You are on page 1of 8

Survivable WDM Mesh Networks, Part I – Protection

S. Ramamurthyt and Biswanath Mukherjee


Department of Computer Science
University of California, Davis, CA 95616, U.S.A.
tCorresponding Author –
Tel: (530) 752-5129 Fax: (530) 752-4767
E-mail: ramu@cs. ucdavis. edu

Abstract may be routed from an input fiber to an output fiber with-


This investigation considers optical networks which em- out undergoing opt oelectronic conversion. An optical signal
ploy wavelength cross-connects that enable the establish- passing through a wavelength-interchanging cross-connect
ment of wavelength- division-multiplexed ( WDM) channels, may be converted from the input wavelength to a different
between node-pairs. In such and other networks, the fail- wavelength at the output port.
ure of a network element (e. g., fiber link, cross-connect, The architecture of a wavelength-routed optical net-
etc.) may cause the failure of several optical channels, work, shown in Fig. 1, consists of wavelengt h cross-connects
thereby leading to large data losses. This study examines (wavelength-selective or wavelength-interchanging) (labeled
different approaches to protect mesh-based WDM optical 1 through 15) interconnected by fiber links. Each network
networks from single-link failures. These approaches are link consists of a pair of unidirectional fiber links. Techno-
based on two basic survivabildy paradigms: (a) path pr-o- logical constraints dictate that the number of WDM chan-
tection/restora tion, and (b) link pr-otection/res toration. In nels that can be supported in a fiber be limited to W (whose
path- and link- protection schemes, backup paths and wave- value is a few tens today, but is expected to improve with
lengths are reserved in advance at the time of call setup. time and technological breakthroughs). An access station is
Path- and link-restoration schemes are dynamic schemes in connect ed to each cross-connect. For clarity of exposition,
which backup paths ar-e discovered (from the spare capac- we will consider the access-station/cross-connect combina-
ity in the network) upon the occurr-ante of a failure. In tion as an integrated unit which we will refer to as a network
part 1 of this study presented in this paper, we formulated node. In this work, we assume that all cross-connects are
Integer Linear Programs to determine the capacity require- wavelength-selective, and there is no wavelength conversion
ments for the above protection schemes for a static trafic in the network.
demand. In part 2 of this study presented m [1], we anal- A connection request is satisfied by establishing a light-
yse the protection-switching times for different protection path from the source node of the connection to the destina-
schemes, propose distributed restoration protocols, and anal- tion node. A lightpath is an all-optical channel which may
yse the restoration-time , and restoration eficiency charac- span multiple fiber links, to provide a circuit-switched inter-
teristics of the different restoration protocols. connection between two nodes. In the absence of wavelength
Keywords and Phrases: Optical network, WDM, converters, a lightpath would occupy the same wavelength
wavelength routing, lightpath, failure, survivability, protec- on all fiber links that it traverses. Two lightpaths on a fiber
tion, restoration, capacity requir-ement, optimization, I.LP. link must be on different wavelength channels to prevent the
interference of the optical signals. Figure 1 shows the follow-
1 Introduction ing wavelength-continuous lightpaths: (a) between Nodes 10
Wavelength-division multiplexing (WDM) [2] divides the and 6 on wavelength Al, and (b) between Nodes 15 and 9 on
tremendous bandwidth of a fiber (potentially a few tens wavelength JZ. In this work, we assume that all connection
of terabits per second) into many non-overlapping wave- requests are unidirectional.
lengths (WDM channels). Each channel can be operated A logical topology consists of a set of Lightpaths set up to
asynchronously and in parallel at any desirable speed, e.g., exploit the relative strengths of both optics and electronics -
peak electronic speed of a few Gbps. An access station may viz., packets of information are carried by the logical topol-
transmit signals on different wavelengths, which are coupled ogy “as far as possible” in the optical domain using optical
into the fiber using wavelength multiplexer. An optical switching, but packet forwarding from lightpath to lightpath
signal passing through a wavelength-selective cross-connect is performed via electronic packet switching, whenever re-
1S. Ramamurthy and B. Mukherjee were supported by NSF quired. The electronic packet-switching may be performed
Grant No. NCR-9508239, and by grants from Pacific Bell and by an ATM switch (or an 1P router), in which case the logical
the UC MICRO Program. topology is operated as an ATM network (or 1P network).

