Content uploaded by Ken-ichi Sato
Author content
All content in this area was uploaded by Ken-ichi Sato
Content may be subject to copyright.
1890
IEICE TRANS. COMMUN., VOL.E90–B, NO.8 AUGUST 2007
SURVEY PAPER
Special Section on Feature Topics on Latest Trends in Optical Networks
Prospects and Challenges of Multi-Layer Optical Networks
Ken-ichi SATO†a) ,Fellow and Hiroshi HASEGAWA†,Member
SUMMARY This paper investigates the prospects and challenges of
hierarchical optical path networks. The merits and issues of introducing
higher order optical paths are elucidated. State of the art of the key enabling
technologies are demonstrated including hierarchical optical cross-connect
switch architectures, hierarchical optical path network design algorithms, a
newly developed waveband filter, and waveband conversion technologies.
key words: optical path, waveband, photonic network, multi-layer optical
paths, optical cross-connect
1. Introduction
Broadband access including ADSL and FTTH is being
rapidly adopted throughout the world and as a result traf-
fic is continually increasing. In order to cope with this,
optical transmission and node technologies are advancing
steadily. The maximum number of WDM wavelengths per
fiber now exceeds one hundred, and the maximum channel
speed of 40 Gb/s are now being intruduced. The key en-
abling technologies that enhance node throughput and cost
reduction, recognized to be optical path technologies [1]–
[3], exploit wavelength routing. Wavelength routing using
ROADMs (Reconfigurable Optical Add/Drop Multiplexers)
has recently been introduced and a large scale deployment is
being conducted in North America to develop cost-effective
networks [4], [5]. GMPLS (Generalized Multi-Protocol La-
bel Switching) controlled OXCs (Optical Cross-connects)
[6] have also been used to create nation-wide testbeds [7]–
[9]. New broadband services including IP TV and high
definition TV are now imminent and further traffic expan-
sion is expected in the near future. Cutting-edge applica-
tions including e-science, all of which need enormous band-
width, are also conceived [10], [11]. This will result in a
significant increase in the number of wavelength paths that
should be cross-connected at nodes, and hence optical node
throughput must be enhanced. One important technology
that can resolve this problem is the introduction of wave-
bands and hierarchical optical path cross-connects (HOXCs)
[12]–[14]. This paper explores recent advances in the re-
search and development of hierarchical multi-layer photonic
networks.
First, we discuss transport network evolution focus-
ing on path technologies. The hierarchical electrical paths
Manuscript received September 20, 2006.
Manuscript revised February 13, 2007.
†The authors are with the Department of Electrical Engineer-
ing and Computer Science, Nagoya University, Nagoya-shi, 464-
8603 Japan.
a) E-mail: sato@nuee.nagoya-u.ac.jp
DOI: 10.1093/ietcom/e90–b.8.1890
and the cross-connection have been utilized since the PDH
(Plesiochronous Digital Hierarchy) era [15], [16]. The digi-
tal path concept essentially reached fruition with the inven-
tion of SDH (Synchronous Digital Hierarchy) and the hi-
erarchical digital path technology formed the basis of mod-
ern telecommunications networks. The optical realization of
the path concept utilizing wavelength routing [17]–[19] has
been intensively investigated since the early 1990’s [1]–[3]
and practical implementation [7]–[9] and commercial appli-
cation started very recently as mentioned above. Some early
considerations of the hierarchical optical path were seen in
the late 1990’s [12]. In Sect.2, the role of the hierarchical
optical path, the next evolution step envisaged for the optical
path, is elucidated.
We then discuss network design issues. The network
should handle hierarchical optical paths, wavelength paths
and wavebands (WBs) that consist of multiple wavelength
paths. Several studies have targeted the development of
hierarchical optical path network design algorithms [20],
[21], and have demonstrated the advantage of introducing
HOXCs; most works evaluated the effectiveness in terms
of the total number of HOXC ports compared to that of
single-layer OXCs. Another study [22], [23] showed that
a HOXC with non-uniform WB size can not only improve
node throughput but also reduce cross-connect node cost.
