Enhancing Fault Tolerance and

Rerouting Strategies in MPLS Networks

A
Summary
For the award of

Doctor of Philosophy
In

Information Technology
Under the faculty of
Computer Science and Information Technology
Submitted to

Rajiv Gandhi Proudyogiki

Vishwavidyalaya
(State Technological University of Madhya Pradesh)
Bhopal, (M.P.), India

by
Ravindra Kumar Singh
(Enrollment No. 0108IT09PD61)

Supervisor Co-supervisor
Dr. Narendra S. Chaudhari Dr. Kanak Saxena
Professor (HAG) Professor
Computer Science and Engineering Department of Computer Applications
Indian Institute of Technology Indore, Samrat Ashok Technological Institute,
Indore, M.P., India Vidisha, M.P., India
Chapter 1

Introduction

1.1 Introduction

Multi-protocol Label Switching (MPLS) is a data forwarding mechanism that is

used in high performance backbone networks. It directs the data from one node
to another with the help of short labels instead of long Internet Protocol (IP)
address which requires complex look up in the routing tables [1]. According to
Network World [2], the various properties that contribute towards the wide success
of MPLS are as follows :

1. MPLS embraced IP
MPLS was the result of efforts of pinning ATM with IP in order to make
circuit oriented architecture for Internet. MPLS encapsulates the IP packet
which comes from the IP router. This is done by adding short, fixed and
appropriate labels after analyzing the IP header of the packet.

2. MPLS is flexible
MPLS is very flexible due to the separate control and data planes. It supports
more than one label in the label stack whose size can be varied depending
on the tunneling requirements.

3. MPLS is protocol neutral

MPLS supports many control plane and data plane protocols seamlessly.
It was designed not to change the existing scenarios but to enhance the
current Internet’s best effort service. This allows MPLS to work with ATM,
1
Chapter 1. Introduction 2

Frame Relay, SONET or Ethernet at the core. If underneath protocols are

enhanced, MPLS enhances too.

4. MPLS is pragmatic
MPLS is pragmatic in the sense that it only introduced two protocols, namely
Label Distribution Protocol (LDP) and Link Management Protocol (LMP).
All the remaining protocols are simply enhancement of the existing ones.
This has been feasible since one does not have to change the whole infras-
tructure to upgrade to MPLS.

5. MPLS is adaptable
MPLS is higly adaptable because it can support new applications and ser-
vices. Both layer 2 and 3 Virtual Private Networks (VPNs) can be easily
deployed with minimal costs.

6. MPLS support metrics

MPLS supports various metrics like delay, traffic volume and latency between
the routers which facilitates favorable returns to the vendors and also helps
them to deliver appropriate QoS to the end user.

7. MPLS scales
Successful technologies are accepted by the masses and therefore,they should
be scalable. Since it’s development, MPLS has been able to make it’s pres-
ence felt over many countries, across most of the public and private networks.

1.2 MPLS Traffic Engineering and Fault Toler-

ance: An Overview

Traffic Engineering (TE) is the process of routing traffic so that the desired traffic
metrics are as per the service level agreement with end user. MPLS facilitates TE
by providing QoS and fault tolerance for the traffic routed in it’s network. MPLS
deploys this by source routing which is done by provisioning virtual paths from
source router (Ingress) to destination router (Egress). Traffic is carried in this
signaled virtual path until there is a failure in the intermediate node or link.

In order to route the traffic of the failed path, two techniques have been proposed
by Internet Engineering Task Force (IETF). They are protection switching and
Chapter 1. Introduction 3

rerouting. In protection switching, one or more backup path is pre-provisioned for

each active path so that the traffic can be routed through the backup path in case
the active path fails. Whereas in rerouting, backup path is provisioned after the
node of link has failed locally to by pass the faulty node or link. Each of these
techniques has it’s own sets of advantages and disadvantages. One component
which is common in all the traffic engineering and fault tolerance methods is the
selection of efficient LSP’s.