0-7803-5420-6/99/$10.00 (c) 1999 IEEE


dynamic restoration, the spare capacity available within the
network is utilized for restoring services affected by a failure.
Generally, dynamic-restoration schemes are more efficient
in utilizing capacity due to the multiplexing of the spare-
capacity requirements, and provide resilience against dif-
ferent kinds of failures, while dedicated-restoration schemes
have a faster restoration time and provide guarantees on the
restoration ability.

Protection/Restoration Schemes

+7
preconfigured
Backup Route and Wavelength
Dynamic Dk.covery of
Backup Route and Wavelength

Figure 1: Architecture of a wavelength-routed optical


net work.

Dedicated Shared Link Path


Backuv Backuv Restorauon Restoration
In such an optical network architecture, the failure of a
network component such as a fiber link can lead to the fail-
ure of all the lightpaths that traverse the failed linkl. Since
each lightpath is expected to operate at a rate of several
AA
Link Path Link Path
Gbps, such a failure can lead to a severe disruption in the Protection Protecuon Protection Protection

network’s traffic. SONET operating over the WDM opti-


cal network provides its own protection schemes (and has Figure 2: Different schemes for surviving single-link fail-
protection times on the order of milliseconds), and higher ures.
protocol layers (such as ATM and 1P) operating over the
optical network may have their own recovery procedures to This study examines different approaches (illustrated in
recover from link failures. An 1P network, for example, re- Fig. 2) to survive single-link failures in an optical net-
covers from link failures by rerouting data packets around work. These approaches are based on two basic survivabil-
the failed link. However, the recovery time for higher lay- ity paradigms: (a) path protection/restoration and (b) link
ers (such as ATM and 1P) is still significantly Iwge (on the protection/restoration.
order of seconds), whereas we expect that restoration times
at the optical layer will be on the order of milliseconds to ● Path protection frestoration:
minimize data losses [3]. Furthermore, it is beneficial to In path protection, the source ancl destination nodes
consider restoration mechanisms in the optical layer for the of each connection statically reserve backup paths on
following reasons [4]: (a) the optical layer can efficiently mul- an end-to-end basis during call setup. In path restora-
tiplex protection resources (such as spare wavelengths and tion, the source and destination nodes of each connec-
fibers) among several higher-layer network applications, and tion that traverses a failed link dynamically discover a
(b) survivability at the optical layer provides protection to backup route on an end-to-end basis (such a backup
higher-layer protocols that may not have built-in protection. path could be on a different wavelength channel) after
There are several approaches to ensure fiber network sur- the link failure.
vivability [5, 6]. Survivable network architectures are based
either on dedicated resources, or on dynamic restoration. In – Dedicated-path protection: In dedicated-
dedicated-resource protection (that includes Automatic Pro- path protection (also called 71+1 protection), at
tection Switching (APS) and Self-Healing Rings), network the time of call setup for each primary path, a
service is restored utilizing the dedicated network resources. link-disjointz backup path and wavelength are re-
The network resources used in dedicated-resource protection served, and dedicated to that call. The backup
may be dedicated for each failure scenario; alternatively, the wavelength reserved on the links of the backup
dedicated network resources used for protection against fail- path are dedicated to that call, and are not shared
ures may be shared among different failure scenarios. In with other backup paths.

1we ~sume in this study that a link is unidirectional, and a 2B y link-disjoint we mean that the backup path for a connec-
link failure is the failure of a unidirectional link. Our work can be tion has no fiber-links in common with the primary path for that
generalized to account for the failure of a bidirectional fiber-link connection.

0-7803-5420-6/99/$10.00 (c) 1999 IEEE


cover a route around the link, for each wavelength that
Primary
___________ traverses the link.

– Dedicated-link protection: In dedicated-link


protection, at the time of call setup, for each link
of the primary path, a backup path and wave-
length are reserved around that link, and are
dedicated to that call. In general, it may not
v Backup - be possible to allocate a dedicated backup path
around each link of the primary call, and on the
(a) Path protection same wavelength as the primary path. For ex-
ample, Fig. 4 shows a bidirectional ring network
with one connection request between Node 1 and
Node 5. The backup path around link (2,3), viz.

