (Callgen) and PSQM
were some of the tools used to evaluate network performance and to determine better solutions to maintain
the same network connectivity in case of link failures.
2. MPLS (Multi-Protocol Label Switching):
MPLS is a multi-protocol because it can work with many protocols including OSPF, RSVP, LDP and BGP.
It is a high performance method for forwarding packets through the network. In the conventional IP
forwarding, the packets at each interface have to perform a route lookup to determine their destination.
MPLS provides a mechanism to perform label switching which is a new technique. It forwards packets
using the concept of label swapping, in which packets carry a label of fixed length. There is enough
information present in this label to tell the packet how to process the data. It includes information such as
the destination, precedence by applications, users, groups, Virtual Private Network (VPN) Membership,
QOS via IP Precedence class of service (COS), and the packet route chosen by IP routing protocols and
possibly by Traffic Engineering.
MPLS is based on IP protocols. It integrates the label-switching paradigm with network layer routing,
which improves the performance of network layer routing. It provides greater flexibility in the delivery of
routing services by allowing new routing services to be added without a change to the forwarding
paradigm. The forwarding is done through the labels, which are a fixed length entity of four bytes. Labels
are assigned to the packets as they enter the MPLS domain at the Ingress routers (Edge routers). All the
MPLS nodes forward packets/cells based on label value (and not on IP information). When a packet enters
a MPLS network, it is classified and encoded as a label. This packet is not analyzed further in the domain.
1
Only the intermediate routers within the MPLS domain take part in label swapping. The MPLS forwarding
example and Generic Label format is shown below in figures 1 and 2 respectively. MPLS has two
functional components of network layer routing. They are the Control component and the Forwarding
component. The Control component is based on IP routing and Forwarding is based on concept of labels.
Data Link IP Header
Layer Header Shim Header Layer 3 Packet data
Label Exp S Time to Live
20 bits 3 bits 1 bit 8 bits
Figure 1: MPLS Label Format
Network 130.10.X.X / 24
Intf Label Dest Intf Label Intf Label Intf
Dest
In In out out In in out
2 0.40 56.2 1 0.30 2 0.30 56.2 1
56.1 Request 56.2 Request 56.2
2 1 2 1 2 1
Mapping 0.40 Mapping 0.30 56.2
3
Intf Dest Intf Label
In out out
2 56.2 1 0.40 2 1
56.3
Figure 2: MPLS Forwarding Paradigm
The label is present between the layer 2 and the layer 3 headers. The labels have only local significance. In
case of layer 2 technologies, this label is replaced by VPI/VCI value like in ATM, and in case of Frame
Relay, the DLCI field replaces it.
3. Traffic Engineering (TE):
Traffic Engineering is essential for both the service provider and ISP backbones to support maximum
utilization of transmission capacity, and for their networks to be very resilient, so that they can withstand
link or node failures. Traffic Engineering enables ISPs to route network traffic through multiple paths, thus,
offering better services to users in terms of throughput and delay.
MPLS TE enables a MPLS backbone to replicate and expand upon the Traffic Engineering capabilities of
Layer 2 technologies like ATM and Frame Relay. Here are some of the capabilities of MPLS TE:
MPLS TE capabilities are integrated into Layer 3, which optimizes the routing of IP traffic,
abiding by the constraints imposed on it.
The flow of routing traffic across a network is determined based on the resources the traffic flow
requires and the resources that are available.
It employs “Constraint-Based Routing”, i.e., it chooses the shortest path to the destination that
meets the resource requirements of the traffic flow. In MPLS TE, the traffic flow has bandwidth
requirements, media requirements.
MPLS TE recovers to link or node failures that change the topology of the backbone by adapting
to a new set of constraints.
2
ISP’s tend to use layer-2 techniques to manage their network bandwidth, which establishes the connection
as VC’s connect to IP backbone. IP for its share tends to have drawbacks as well. It was seen that a single
link failure resulted in the breaking down of VC’s, forcing a major re-convergence among the routing
protocols. The IP has no information about metrics associated with the physical topology, and, the routes
chosen may be sub-optimal. In short, the conventional solution is more costly, and there are complications
in the network management. MPLS is more scalable, manageable, and cost effective solution. In the future,
if an ISP needs to deploy new revenue-generating services, the MPLS forwarding infrastructure can remain
the same, while new services are built by simply changing the way packets are assigned to an LSP.