1.3 LSP Selection

As stated in section 1.2, LSP selection is the core process in any TE method.
Before the IP packet is assigned any route, a path having appropriate QoS is
signaled from the whole network as per the destination address of the packet and
network policies.

1.4 Motivation

Most of the research in MPLS fault tolerance can be grouped under two classes
i.e. Protection Switching and Rerouting [3]. Similarly, the fault tolerance can be
classified as local or global depending on whether a part of the path or the entire
path is protected by the recovery technique.

For improving both these techniques,researchers have carried out various works
so as to minimize the recovery time by modifying their various phases. But in
all the proposed techniques, selection of an efficient path is a crucial parameter.
Therefore, in this thesis, we propose various solutions to the above mentioned
problem of efficient LSP selection.

1.5 Research Problem

Pertaining to the discussion in section 1.4, the following problems were identified:

1. How to select a LSP from the list of established LSP pairs so as to reduce
the blocking probability in accordance with the QoS of LSP requests ?
Chapter 1. Introduction 4

2. How to enhance the LSP pair algorithm in order to tune the bandwidth
available in the LSP pairs according to the traffic pattern ?

3. How to establish a LSP in the network which reduces blocking probability

and average bandwidth in feasible time complexity ?

4. How to enhance the robustness of a network while establishing LSPs in

accordance with the traffic conditions and path requests ?

5. How to reduce the time of the LSP selection algorithm with some compromise
in established optimal LSP ?

1.6 Background Literature Review

Authors in [4] proposes an approach for selecting an LSP pair among available
parallel LSP pairs with the aim of minimizing total blocking request while consid-
ering other desirable requirements such as probability of congestion, packet loss.
MPLS was essentially proposed for fast forwarding the packets over the Internet
[1]. However it has other capabilities which are used for the traffic engineering
and efficient resource utilization. It also facilitates source routing by using the
pre-signaled path known as LSP. Optimized routing of these LSPs is very impor-
tant which in turn is done by using the major building block, CBR [5]. Authors
in [5] addresses MPLS traffic engineering and its requirements in Large Internet
backbone. They suggest two conditions leading to the congestion in networks:

1. When the network resources are not sufficient to accommodate the traffic.

2. When the resources are sufficient but the traffic is unevenly distributed such
that some parts of the network become over occupied while other parts
remain under occupied.

For condition 1 we network hardware has to be upgraded, which is not always

feasible. So most of the work has been reported for solving condition 2. All the
parameters of QoS can be translated in to Bandwidth Guarantees [6]. Authors in
[6] also show that load balancing algorithms meant for distributing the load equally
in to the network have the adverse performance when the network is lightly loaded
by directing the traffic to longer links. If the traffic gets enough bandwidth then
Chapter 1. Introduction 5

the metrics like packet delay, packet disorder and blocking probabilities will be
having the optimum values. [7] gives the idea of maxflow, which is the value
of maximum bandwidth that a pair of nodes can transmit at the time. MinHop
algorithms [8] are routing algorithms which route the traffic on the basis of number
of Hops. This leads to the network congestion in high loads since maximum traffic
is routed by the shortest paths based on the number of links and the longer paths
remain idle. [9] Modified the MinHop algorithm by introducing the concept of
Shortest Widest Path (SWP) and Widest Shortest Path (WSP) which introduces
some bandwidth requirement in the MinHop. Load Minimization [10] is the form
of load balancing which routes the traffic on lightly loaded links.

Minimum Interface Routing Algorithm (MIRA) proposed in [11, 12] routes the
flow by calculating maxflow and finding the correlation between the max flow and
the maximum amount of bandwidth that can be routed between the pair of nodes.
MIRA identifies the critical links and assigns them the proportional weights. After
that the traffic is routed according to the assigned weights. Authors in [13] have
proposed an integrated solution for traffic optimization by combining the MinHop,
MIRA and Load balancing algorithms. They calculate the link cost by using all
three algorithms and assigning suitable weights depending on the load in the net-
work.As they have suggested, MIRA has all the properties to become a good load
balancing algorithm but has the only drawback of computational complexity. This
complexity lies in their solution as well since in the cost function link criticality is
calculated every time even when the network is lightly loaded. So this increases
the complexity and moreover it is more than MIRA since complexity of other two
algorithm adds as well. Authors in [14] update the previous proposed methods by
various authors on fault tolerance in MPLS networks. Recommendations of the
transmission of traffic of failed LSP by one of more failure free LSPs have been
made.