SOq+[>
(2,1,8,7,6,5,4,3), and the backup path around link
(3,4), viz. (3,2,1,8,7,6,5,4), share links in common
and hence cannot be dedicated the same wave-
length. Hence, we will not consider dedicated-link
protection in this work3.
:.4.. .. ............... .:5
Backup

(b) Link protection

Figure 3: Protection schemes.

– Shared-path protection: In shared-path pro-


tection, at the time of call setup for a primary
path, a link-disjoint backup path and wavelength
are also reserved. However, the backup wave-
length reserved on the links of the backup path
may be shared with other backup paths. As a re-
sult, backup channels are multiplexed among dif-
ferent failure scenarios (which are not expected to Figure 4: Illustrative example showing that dedicated-
occur simultaneously), and therefore shared-path link protection is not possible in a bidirectional ring
protection is more capacity efficient when com- network.
pared with dedicated-path protection.
– Path restoration: In path restoration, the – Shared-link protection: In shared-link protec-
source and destination nodes of each connec- tion, at the time of call setup, for each link of the
tion traversing the failed link participate in a primary path, a backup path and wavelength are
distributed algorithm to dynamically discover a reserved around that link. However, the backup
backup route and wavelength on an end-to-end wavelengths reserved on the IIinks of the backup
basis (such a backup path could be on a different path may be shared with other backup paths. As
wavelength channel). If no new route (and as- a result, backup channels are multiplexed among
sociated wavelength) are discovered for a broken different failure scenarios (which are not expected
connection, that connection is blocked. to occur simultaneously), and therefore shared-
link protection is more capacity efficient ~hen
● Link protection/restoration: compared with dedicated-link protection.
In link protection/restoration (illustrated in Fig. 3(b)), — Link restoration: In link restoration, the end-
all the connections that traverse the failed link are
nodes of the failed link participate in a distributed
rerouted around that link. The source and destination
algorithm to dynamically discover a route around
nodes of the connections traversing the failed link are
the link, for each wavelength that traverses the
oblivious to the link failure. In link protection, during
call setup, backup paths and wavelength are reserved 31f wavelength
converters are present in the network, then
around eack link of the primary path. In link restora- dedicated-link protection is possible by having the backup paths
tion, the end-nodes of the failed link dynamically dis- around (2,3) and (3,4) on different wavelengths.