The goal of traffic engineering is to maximize the utilization of the network resources. MPLS TE
automatically establishes and maintains a tunnel across the backbone using RSVP. The tunnel path is
determined based on the tunnel resource requirements and network resources, such as bandwidth. These
paths are calculated based on the resources required and the resources available (Constraint-Based
Routing). In TE, there are multiple paths to the destination. When MPLS TE is configured, only one route
is seen in the routing table. All other routes to the destination are present in the routing protocols database.
A packet crossing the MPLS backbone travels on a single tunnel that connects the ingress point to the
egress point and is considered to have a metric one hop. MPLS TE is built on some of the following IOS
mechanisms:
LSP (label switch paths) tunnels, which are signaled through RSVP, with TE extensions. LSP
tunnels have configured destinations and are unidirectional.
An MPLS TE link management module that determines paths to use for LSP tunnels.
An MPLS TE link Management module that performs link admission and bookkeeping of the
resource information to be flooded.
Label switching forwarding, which provides routers with a Layer-2 like ability to direct traffic
across multiple hops as directed by resource-based routing algorithms.
4. ATM (Asynchronous Transfer Mode):
ATM is an International Telecommunication Union- Telecommunication Standardization Sector (ITU-T)
standard capable of multiplexing different services including voice, video and data, wherein information is
conveyed in small, fixed size cells. ATM networks are connection oriented. It is emerging as a technology
of choice for the public. It is a cell switching and multiplexing technology that combines the benefits of
circuit switching with those of packet switching. It provides scalable bandwidth from a few megabits per
second to many gigabits per second. Each cell consists of 53 bytes. The first 5 bytes contain cell header
information, and the remaining 48 contain the “payload”(user information). The basic ATM cell format is
shown in figure 3.
Generic Flow Control Virtual Path IE
Virtual Path IE Virtual Channel IE
Generic Flow Control 5 bytes
Virtual Channel IE Payload Type IE CLP
Header Error Check
Payload (48 bytes) 48 bytes
Figure 3: ATM Cell Format
3
The different parts of the cell are described below:
Generic Flow Control (GFC) identifies multiple stations that share a single interface.
Virtual Path Identifier (VPI) identifies the next destination of cell with help of VCI.
Virtual Channel Identifier(VCI) identifies the next destination of cell with help of VPI.
Payload Type (PT) indicates about congestion, frame type and data.
The operation of ATM is very simple. The two types of ATM connections are Virtual paths (VP), which
are identified by virtual path identifiers (VPI) and Virtual channels (VC), which are identified by virtual
path identifiers (VPI) and virtual channel identifiers (VCI). A virtual path is a bundle of virtual channels,
all of which are across the network on the basis of a common VPI. The information about the VPI/VCI
value is available in the local translation table. Whenever a cell is received on the incoming interface with a
known VPI/VCI value, the switch looks into its table and determines the outgoing port of the connection
and a new VPI/VCI value of the connection on that link. The switch then transmits the cell with the
appropriate VPI/VCI value. The VPI/VCI values are local to the switch and can be remapped at each
switch if required. There are three types of ATM services:
Permanent virtual circuit (PVC) in which manual configuration is required.
Switched virtual circuit (SVC) in which, SVCs are created and released dynamically and remain in
use only as long as the data is being transferred.
Connectionless service.
4.1. ATM Service categories:
The service categories of ATM allow the user flexible access to network resources and the ability to find
satisfactory compromise between performance and cost. According to the ATM Forum, there are five
service categories:
Constant Bit Rate (CBR)
Variable Bit Rate real time (VBRrt)
Variable Bit Rate non-real time (VBRnrt)
Unspecified Bit rate (UBR)
Available Bit Rate (ABR)
VBRrt is preferred over CBR since there are many constraints for CBR when real time traffic is considered.
Moreover, VBRrt doesn’t require constant bandwidth. Since both real time and non-real time traffic are
being used in this study, ABR and VBRrt were deployed for real time traffic and ABR and VBRnrt were
used for non-real time traffic.