Paper [15] suggests an algorithm to control the admission of traffic from the edges
of the network using the threshold characteristics like bandwidth of the network
state. The paper essentially states that for efficient admission control, emphasis
should be given on consideration of the network state with the state of flow in
the network. These network states are computed by the shortest path algorithms
run beforehand in the background. Paper [13] performs comparative study of four
LSP selection methods i.e. Minimim Hop (MinHop) [8], Load balancing, Mini-
mumLength (MinLength) and MIRA [11]. MinHop algorithm selects the LSPs
Chapter 1. Introduction 6

considering the path length which is the number of intermediate hops. Load bal-
ancing tries to distribute the traffic demands into the entire network by balancing
the load as per the residue bandwidth of the link. Minimum length algorithm
engineers the traffic on the basis of physical length of the link. MIRA defines the
critical link as the link which can result in affecting the MaxFlow [7] between the
node pairs. MIRA delivers the best performance but has very high complexity
since it computes the MaxFlow frequently [13]. The paper proposes an integrated
solution by combining load balancing, MIRA and MinHop. Although it states that
MIRA is computationally complex yet, it has been utilized more number of times
than that in the original MIRA algorithm [11]. The paper [16] proposes a prob-
abilistic algorithm for improving survivability of the selected paths for the traffic
demands in the network. It proposes four cost functions and computes their perfor-
mance by sequentially implementing the cost functions in the algorithms. Results
vary with the sequence of the cost function deployed in the algorithms since there
are trade-offs between cost functions. If the paths are selected on the basis of
their failure probability history then load balancing gets affected i.e. the network
is secured at the cost of resource consumption. Authors in [17] propose a model
for link and node disjoint loop free path selection for 1:1 backup path protected
network.

In paper [18] authors have proposed an integrated solution by using the different
selection algorithms depending on the load in the network. Authors in paper [19]
present the model for problem of embedding the virtual network onto the physical
substrate network. This has been done by selecting the appropriate path keeping
in consideration the CPU capacity and bandwidth of the virtual network. The
problem is then relaxed by reducing the restriction of integer constraints.

In paper [20] this problem is further elaborated and solved by assuming that the
substrate network is not fault resistant. Authors propose the algorithm for sur-
vivable virtual network embedding on the substrate network. In the above works
authors except that of paper [16] did not consider the link failure probability as
the cost function. Authors in paper [16] differentiate the links into high availabil-
ity and low availability links based on the threshold of the link failure probability
and then establish the paths comprising of high availability links. For the low
availability links they propose backup paths. Due to the trade-off between the
cost functions, applying the cost function on the output of previous cost function
do not give the optimized output.
Chapter 1. Introduction 7

Routing algorithms motivated by Ant Colony Optimization (ACO) are also re-
ported in the literature. Authors in [21] conducted an experiment called double
bridge experiment. In this model a food source was connected with ant nest by two
paths of equal length. Ants, at the outset explore one of the two paths randomly.
In the later stages due to the random fluctuations, one of the two paths gets more
explored which in turn leads higher concentration of the pheromone deposited in
the path. Ants follow the paths with higher concentration and more over these
pheromones gets evaporated with due coarse of time. So this leads to the single
path followed by majority of ants despite both paths having same length.

A variant of double bridge experiment was conducted in [22]. In this experiment

two path from ant nest to the food source were of different length. With time,
the shorter path received more pheromones since ants traversing that were able
to return to the nest earlier than those who explored longer paths. So eventually
more ants selected the shorter paths and it became the favorable path from nest
to food . If in the case a more shorter path is established then also it will not be
explored by the ants because of the very little of no pheromone in that path. A
probability model was proposed by authors in [23] by observing the behavior of
ants .