0-7803-5420-6/99/$10.00 (c) 1999 IEEE


link. If no new route (and associated wavelength) and given. Our objective is to minimize the total number of
are discovered for a broken connection, that con- wavelengths used on all the links in the network (for both
nection is blocked. the primary paths and backup paths). The ILP solution also
determines the routing and wavelength assignment of the
In this paper, we examine the wavelength capacity re- primary and backup paths. Generally, capacity efficiency
quirements, and routing and wavelength assignment of pri- can be measured in two ways: (a) given a certain capacity,
mary and backup paths for different protection schemes for maximize the protected carried demand [7], or (b) given a
a static traffic demand. In a follow-up paper [1], we exam- certain demand, and given a 100% restoration requirement,
ine the protection-switching time requirements for path- and minimize the total capacity used. In our formulations, we
link-protection schemes, and propose distributed protocols require that all demands should be protected, and we mini-
for restoration, and study the restoration time requirements, mize the total capacity used. ILPs 1, 2, and 3 minimize the
and restoration efficiency for path and link restoration. capacity utilizations for dedicated-path protection, shared-
The design of a survivable optical network has been stud- path protection, and shared-link protection, respectively.
ied in [7, 8, 9, 4, 3, 10, 11, 12, 20, 21, 22]. In [7], the authors 2.1 Not at ion
propose physical protection schemes and a path-restoration We define the notation employed to formulate the ILPs.
scheme based on 1+1 protection. The work in [8] consid- We are given the following: (a) the network topology rep-
ers different approaches for fault-tolerant design of optical resented as a directed graph G, (b) a demand matrix, i.e.,
ring networks. The work in [4, 3] addresses the issues in the number of lightpath requests between node-pairs, and
designing a survivable optical layer. In [10], the authors (c) alternate routing tables at each node. Also given are the
propose an algorithm that protects optical mesh networks following.
from link and node failures. In [9], the authors examine
provisioning algorithms to provision wavelength capacity in ● N: Nodes in the network (numbered 1 through N).
a WDM optical network. In [12], the authors propose an- ● Node-pairs are numbered 1 through N x (N – 1).
alytical methods to estimate capacity utilization in optical ● E: Links in the network (numbere[i 1 through E).
networks that are resilient against single-link failures. In
● W: Maximum number of wavelengths on a link.
[20], the authors examine 1+1 protection, and other WDM
network architectures with optical protection. ● Ri: Set of alternate routes for node-pair i.
Path- and link-restoration schemes have been extensively ● M’ = [RI: Number of alternate routes between node-
researched in circuit-switched transport networks [5, 6, 13, pair i. Let ill be the maximum number of alternate
14, 15]. In [13], the authors report that path restoration routes between any node-pair, i.e., M = Maxi M%.
provides about 19% improvement in spare-capacity utiliza- ● R;: Set of eligible alternate routes between node-pair
tion over link restoration in circuit-switched transport net- i after link j fails.
works. Distributed protocols for restoration have been ex-
● end-nodes(j): the set of alternate routes between the
tensively researched in circuit-switched transport networks
node-pair adjacent to link j.
[16, 17, 19, 18] and in ATM networks [15, 6].
1.1 Out line of Remaining Sections ● d,: Demand for node-pair i, in terms of number of
In Section 2, we develop Integer Linear Program (ILP) lightpath requests.
formulations for the routing and wavelength assignment We require the ILPs to solve for the following variables:
problem, and wavelength utilization for a static traffic de-
● Sj: Number of spare wavelengths used on link j.
mand, for each of the different protection schemes. Sec-
tion 3 presents numerical results for wavelength capacity ● wj: Number of wavelengths used by primary lightpaths
utilization on a representative network topology for differ- on link j.
ent protection schemes. Section 4 concludes this work with ● ~~r takes on the value of 1 if the rth route between
a discussion of its main contributions. node-pair i utilizes wavelength w before any link fail-
ure; O otherwise. These variables are employed in all
2 Problem Formulation ILPs.
In this section, we develop ILP formulations of path- and
● o~ takes on the value of 1 if the backup router between
link-protection schemes to protect against all possible single-
node-pair i utilizes wavelength w; O otherwise. These
link failures. We assume that the network topology and a
variables are employed only in ILP1.
demand matrix (consisting of the number of connections to
be established between each node-pair) are given. We also ● 6~~~ takes on the value of 1 if wavelength w is utilized
assume that the set of alternate routes4 (that are used to on restoration route r between node-pair i, when link
satisfy any demand) between each node-pair is precomputed k fails; O otherwise. These variables are employed only
in ILP2.
41n alternate routin~, each network node has a routing table
9 l~r takes on the value of 1 if wavelength w is utilized
which contains a list of a limited number of fixed routes to each
destination node.A connection request arriving at the node uti-
on restoration route r between the node-pair that is
lizes one of the routes to the destination node from the set of adjacent to j, when link j breaks; O otherwise. These
available routes. variables are employed only in ILP3.

0-7803-5420-6/99/$10.00 (c) 1999 IEEE


● m~ takes on the value of 1 is wavelength w is utilized Number of channels on each link is bourlded:
by some restoration route r that traverses link j, O
otherwise. These variables are emdoved in ILP2 and Wj+sjsw l~jz;E (9)
ILP3.
Before Fault:
2.2 ILP Formulations Demand between each node-pair is satisfied:
2.2.. 1 ILP1: Dedicated-Path Protection

Minimize the total capacity used: .=1 7JJ=1

E
Definition of the number of primary lightpaths traversing a
Minimize ~(wj + s,) (1) link:
j=l

Number of lightpaths on each link is bounded:

(W, +sj)<w l<j<E (2)

Definition of the spare capacity required on link k:


Demand between each node-pair i is satisfied:

M. W
(12)
“=z 7.=1
z~~” l~i~N(N-l)
W=l
(3)
UJ=l

Number of primary lightpaths traversing link j: Constraints indicating if wavelength w is utilized for some
restoration path on link k:
N(N–1) w