4.1.1. Variable Bit Rate real time (VBRrt):
The types of applications supported by the VBRrt service category are ATM voice with bandwidth
compression and silence suppression and some types of multimedia applications. It is characterized by Peak
Cell Rate (PCR), which signifies the rate at which the virtual circuit can transmit in Kbps, Sustained Cell
Rate (SCR), which signifies the average rate at which the virtual circuit can transmit in Kbps and maximum
Burst Size (MBS), which measures the burstiness of a special connection.
4.1.2. Variable Bit Rate non-real time (VBRnrt):
This service category is intended for non-real time, bursty applications that require service guarantee.
Typical applications supported are critical response time transaction processing applications. Similar to
VBRrt, VBRnrt is also characterized by PCR, SCR and MBS.
4.1.3. Available Bit Rate (ABR):
This service category is intended for non-real time bursty applications. Typical applications supported are
those that adapt to network feedback. ABRs are characterized by PCR and MCR.
4
4.2. Traffic Management (TM) Functions:
There are several ATM TM functions that are used to monitor and control traffic and congestion. Some of
them include Connection-Admission-Control (CAC), Traffic Policing, Traffic Shaping, priority Control,
Buffer Management, Frame Discard and Route Management. The test scenarios in this study incorporate
CAC and Traffic Shaping.
In ATM, the network first establishes a Virtual Connection. CAC is a switch function in ATM that checks
to see that the connection is accepted only when sufficient resources are available at each successive link in
the circuit, thus avoiding congestion. Traffic shaping is a technique in which bandwidth utilization is
optimized. Traffic shaping was used in test scenarios to throttle the rate at which the packets are being sent
to the destination through a particular link.
5. Experimental Setup:
Since both real time and non-real time traffic was used to test the effects of link failure on the network, they
have divided this into two phases. Phase-1 talks about non-real time traffic, and Phase-2 focuses on real
time traffic. The routers used were all Cisco 3600 series routers. The switches used were all Light-Stream
LS1010. The links used to connect the routers and the ATMs are the OC-3 links (155.4 Mbps). The OC-3
links were chosen because of their reliability and huge bandwidth capacity, which usually is the case with
current ISP’s. OSPF was the routing protocol used in both the cases for network connectivity. On all
routers, the physical interfaces were divided into sub-interfaces, and PVC’s were configured on them. The
scenario for Phase-1 is shown in Figure 4.
5.1. Phase-1:
The main aim was to implement MPLS TE over ATM and observe the effects of link failure and link
loading. The network was setup such that there were two paths from source to destination. MPLS TE was
configured on all the routers, and tag switching was enabled on the ATM switches. MPLS TE was
configured to obtain a dynamic tunnel from source (R4) to destination (R7). Initially, the packets would
take the pre-established path through the tunnel.
Pagent
R1 on R2
5.1
ATM LSR R8
1.2
R3
R4 4.2
ATM 2
1.1
2.1
ATM 1
Pagent Path
5.2
2.2 ATM 3 4.1
Edge LSR
9.1 3.1 3.2
R5 R7
9.2 LSR: Label Switch Router
LC-ATM Interfaces
LC-ATM: Label Controlled ATM
Pagent Links used : OC-3 ( 155.4 Mbps)
on R6
Path 1
Path 2
Pagent Path
Figure 4: Phase 1: MPLS TE over ATM ( NRT)
5
Once this path was congested or over-utilized, MPLS TE with ATM Traffic Management would make the
Ingress router switch the link from the over-utilized path to the under-utilized path. This was possible since
the Ingress router’s routing protocol database (OSPF database in our case) had information about other
paths to the destination. The congestion for the OC-3 links was achieved by connecting extra routers that
were running the Pagent (Packet Generator) IOS image. Pagent is a traffic-generating tool that is used to
generate non-real time traffic (i.e., data traffic of large packet sizes). This Pagent uses two tools known as
TGN (used to generate traffic) and PKTS (used to capture packets). The Pagent traffic was running
continuously from R2 to R6 passing through the ATM network as shown in Figure 4 (indicated by the
dotted line). At the same time, ping traffic was sent from the source to the destination router. The ATM
Traffic Management scheme used in this scenario was CAC (Connection Admission Control). The tunnel
was created dynamically from R4 to R7. The service categories configured in this phase were ABR and
VBRnrt. VBRnrt was assigned to receive 95% of the traffic, and ABR was assigned to receive the
remaining amount of traffic. The pagent was used to congest the network. A continuous stream of data was
sent through the network that occupied 95% of the bandwidth when Pagent was used. Other ping traffic
(Extended ping traffic with don’t fragment bit set to “yes”) was sent from router R1 to R5. When data
traffic was sent from R4 to R7 when the link between ATM1 and ATM3 was congested because of the
Pagent traffic and the extended ping traffic flowing through it, the MPLS TE switched the tunnel from the
over-utilized path to the under-utilized path (shown by the dashed line in the diagram). Therefore network
connectivity was maintained. A sample configuration showing the commands used to configure the ATM
and MPLS supporting routers and the rules to be followed while configuring MPLS TE is shown in the
appendix.