1.7 Contributions of the Research Work

Based on the above mentioned research problems and objectives, the following
solutions have been proposed in this thesis.

1. The first problem is to select an LSP pair from a pool of LSP pairs in
order to prevent the dead lock condition, where LSP request gets rejected
due to unavailability of upstream and downstream bandwidth in the same
LSP pair. We have designed an algorithm which in the process of assigning
the LSP to request, reduces the gap between the bandwidth of upstream
and downstream LSP in a LSP pair. In this way, bandwidth in both the
directions is reduced in proportion to the request, which in turn prevents
deadlock and subsequently leads to lowering of request blocking.
Chapter 1. Introduction 8

2. For the second problem, the LSP pair selection algorithm is further enhanced
so as to select the LSP pair that has highest mismatch in the ratio of up-
stream to downstream bandwidth. This tunes the bandwidth for future
incoming requests.

3. Third problem is to establish a LSP with optimized traffic metrics. For this
problem, we use an integrated optimization model which uses the MIRA,
shortest path and load balancing algorithms depending on the network con-
dition.

4. Next problem is to design a LSP optimization algorithm which increases the

network robustness. We propose a model which establishes the LSP from the
link that has a very high availability history and proximity with the source
or the destination.

5. Last problem addressed in this thesis is to reduce the time complexity of

LSP selection algorithm at the cost of minor reduction in the optimal selected
LSP. We modify the existing ACO algorithm in the literature for optimal LSP
selection. After carrying out experiments on standard data sets, we obtain
that the time taken is reduced by >50% with insignificant degradation in
the optimal LSP selection.

1.8 Organization of Thesis

In this thesis, a review and a detailed comparison of the MPLS LSP selection for
constraint based routing has been presented. In the first phase (Chapter 1), MPLS
fault tolerance is introduced with three measures, namely prevention, detection
and recovery to cope up with the fault. The various proposals given by different
researchers for Constraint-based Routing (CBR) along with their limitations, that
are the source of motivation for our present work have been broadly illustrated.
Solutions are also presented which are described in the subsequent chapters.

In Chapter 2, MPLS is introduced with its attributes and methods for provid-
ing traffic engineering. Fault tolerance proposals by Internet Engineering task
force have been described along with their limitations. This chapter serves as a
background for the subsequent chapters.
Chapter 1. Introduction 9

In Chapter 3, various models for routing under QoS constraints proposed in the
literature have been discussed and analyzed.

In Chapter 4, work done by various authors for MPLS Fault tolerance is described.
Its main focus is on highlighting the contributions in load balancing and LSP
selection techniques for traffic engineering. Chapter 5 introduces various metrics
for comparison of LSP selection algorithms proposed in the thesis with those in
the literature.

In Chapter 6, a method for LSP pair selection is proposed which reduces the
request blocking and average bandwidth consumed by requests. The method al-
locates appropriate LSP, taking into consideration attributes like bandwidth re-
quested and threshold delay. The objective is to reduce cases where either the
upstream or the downstream LSP instead of both have spare bandwidth which
consequently leads to non-allocation of LSP to the LSP pair request. The pro-
posed method attempts to reduce this situation by selecting such a LSP pair that
minimizes the gap between the bandwidths of upstream and downstream LSPs.

In Chapter 7, the LSP pair selection method proposed in Chapter 6 has been
enhanced by allocating the LSP which has the maximum ratio of the upstream and
downstream bandwidth. In this way the algorithm tries to allocate the LSP pair
which has largest gap between the available upstream and downstream bandwidth
with a view to normalize upstream and downstream bandwidths across all LSPs
with every allocation request.

In Chapter 8, the problem for signalling optimal paths for generating the LSP pair
from the network has been discussed. This chapter analyses various proposals in
the literature and finds that there is no single algorithm which addresses all the
issues. An integrated solution is therefore proposed using a model which uses
appropriate algorithms depending on the traffic conditions. The solution is then
compared with few algorithms reported in literature and is found to perform better
in most scenarios.