(4) N(N–1) E

(13)
Number of spare channels utilized for link j:
l<k<E,l<w<W
N(N–1) w
(5)

Wavelength-continuity constraint, i.e., only one primary or


backup lightpath can use wavelength w on link j:

N(N–1)

Wavelength-continuit y constraint i.e., only one primary or


backup lightpath can use wavelength w on link j:

When a link j fails, demands between all node-pairs can still


(15)
be satisfied:

After Fault:
Definition of the total number of rerouted lightpaths be-
l<j<E,l<i<iV(iV-1) tween node-pair i when link j fails:

2.2.2 ILP2: Shared-Path Protection


(16)

Minimize the total capacity used:

Minimize ~(wj + Sj) (8)


‘j=]

0-7803-5420-6/99/$10.00 (c) 1999 IEEE


2.2.3 ILP3: Shared-Link Protection 2.3 Example ILP Solutions
In this sect;on, we present worked examples to illus-
trate the problems and understand the solutions provided
Minimize the total capacity used: by the ILPs. Consider the network illustrated in Fig. 1. As-
sume that the demand consists of two lightpaths: (a) the
Minimize ~(w, + s,) first lightpath from Node 10 to Node 6 and (b) the second
(17)
lijzhtpath from Node 15 to Node 9. The routes and wave-
J=l
le;g;hs of primary and backup lightpaths for the dedicated-
Number of lightpaths on each link is bounded: path protection (as solved by ILP1) are illustrated in Ta-
ble 1. The total capacity utilization of this solution is 16
wavelength-links (where 1 wavelength-link is a wavelength
used on a link): 6 wavelength-links for the primary light-
Demand between each node-pair i is satisfied:
paths, and 10 wavelength-links for the backup lightpaths.

Primary Lightpath Backup Lightpath


10+6 (10,11,1,6) - A, (10,9,7,6) - AI
Define the number of primary lightpaths traversing each 15+9 (15,6,7,9) - J, (15,14,12,13,1,11,10,9)
link:
N(N–1) w
Table 1: The routes and wavelengths of primary and
backup lightpaths under dedicated-path protection.

Definition of the spare capacity required on link k:


The routes and wavelength assignments for the primary
and backup lightpaths as produced by for shared-path pro-
tection by ILP2 are illustrated in Table 2. We note that
this solution utilizes 6 wavelength-links for primary paths
and 9 wavelength-links for backup paths for a total of 15
Constraints indicating if wavelength w is utilized for some wavelength-links. We note that the two working lightpaths
restoration path on link j: - (10,11,1,6), and (15,6,7,9) - are link disjoint. As a result,
upon any link failure, at most one of the two lightpaths can
fail, i.e., both lightpaths cannot fail simultaneously upon
any single-link failure. Therefore, the backup lightpaths can
share wavelengths since they will not be activated simultane-
ously. This observation leads to the routes and wavelength
assignments for the working and backup lightpaths illus-
trated in Table 2. We note that wavelength AI is shared by
both of the backup routes on link (10,9).

Primary Lightpath Backup Lightpath


10+6 (10,11,1,6) - AI (10,9,7,6) - A,
Wavelength-continuit y constraint, i.e., only one primary or 15+9 (15,6,7,9) - A, (15,14,12,13,1,11,10,9)
link restoration lightpath can use a wavelength w on link j: — Al

Table 2: The routes and wavelengths of primary and


backup lightpaths under shared-path protection.

The routes and wavelength assignments for the primary


Link restoration demands are met after link j fails for each
and backup lightpaths as produced by ILP3 (shared-link
wavelength w:
protection) are illustrated in Table 3. The solution utilizes
a total of 24 wavelength-links comprising of 6 wavelength-
links for primary paths and 18 wavelength-links for backup
paths. We note that backup wavelength-links aze not dedi-
cated, and hence, for example, the wavelength JZ is shared
by backup paths on links (10,11), (11,1), and (1,7).