5.2. Phase-2:
The same test scenario was repeated for real time traffic with some minor additions to the network and
changes in the configurations as shown in Figure 5.
The PSQM servers were used to measure the quality of voice when it was passed through the network.
Two Callgen supporting routers were connected via gateways to the source and the destination router.
These were in turn connected to a PSQM server. Callgen is a tool used to generate voice calls. The calls
were originated from “callgenR1” and were terminated at “callgenR2”. Traffic shaping was enabled on the
link between the gateways and router R4. Traffic shaping was configured on the Ethernet interface, which
throttled the rate at which the calls were being sent through the network. This was necessary since the links
used were OC-3 links (155.4 Mbps). By traffic shaping, the physical bandwidth of the link was altered to
the amount specified. The call received at the terminating end was compared with the call already stored in
the PSQM server to generate the PSQM score. PSQM uses an algorithm that gives scores by comparing the
voice call received through the network with that of the original call and measuring the voice quality
degradation. Since this phase involved real time traffic the ATM service categories configured on the
interfaces were ABR and VBRrt. ABR was configured to receive 95% of the traffic, and VBRrt handled the
remaining traffic. Extended ping traffic from R1 to R5 was used to congest the link. The ATM traffic
management technique used was Connection Admission Control. Appropriate peak cell rate (PCR),
sustained cell rate (SCR) and Maximum burst size (MBS) values were set for the required service category.
The configuration of the above parameters was very important, since they influence the quality of voice
calls sent through the network. The synchronization types were used here, which specifies the types of
tones to synchronize both ends i.e., the receiving and the transmitting end before running PSQM.
6
Pagent
R1
on R2
5.1
R8
1.2
R3
R4 4.2
ATM 2
1.1
T1 Link
2.1
ATM 1
R9
Pagent Path
GW 5.2
2.2 ATM 3 4.1 R7
CallgenR1
9.1 3.1 3.2
R5 GW
9.2 R10
Pagent
PSQM Server T1 Link
on R6
Callgen R2
LSR: Label Switch Router
LC-ATM: Label Controlled ATM
Links used : OC-3 ( 155.4 Mbps)
GW : Gateway
Path 1
Path 2
Pagent Path
Figure 5: Phase 2: MPLS TE over ATM ( RT)
In the tables 1,2 and 3 PSQM scores for different codecs and synchronization types such as 3-tone, DTMF
and no-sync has been shown respectively. The default is 3-tone. The DTMF is used here because the
current callgen design is implemented by dialing DTMF tones back and forth between the originating and
the terminating ends. The other type of sync i.e., no-sync is where in which there is no synchronization
between the receiving end and the transmitting end and hence the scores obtained were not better when
compared to others. The authors also conducted experiments for the above setup for different bandwidths
and measured its effect on the voice quality. The sample configurations done over the network has been
shown in the appendix.
6. Results:
6.1. Phase 1:
Phase-1 transmitted only non-real time traffic. Ping traffic was sent from source to destination. The
switching operation of the tunnel when the link was loaded is shown by trace route before and after the
tunnel was switched in Figures 6 and 7 respectively. The figures clearly indicate the change in the route and
the label information adopted by the packets before and after the switching. Separate ping traffic was also
sent which showed that at the time of switching some packets were dropped. Once the link was switched to
the next available path to the destination then all the data was sent through that path with no packet drops.