In Chapter 9, a model has been proposed that introduces two parameters in cost
function, namely the failure history of the link and its distance from the source
which decides the probability of selecting a link in the LSP. Experiments have
been performed and the resultant network is found to be reliable and has reduced
average bandwidth and blocking probability. Network optimization models have
the limitation that they suffer from considerable time complexity, so in Chapter 10,
Chapter 1. Introduction 10

an improved Ant Colony Optimization model has been proposed for LSP selection.
Experiments have been performed on well-known network topologies and it was
found that our improved ACO finds close to optimal LSP in >50% less time
as compared to algorithms and algorithm selection methods presented in earlier
chapter.

To summarize, an in-depth analysis has been carried out and a formulation to

evaluate network protection level is presented. This evaluation is based on maxi-
mization of network reliability and minimization of failure impact in terms of QoS
degradation. A scalable proposal with reference to resource consumption that of-
fers the required level of protection across network scenarios for different traffic
services is comprehensively and experimentally analysed.
Chapter 2

Performance Analysis and Results

At the outset we addresses the problem of LSP pair selection among multiple
parallel LSP pairs between same ingress and egress LSR. An iterative algorithm
of marginal complexity was proposed involving simple arithmetic calculations at
each step. This is expected to minimize the overall bandwidth required taking
care of the congestion control and efficiency in the network. The proposal was
supported by simulation of the proposed algorithm and plotting the result as the
LSP request rejection probability against a certain number of LSP requests. Result
in Figure 2.1 and 2.2 show that the proposed algorithm reduces request rejection
by a considerable amount (more than 15%).

We further attempt enhancements to the proposed algorithm to increase further

performance vs. blocked requests and bandwidth utilization. While such enhance-
ments might increase complexity of LSP pair selection algorithm, we also attempt
quantification of additional complexity introduced.
160 60
Conventional Algorithm Conventional Algorithm
Proposed Algorithm Proposed Algorithm
140
Percentage of Blocked Requests

50
Number of Requests Blocked

120
40
100

80 30

60
20
40
10
20

0 0
0 50 100 150 200 250 300 0 50 100 150 200 250 300
Number of LSP Requests Number of LSP Requests

Figure 2.1: Number of Requests Blocked Figure 2.2: Requests Blocked in %

We enhance LSP pair selection algorithm proposed above. First the inability of
the proposed algorithm to perform satisfactory for the requests greater than 300 is
11
Chapter 2. Performance Analysis and Results 12

observed. The algorithm is then chained by including the conventional algorithm

when none of the LSP pairs get selected. This algorithm in further normalized by
sorting the LSP pairs according to the proportion of mismatch ratio of upwards
to downwards bandwidth. Then the allocation of LSP pair starts from the great-
est mismatch LSP pair to the lesser mismatch ratio. In this way the normalized
algorithm attempts to reduce the mismatch proportion in the LSP thereby pre-
venting the deadlock condition when the upwards and downwards bandwidth is
available but not in the same LSP pair. Plotted results of Figure 2.3 and Figure
2.4 indicates the reduced blocking rate with negligible increase in time complexity
of the algorithm. Next chapter we give a formalism for establishment of efficient
and cost effective LSPs in the network.
700 2.5e+06
Conventional LSP Selection Conventional LSP Selection
Enhanced LSP Selection Enhanced LSP Selection
Cumulative time in NanoSeconds
600
Number of Requests Blocked

Normalized LSP Selection Normalized LSP Selection

2e+06
500

1.5e+06
400

300
1e+06

200
500,000
100

0 0
0 500 1,000 1,500 2,000 2,500 3,000 0 500 1,000 1,500 2,000 2,500 3,000
Number of LSP Requests Number of LSP Requests

Figure 2.3: Normalized LSP Selection Figure 2.4: Cumulative time taken LSP
Algorithm Selection Algorithms