0-7803-5420-6/99/$10.00 (c) 1999 IEEE


for ILPs 1, 2, and 3, respectively.
Primary Lightpath
10-+6 (10,9,7,6) - ~,
I 15 + 9 I (15,6,7,9) - & I Connections No Dedicated Shared Shared
Protection path path link
Failed link Restoration Lightpath 10 29 70 57* 96*
(6,7) (6,1,7) - A, 15 31 82 65* 92*
(7,6) (7,8,6) - AZ 20 45 122 85* 157*
(7,9) (7,1,11,10,9) - Al 25 59 163 99* 189*
(9,7) (9,10,11,1,7) - A-z 30 74 200 116* 243*
(10,9) (10,11,1,7,9) - AZ 35 84 228 126* 254*
(15,6) (15,14,12,13,5,6) - ~,

Table 3: The routes and wavelengths of primary and


backup lightpaths under shared-link protection. Table 4: Capacity utilization (in wavelength-links)
for path- and link- protection schemes for the
interconnected-rings network, with a random traffic de-
mand.
2.4 Solution Approach
The routing and wavelength assignment (RWA) problem
(with no protection for any demands) has been shown to 4 Conclusion
be NP-complete [23]. We anticipate that the problems for- Optical networks based on WDM technology can poten-
mulated in ILPs 1, 2, and 3 are NP-complete as well. We tially transfer several Gbps of data on each fiber link in the
utilized the CPLEX 4.0 software package to solve the in- network. However, the high capacity of a link has the draw-
stances of the ILPs generated for a representative network back that a link failure can potentially lead to the loss of a
topology. We note that the number of variables and the large amount of data.
number of equations for the ILPs grow rapidly with the size This study examined different approaches to survive
of the network; therefore, the ILP formulations are prac- single-link failures in an optical network. These approaches
tical only for small networks (a few tens of nodes). For are based on two basic survivability paradigms: (a) path
larger networks (a few hundreds of nodes), we need to em- protection/restoration, and (b) link protection/restoration.
ploy heuristic methods. For a one-time static network de- In path- and link- protection schemes, backup paths and
sign, it might be feasible to solve larger instances of the wavelengths are reserved in advance at the time of call setup.
ILPs (over a longer time duration), however for rapid real- Path- and link-restoration schemes are dynamic schemes in
time solutions (as may be required by the network control which backup paths are discovered (from the spare capacity
and management software to respond to changing network in the network) upon the occurrence of a failure. We formu-
conditions), solving ILPs may not be practical. lated ILPs to determine the capacity utilization for different
protection schemes for a static traffic demand. The numeri-
cal results obtained for a representative network topology
3 Illustrative Examples and Discussion and for random demands indicate that shared-path pro-
We performed our studies on an example 16-wavelength tection provides significant savings in capacity utilization
network of interconnected rings illustrated in Fig. 1. This over dedicated-path and shared-link protection schemes, and
topology was chosen to be representative of typical mesh dedicated-path protection provides marginal savings in ca-
topologies employed in telecommunications networks. We pacity utilization over shared-link protection.
chose a set of four alternate routes between each node-pair,
ensuring that all link-disjoint routes between the node-pair References
are included. For this network topology, we ran ILPs 1, 2, [1] S. Ramamurthy and B. Mukherjee, “Survivable WDM
and 3 on random demands, where each random demand had Mesh Networks, Part II - Restoration” Proc., ICC ’99,
between 10 and 35 connection requests. to appear.
[2] B. Mukherjee, Optical Communication Networks, New
3.1 Results
York: McGraw-Hill, July 1997.
We tabulate the results for the interconnected-rings net-
[3] P. Bonenfant, “Optical Layer Survivability: A Com-
work in Table 4. The first column indicates the number of
prehensive Approach,” Proc., OFC ’98, San Jose, CA,
connections in the demand. The second column indicates
vol. 2, pp. !270-271, I?ebrus.ry 1998.
the capacity utilization of the optimal routing and wave-
length assignment of the lightpaths obtained from the RWA 5Numberg that are asterisked indicate the best SOlutiOn re-
ILP formulation without any protection [23]. The third, portedby CPLEX running for 12 hours on an otherwise unloaded
fourth, and fifth columns indicate the capacity utilization HP workstation.