The network restoration time for this phase was observed to be in the range of 25 - 40 seconds. When
network was configured without MPLS TE then the network restoration time was in the range of minutes.
The time taken to switch from one path to another path was observed to be faster when MPLS TE and
ATM TM was configured over the network. This was because ATM TM used CAC, which is one of the
congestion avoidance algorithms.
7
rack7r5#ping 10.10.10.5
Type escape sequence to abort.
Sending 5, 100-byte ICMP Echos to 10.10.10.5, timeout is 2 seconds:
...!!
Success rate is 40 percent (2/5), round-trip min/avg/max = 28/28/28 ms
Ping traffic showing packets dropped when tunnel is being switched:
rack7r5#trace 10.10.10.5
Type escape sequence to abort.
Tracing the route to 10.10.10.5
1 100.1.2.2 [MPLS: Label 27 Exp 0] 92 msec 204 msec 160 msec
2 100.1.3.2 116 msec 112 msec *
Figure 6: Route taken before change of Tunnels (with Label info):
rack7r5#ping 10.10.10.5
Type escape sequence to abort.
Sending 5, 100-byte ICMP Echos to 10.10.10.5, timeout is 2 seconds:
!!!!!
Success rate is 100 percent (5/5), round-trip min/avg/max = 28/34/60 ms
Ping showing 100% success after the change of Tunnels:
rack7r5#trace 10.10.10.5
Type escape sequence to abort.
Tracing the route to 10.10.10.5
1 100.1.1.2 [MPLS: Label 29 Exp 0] 36 msec 36 msec 36 msec
2 100.1.4.1 20 msec 16 msec *
Figure 7: Route taken after the change in Tunnels (with label info):
6.2. Phase 2:
This phase, real-time traffic was transmitted. The PSQM scores were generated for different codecs and for
different synchronization types. The first trace route in Figure 8 shows the path taken by voice packets
from source to destination. The second trace route in Figure 9 shows the switched path taken by the voice
packets when the link is congested. It also shows the change in the MPLS labels chosen for both paths. The
network restoration time for this phase was observed to be same as that of phase-1. The quality of voice
calls before and after passing it through the network can be determined using the PSQM scores. PSQM
uses an algorithm that evaluates the quality of voice by comparing two speech samples and gives a score. A
PSQM score is in the range of 0-10. Any scores less than 2 indicates a good quality, PSQM scores in the
range of 2-6 are subjective, and any score above 6 is unacceptable. PSQM scores were recorded for all
kinds of codecs. The PSQM scores generated for different scenarios (A network with just OSPF without
MPLS, A network with MPLS but without TE, A network with MPLS TE but without ATM TM, A
network with MPLS TE with ATM TM) were compared and are shown in Tables 1,2 and 3. From these
tables we can observe that the PSQM scores are better for scenario with MPLS TE and ATM TM for all
cases. This is because the ATM TM uses the congestion avoidance algorithms such as CAC and traffic
shaping. The success of the calls was also achieved by setting appropriate traffic parameters for the
different service categories used. It was also observed that as the PCR value for the switch increased the
success rate of the call also increased.
rack7r5#trace 10.10.10.5
Type escape sequence to abort.
Tracing the route to 10.10.10.5
1 100.1.2.2 [MPLS: Label 16 Exp 0] 152 msec 76 msec 80 msec
2 100.1.3.2 64 msec 248 msec *
Figure 8: Route taken before starting the
call:
rack7r5#trace 10.10.10.5
Type escape sequence to abort.