We propose an integrated solution for allocation of LSPs after the link or node
fault in the network. Simulation suggest that none of the algorithm is suitable
for all loads in the network so an integrated solution was proposed which uses the
characteristics of each algorithm for the load it is suitable for. This integrated
solution can be further taken for extensive theoretical research and simulation for
the optimized results. Results are plotted for satisfactory performance in request
blocking probability (Figure 2.5) and mean load (Figure 2.6) of the network.
0.3 0.3
Proposed Algorithm MIRA Load Balancing Proposed Algorithm
MIRA
0.25 0.25 Load Balancing
Blocking Probability

Average Load in MB

0.2 0.2

0.15 0.15

0.1 0.1

0.05 0.05

0 0
0 1,000 2,000 3,000 4,000 5,000 0 1,000 2,000 3,000 4,000 5,000
Number of LSPs Number of LSPs

Figure 2.5: Blocking Probability Figure 2.6: Mean Load

Chapter 2. Performance Analysis and Results 13

Next for enhancing the reliability, we present a model for path allocation for dy-
namic LSP request in a MPLS network. A novel cost function with three metrics
is proposed. Proposed cost function has been simulated and was found to increase
the survivability of network considerably when compared with five other algo-
rithms mentioned in the literature. Figure 2.7 and 2.8 depict performance of our
model with other algorithms in the literature. Network Protection Degree (NPD)
in Figure 2.8 is the number of LSPs having one or mode link having the failure
probability less then 1% divided by total number of LSPs. Failure Impact Degree
(FID) in Figure 2.7 is total number of links having distance more than 1 from the
ingress or egress LSRs.
0.85
MinHop Load Balancing Residue Bandwidth MinHop Load Balancing Residue Bandwidth
1.2 Link Cost MIRA Proposed Algorithm Link Cost MIRA Proposed Algorithm

Network Protection Degree

0.8
Failure Impact Degree

0.8
0.75

0.6

0.7
0.4

0.2 0.65
0 500 1,000 1,500 2,000 0 500 1,000 1,500 2,000
Number of LSPs Number of LSPs

Figure 2.7: Failure Impact Degree Figure 2.8: Network Protection Degree

0.84 400
Probabilistic LSP Selection
Probabilistic LSP Selection Ant Colony Optimization
Ant Colony Optimization 350
0.82
Network Protection Degree

300
0.8
Cumulative Time Taken

250
0.78
200
0.76
150

0.74
100

0.72
50

0.7 0
0 500 1,000 1,500 2,000 0 500 1,000 1,500 2,000
Number of LSPs Number of LSPs

Figure 2.9: Network Protection Degree Figure 2.10: Cumulative Time Taken

A modified ACO was used to optimize the LSP establishment model described in
previous paragraph. In this modified ACO, cost functions like link cost and link
failure probability were also included for determining the probability of selection
of link in an LSP. Result in Figure 2.9 indicate that the paths found were near
to optimal and the network scored satisfactory in the reliability and protection
cost metrics. This was due to the inclusion of other cost functions in the link
selection. Moreover the biggest advantage of using ACO was reduction of time
complexity (Figure 2.10)of the algorithm with a considerable amount. So ACO,
Chapter 2. Performance Analysis and Results 14

should be considered as the viable approach in the domain of network routing

optimization.
Chapter 3

Conclusion

This chapter lists the contributions of this thesis. In this thesis, we performed an
exhaustive study of the fault tolerance and recovery mechanisms in the context
of MPLS networks. A key aspect to achieve fault tolerance and recovery is the
selection and establishment of LSP in the network. We compared the different
algorithms proposed in the literature to achieve this and their trade-offs in terms
of time taken, request-blocking probability, complexity, resource consumption and
other constraints. We proposed a novel algorithm for LSP pair selection and
establishment in this work.

In this work we designed an algorithm which in the process of assigning the LSP
to request, reduces the gap between the bandwidth of upstream and downstream
LSP in a LSP pair. In this way, bandwidth in both the directions is reduced
in proportion to the request, which in turn prevents deadlock and subsequently
leads to lowering of request blocking. The LSP pair selection algorithm is further
enhanced so as to select the LSP pair that has highest mismatch in the ratio of up-
stream to downstream bandwidth. This tunes the bandwidth for future incoming
requests. Further to enhance the LSP selection based on network conditions, we
developed an optimized model which adds to the network robustness by establish-
ing the LSP from the link that has a very high availability history and proximity
with the source or the destination. We used Ant Colony Optimization technique
to further reduce the time complexity of our algorithm.