0-7803-5420-6/99/$10.00 (c) 1999 IEEE


[4] O. Gerstel, “Opportunities for Optical Protection and [18] H. Sakauchi, Y. Nishimura, and S. Hasegawa, “A Self-
Restoration,” Proc., OFC ’98, San Jose, CA, vol. 2, pp. Healing Network with an Economical Spare-Channel
269-270, February 1998. Assignment ,“ Proc., IEEE Globecom ’90, San Diego,
CA, pp. 438-443, December 1990.
[5] T. Wu, Fiber Network Service Survivabdity, Norwood,
MA: Artech House, 1992. [19] C. E. Chow, J. Bicknell, S. McCaughey, and S. Syed,
“A Fast Distributed Network Restoration Algorithm,”
[6] T. Wu, “Emerging Technologies for Fiber Network Sur- Proc., IEEE IPCCC ’93, Tempe, AZ, pp. 261-267,
vivability,” IEEE Communications Magazine, vol. 33, March 1993.
no. 2, pp. 58-74, February 1995.
[20] D. Marcenac, “Optical Protection Switching,” WDM
[7] J. Armitage, O. Crochat, and J.-Y. Le Boudec, “Design Forum, London, June 1998.
of a Survivable WDM Photonic Network,” Proc., IEEE
[21] F. Tillerot, “Protection Techniques for WDM Trans-
Infocom ’97, Kobe, Japan, pp. 244-252, April 1997.
port Networks,” WDM Forum, London, June 1998.
[8] O. Gerstel, R. Ramaswami, and G. Sasaki, “Fault
[22] C. Coltro, “WDM Protection,” WDM Forum, London,
Tolerant Multiwavelength Optical Rings with Limited
June 1998.
Wavelength Conversion,” Proc., IEEE Infocom ’97,
Kobe, Japan, pp. 508-516, April 1997. [23] R. Ramaswami and K. N. Sivarajan, “Routing and
Wavelength Assignment in All-Optical Networks,”
[9] M. Alanyali and E. Ayanoglu, “Provisioning Algo- IEEE/A CM Transactions on Networking, vol. 3, no.
rithms for WDM Optical Networks,” Proc., IEEE In- 5, pp. 489-500, October 1995.
focom ’98, San Francisco, CA, March 1998.

[10] S. G. Finn, M. Medard, and R. A. Barry, “A New Algo-


rithm for Bi-Directional Link Self-Healing for Arbitrary
Redundant Networks,” Proc., OFC ’98, San Jose, CA,
vol. 2, pp. 298-299, February 1998.

[11] E. Karasan and E. Goldstein, “Optical Restoration at


the Wavelength-Multiplex Section Level in WDM Mesh
Networks,” Proc., OFC ’98, San Jose, CA, vol. 2, pp.
300-301, February 1998.

[12] E. Limal, S. L. Danielsen, and K. E. Stubkjaer, “Ca-


pacity Utilization in Resilient Wavelength-Routed Op-
tical Networks Using Link Restoration,” Proc., OFC
’98, San Jose, CA, vol. 2, pp. 297-298, February 1998.

[13] R. R. Iraschko, M. H. MacGregor, and W. D. Grover,


“Optimal Capacity Placement for Path Restoration in
Mesh Survivable Networks,” Proc., ICC ’96, Dallas,
TX, pp. 1568-1574, June 1996.

[14] M. Herzberg, S. J. Bye, and A. Utano, “The Hop-Limit


Approach for Spare-Capacity Assignment in Survivable
Networks,” IEEE/A CM Transactions on Networking,
vol. 3, no. 6, pp. 775-784, December 1995.

[15] J. Anderson, B. T. Doshi, S. Dravida, and P. Har-


shavardhana, “Fast Restoration of ATM Networks,”
IEEE Journal on Selected Areas in Communications,
vol. 12, no. 1, pp. 128-138, January 1994.

[16] W. D. Grover, “The Selthealing Network: A Fast Dis-


tributed Restoration Technique for Networks Using
Digit al Crossconnect Machines,” Proc., IEEE Globe-
com ’87, pp.28.2.1-28.2.6, 1987.

[17] H. Komine, T. Chujo, T. Ogura, and T. Soejima,


“A Distributed Restoration Algorithm for Multiple-
Link and Node Failures of Transport Networks,” Proc.,
IEEE Globecom ’90, San Diego, CA, pp. 459-463, De-
cember 1990.

0-7803-5420-6/99/$10.00 (c) 1999 IEEE

You might also like