Tracing the route to 10.10.10.5
1 100.1.1.2 [MPLS: Label 32 Exp 0] 36 msec 36 msec 36 msec
2 100.1.4.1 20 msec 20 msec *
Figure 9: Route taken by the call after change of the tunnel:
8
Codecs PSQM scores w/o PSQM scores with PSQM scores with PSQM scores with
(Kbps) MPLS TE (just MPLS but no TE MPLS TE and no ATM MPLS TE and
OSPF) TM ATM TM
G.711 (64) 1.419 1.273 2.435 0.972
G.726 (32) 1.942 1.273 2.420 1.329
G.729 (8) 1.460 1.411 3.925 1.203
G.723(6.3) 1.296 1.486 2.223 1.015
Table.1: Synchronization type: 3-tone
Codecs PSQM scores w/o PSQM scores with PSQM scores with PSQM scores with
MPLS TE (just MPLS but no TE MPLS TE and no ATM MPLS TE and
OSPF) TM ATM TM
G.711 0.669 0.670 3.240 0.627
G.726 0.683 1.104 2.486 0.648
G.729 0.709 1.242 3.541 0.674
G.723 0.742 0.921 2.847 0.691
Table 2: Synchronization type: (DTMF)
Codecs PSQM scores w/o PSQM scores with PSQM scores with PSQM scores with
MPLS TE (just MPLS but no TE MPLS TE and no ATM MPLS TE and
OSPF) TM ATM TM
G.711 2.012 2.145 4.129 1.746
G.726 2.421 2.104 5.031 2.124
G.729 2.742 2.546 5.112 1.842
G.723 2.396 2.286 4.912 1.910
Table 3: Synchronization type: no-sync
7. Observations:
From the above tables 1,2 and 3 it can be seen that G.711 yielded better results as expected. It was also
seen that performance to bandwidth ratio for G.729 was better when compared to other codec standards.
The PSQM scores for sync type DTMF were observed to be better than that of other synchronization types.
This is just because the current path-confirmation design of Callgen is implemented to suit DTMF.
8. Conclusion:
From the results shown, it can be concluded that by configuring MPLS TE with ATM TM the network
connectivity was restored faster when the link failed due to congestion. This was observed to be the same
for both non-real time and real time traffic.
For the real time traffic, it was observed that the voice quality altered when codecs and scenarios changed.
It can also be observed from the tables 1,2 and 3 that there is change in the scores for the scenarios with
MPLS TE and without MPLS TE. This can be attributed to equal cost load balancing property of the OSPF
routing protocol being used when MPLS TE is not configured. But with MPLS TE configured, the packets
take only one path to the destination and whenever the network is congested the voice calls get
deteriorated. MPLS TE with ATM TM solves this issue by using congestion avoidance algorithms and by
configuring traffic parameters for the service categories such as PCR, SCR and MBS. Congestion
avoidance algorithm used sense the link congestion early and switches the cells to the other path. The
authors suggest the use of MPLS TE with ATM TM for real time traffic.
9. Future work:
Considerable progress has been made in the fields of ATM and MPLS. ATM QOS and MPLS QOS can be
implemented on the network using Cisco 7200 series or higher routers, since they support QOS. Thus,
network performance can be further enhanced. The QOS features can be explored with the same 3600
series routers by replacing ATM LS1010 switches with MGX 8500 series switches.
9
The voice applications in the present scenarios have been tested for a limited number of calls. For further
research, the number of voice calls can be increased and the performance of the voice calls can be tested
and a standard can be determined. The experiment can be extended further and different applications like
video and voice calls can be sent over the same network and QOS can be implemented to obtain better
performance for the applications.
10. Acknowledgement:
I would like to thank Dr. Ravi Pendse for giving us an opportunity to work on this project and for his
excellent guidance and support throughout the project. I would also like to thank Mr. Ryan Doll, Mr. Ravi
Bhagvatulla, Ms. Tish Immich and Ms. Stecey Bailey for their coordination towards the successful
completion of our project.
11. References:
[1] Bruce David & Yakov Rekhter, “ MPLS Technology and Applications”
[2] Galina Diker Pildush, “Cisco ATM Solutions”, Cisco Press, August 2000.
[3] “http://www.mplsrc.com”, MPLS Forum
[4] “http://www.faqs.org/rfcs/rfc3031.html”, RFC 3031
[5] “http://www.cisco.com/warp/public/105/mpls_te_ospf.html”, Cisco Systems
documentation.
[6] “http://www.cisco.com”, Cisco Systems Documentation.
[7] Salman Bhaktiari, “ Performance Evaluation of a Dual-Leaky Bucket Algorithm”
Directed Studies, Wichita State University, 2001
12. Appendix
12.1. Rules to be followed while configuring MPLS TE:
It is mandatory to set up a loopback interface with an IP mask of 32 bits. This address will be used
for the setup of the MPLS network and TE by the routing protocol. This loopback address must be
reachable via the global routing table.