To summarize, we evaluated the existing algorithms for LSP pair selection and
establishment on the standard data sets and identified a metric. We then used this
metric to compare the performance of existing algorithms against the algorithms
15
Chapter 3. Conlusion 16

we proposed in this work. We showed by means of simulations that our algorithm

performs almost twice as better than the existing algorithms in terms of time
taken for path establishment. Our algorithm also supports low request blocking
probability, high bandwidth utilization and less congestion in the network.
Bibliography

[1] E. Rosen, A. Viswanathan, and R. Callon, “Multiprotocol Label Switching

Architecture,” IETF, RFC 3031, January 2001. [Online]. Available:
http://tools.ietf.org/html/rfc3031

[2] C. D. Marsan, “7 reasons MPLS has been wildly successful,” Network

World, March 2009. [Online]. Available: http://www.networkworld.com/
news/2009/032709-mpls-turns-12.html

[3] V. Sharma, F. Hellstrand, B. Mack-Crane, S. Makam, K. Owens,

C. Huang, J. Weil, B. Cain, L. Anderson, B. Jamoussi, A. Chiu, and
S. Civanlar, “Framework for Multi-protocol Label Wwitching (MPLS)-
based Recovery,” IETF, RFC 3469, February 2003. [Online]. Available:
http://tools.ietf.org/html/rfc3469

[4] S. i. Kuribayashi and S. Tsumura, “Optimal LSP selection method in MPLS

networks,” in In IEEE Pacific Rim Conference on Communications, Com-
puters and Signal Processing, 22-24 August 2007, pp. 371–374.

[5] D. Awduche, J. Malcolm, J. Agogbua, M. O’Dell, and J. McManus,

“Requirements for Traffic Engineering Over MPLS,” IETF, RFC 2702,
September 1999. [Online]. Available: http://tools.ietf.org/html/rfc2702

[6] Q. Ma and P. Steenkiste, “On Path Selection for Traffic with Bandwidth
Guarantees,” in Proceedings of International Conference on Network Proto-
cols, 1997, pp. 191–202.

[7] R. K. Ahuja, T. L. Magnanti, and J. B. Orlin, Network Flows: Theory, Algo-

rithms, and Applications. Englewood Cliffs, NJ: Prentice-Hall, 1993.

[8] J. Moy, OSPF: Anatomy of an Internet Routing Protocol. New York:

Addison-Wesley, 1998.

17
Bibliography 18

[9] R. A. Guerin, A. Orda, and D. Williams, “QoS Routing Mechanisms

and OSPF Extensions,” in IEEE, Global Telecommunications Conference,
GLOBECOM ’97, vol. 3, 1997, pp. 1903–1908.

[10] K. Long, Z. Zhang, and S. Cheng, “Load Balancing Algorithms in MPLS

Traffic Engineering,” in Proceeding of IEEE International Conference on High
Performance Switching and Routing, Dallas, 29-31 May 2001, pp. 175–179.

[11] K. Kar, M. Kodialam, and T. V. Lakshman, “Minimum Interference Routing

of Bandwidth Guaranteed Tunnels with MPLS Traffic Engineering Applica-
tions,” Selected Areas in Communications, IEEE Journal on, vol. 18, no. 12,
pp. 2566–2579, Dec 2000.

[12] M. Kodialam and T. V. Lakshman, “Dynamic Routing of Locally Restorable

Bandwidth Guaranteed Tunnels using Aggregated Link Usage Information,”
in Proceedings of IEEE Twentieth Annual Joint Conference of the IEEE Com-
puter and Communications Societies, INFOCOM 2001, vol. 1, 2001, pp. 376–
385.

[13] G. T. S. Lahoud and L. Toutain, “Classification and Evaluation of Constraint-

Based Routing Algorithms for MPLS Traffic Engineering,” in French Sixth
Meetings on Algorithmic Aspects of Telecommunications, AlgoTel 2004, Batz-
sur-Mer, France, 2004.

[14] J. W. Lin and H. Y. Liu, “Redirection Based Recovery for MPLS Network
Systems,” Journal of Systems and Software, vol. 83, no. 4, pp. 609–620, 2010.

[15] A. Bosco, R. Mameli, E. Manconi, and F. Ubaldi, “Edge Distributed Ad-

mission Aontrol in MPLS Networks,” IEEE Communications Letters, vol. 7,
no. 2, pp. 88–90, February 2003.

[16] M. Amin, K. H. Ho, G. Pavlou, and M. Howarth, “Improving Survivability

through Traffic Engineering in MPLS Networks,” in 10th IEEE Symposium
on Computers and Communications (ISCC’05), Cartagena, Murcia, Spain,
27-30 June 2005, pp. 758–763.

[17] M. N. -Pour and V. Desai, “Loop-free Traffic Engineering with Path Protec-
tion in MPLS VPNs,” Computer Networks, vol. 52, no. 12, pp. 2360–2372,
August 2008.
Bibliography 19

[18] R. K. Singh and N. S. Chaudhari, “Integrated Load Balancing Approach for

Fault Tolerance in MPLS Networks,” in Communication Systems and Network
Technologies (CSNT), 2013 International Conference on, April 2013, pp. 295–
298.

[19] M. R. R. M. Chowdhury and R. Baoutaba, “ViNE-Yard: Virtual Net-

work Embedding Algorithms With Coordinated Node and Link Mapping,”
IEEE/ACM Transactions on Networking, vol. 20, no. 1, pp. 206–219, 2012.

[20] M. R. Rahman and R. Baoutaba, “SVNE: Survivable Virtual Network Em-

bedding Algorithms for Network Virtualization,” IEEE/ACM Transactionson
Network and Service Management, vol. 10, no. 2, pp. 105–118, 2013.

[21] J. Pasteels, J. L. Deneubourg, and S. Goss, “Self-organization mechanisms in

ant societies (I): Trail recruitment to newly discovered food sources,” Expe-
rientia Supplementum, vol. 54, pp. 155–175, 1987.

[22] S. Goss, S. Aron, J. L. Deneubourg, and J. M. Pasteels, “Self-organized Short-

cuts in the Argentine Ant,” Naturwissenschaften, vol. 76, no. 12, pp. 579–581,
1989.

[23] J. L. Deneubourg, S. Aron, S. Goss, and J. M. Pasteels, “The Self-organizing

Exploratory Pattern of the Argentine Ant,” Journal of Insect Behavior, vol. 3,
no. 2, pp. 159–168, 1990.
List of Research Papers Pub-
lished/Presented/Communicated

[1] R. K. Singh, N. S. Chaudhari, and K. Saxena, “Load Balancing in IP/MPLS

Networks: A Survey,” Communications and Network, vol. 4, no. 2, pp. 151–
156, 2012.

[2] R. K. Singh, K. Saxena, and R. Singh, “Traffic Aware LSP Selection Method in
MPLS Networks,” in Information Processing and Management, Communica-
tions in Computer and Information Science, Springer Berlin Heidelberg, 2010,
vol. 70, pp. 455–457.

[3] R. K. Singh, N. S. Chaudhari, and K. Saxena, “Enhanced Traffic Aware LSP

Selection Method in MPLS Networks,” in 2012 Ninth International Conference
on Wireless and Optical Communications Networks (WOCN), Sept 2012, pp.
1–4.

[4] R. K. Singh and N. S. Chaudhari, “Integrated Load Balancing Approach for

Fault Tolerance in MPLS Networks,” in 2013 International Conference on
Communication Systems and Network Technologies (CSNT), April 2013, pp.
295–298.

[5] R. K. Singh, N. S. Chaudhari, and K. Saxena, “Probabilistic Selection of QoS

Paths for Improving Survivability in MPLS Networks,” Communications and
Network, vol. 5, no. 4, pp. 328–336, 2013.

Ravindra Kumar Singh

20