Set up a routing protocol for the MPLS network. It must be a link-state protocol (IS-IS or OSPF).
Enable ip cef, tag-switching ip ( MPLS ip), mpls traffic-engineering tunnel.
Enable RSVP on concerned interfaces.
Setup tunnels to be used for TE. There are many options that can be configured for MPLS TE
Tunnel, but the tunnel mode mpls traffic-eng command is mandatory. The tunnel mpls traffic-
eng autoroute announce command announces the presence of the tunnel by the routing protocol.
The use of ip unnumbered loopbackN for the IP address of the tunnel interfaces is a must.
12.2. Basic Configurations:
Some of the basic configurations are shown in next few sections, which are used to configure
MPLS TE over ATM.
12.2.1. Configuring the MPLS TE dynamic tunnels:
> configure terminal
> mpls traffic-eng tunnels
> interface tunnel1
> ip unnumber lo0
> tunnel destination 10.10.10.5
> tunnel mode mpls traffic-eng
> tunnel mpls traffic-eng auto announce
10
> tunnel mpls traffic-eng priority 1 1
> tunnel mpls traffic-eng bandwidth 512
> tunnel mpls traffic-eng path-option 1 dynamic
> int atm2/0.1 p
> band 512
> mpls traffic-eng tunnels
> ip rsvp bandwidth 512 512
> router ospf 100
> mpls traffic-eng area 0
> mpls traffic-eng router-id lo0
12.2.2. ATM switch LS1010: (sample configuration)
> configure terminal
> int atm0/0/1
> atm cac service-category abr permit
> atm output-queue force abr-ubr max-size 256
> atm cac max-peak-cell-rate abr recieve 50
> atm cac max-cdvt abr receive 5
> atm cac max-min-cell-rate abr recieve 50
> atm cac link-sharing max-bandwidth abr receive 1
> atm pacing 50 force
> interface ATM0/0/0
> atm output-queue force vbr-nrt max-size 65280
> atm cac max-peak-cell-rate vbr transmit 9999999
> atm cac max-peak-cell-rate abr transmit 10
> atm cac max-sustained-cell-rate transmit 9999999
> atm cac max-cdvt vbr transmit 9999999
> atm cac max-cdvt abr transmit 10
> atm cac max-min-cell-rate abr transmit 10
> atm cac link-sharing min-bandwidth vbr transmit 95
> atm cac link-sharing max-bandwidth abr transmit 1
> atm cac service-category vbr-nrt permit
> atm cac service-category abr permit
> interface ATM0/0/2
> atm cac service-category vbr-nrt permit
> atm cac max-peak-cell-rate vbr recieve 9999999
> atm cac max-cdvt vbr transmit 214748364
> atm cac link-sharing min-bandwidth vbr receive 95
12.2.3. Callgen / PSQM configurations: (sample configuration)
Global Configuration Mode:
> controller T1 4/0
> framing esf
> clock source internal
> linecode b8zs
> ds0-group 0 timeslots 1-24 type fxs-loop-start
Dial-Peer Configuration Mode Commands:
> dial-peer voice 89 voip
> destination-pattern 8888
> session target ipv4:100.1.20.1
> codec g711ulaw
11
> dial-peer voice 99 pots
> destination-pattern 9999
> port 4/0:0
> forward-digits all
12.2.4. Pagent Configuration: (sample configuration)
tgn pkts
clear conf clear all
fa0 free-buffer all
add ip timestamp
l3-src-addr 100.1.7.2 add 999999
l3-dest-addr 100.1.9.2 name proj
rate 900 select-buffer name proj
length 1000 fa0
fa0 add 900000
add tcp timestamp fa0 in
l3-src-addr 100.1.7.2 fa0 capture in on
l3-dest-addr 100.1.9.2 fa0 filter ip capture in ip-src 100.1.7.2
rate 900 fa0 filter tcp capture in ip-src 100.1.7.2
length 1000
all l2-src-addr 0000.1111.1111
all l2-dest-addr 0000.2222.2222
12