These design algorithms tend to require a complex proce-
dure for resolving wavelength and the waveband assignment
problem when wavelength/waveband conversion is not uti-
lized. Recent algorithms [21] introduce a novel trafficde-
mand expression in a Cartesian product space and are ex-
plained here in detail.
Next, we discuss some key hardware technologies that
are needed to develop HOXCs. They include hierarchi-
cal optical cross-connect architectures that consist of wave-
band cross-connects (BXCs) and wavelength path cross-
connects (WXCs). Few studies have been done so far on
these switch architectures. The switch scale necessary for
single-layer OXCs and HOXCs are evaluated considering
the different levels of switch flexibility that yield graceful
expansion. The conditions wherein HOXCs are superior to
single layer WXCs are clarified. We then discuss wave-
band multi/demultiplexers (WB MUX/DEMUX). We de-
scribe a newly developed WB MUX/DEMUX [24], [25] that
uses two concatenated cyclic AWGs (Arrayed-Waveguide
Gratings). The salient feature of the device is that it can
accommodate multiple input fibers simultaneously and the
port utilization of the device is shown to be 100% with bi-
Copyright c
2007 The Institute of Electronics, Information and Communication Engineers
SATO and HASEGAWA: PROSPECTS AND CHALLENGES OF MULTI-LAYER OPTICAL NETWORKS
1891
Fig. 1 Transport network technology evolution in Japan.
directional input fibers. We finally discuss recent advances
in waveband conversion technologies. The technologies dis-
cussed in this paper will need to be consolidated to realize
hierarchical optical path networks that offer a solution to the
expected traffic increase and the ever increasing power con-
sumption of node systems.
2. Hierarchical Path Network Architectures
2.1 Role of Path and Hierarchical Electrical Path Archi-
tectures
A layered transport network architecture has been developed
that offers simplified network design, development, and op-
eration, and offers smooth network evolution [26]. It also
makes it easy for each network layer to evolve indepen-
dently of the others by capitalizing on the introduction of
new technologies specific to each layer. The layering con-
cept has been extensively discussed within ITU-T for the
SDH transport network [27], [28]. Traditionally, the net-
work can be divided into three layers from the viewpoint of
functions: a physical media layer, a transmission path layer,
and a communication circuit (/flow) layer. The circuit/flow
provides end-to-end communication dynamically or on the
basis of short term provisioning. The circuit/flow layers
are dedicated to specific services. The transmission me-
dia, which physically interconnect nodes and/or subscribers,
is constructed based on long-term provisioning; geographi-
cal conditions are taken into consideration. The path layer
bridges these two layers and plays an important role in con-
structing reliable and flexible networks [26]. Important at-
tributes of a path are; a grouped circuit/flow serving as a unit
of network operation (including traffic engineering), design
and provisioning, and an object to be manipulated for re-
covering from node and transmission line/system failures.
Network flexibility can also be enhanced with path layer
control. Path layer functions can be materialized with intel-
ligent ADM (Add/Drop Multiplexer) systems and/or digital
cross-connect systems and control systems.
Different paths have been introduced in accordance
Tab l e 1 Multiplexing and path realization technologies.
with the evolution of transfer techniques; digital paths
(virtual containers) in SDH [29], Virtual Paths in ATM
(Asynchronous Transfer Mode) [30]–[32], Label Switched
Paths in IP/MPLS (Multi-Protocol Label Switching [33],
[34]), and optical paths for photonic networks [1]–[3]. The
Japanese example of path technology evolution is depicted
in Fig. 1. The characteristics of different paths are summa-
rized in Table 1. GMPLS [35], [36] has been proposed as
a control technique. It aims at providing a common control
mechanism among different path layers (and fiber layer) that
utilizes IP-based signaling; intensive technical development
and standardization activities have been undertaken.
2.1.1 Hierarchical Path Arrangement
Hierarchical path arrangements have been developed and
widely utilized. One typical example is a hierarchical SDH
digital path network, which is shown in Fig. 2 [27]. In circuit
switching, connection admission control is done on the basis
of the usage of paths that have been pre-established between
switches to meet traffic demands. The role of the path is
called service-access. In SDH networks, the service-access
path bandwidth is 1.5 Mb/s(24×64 kb/s channels; Japan
and North America) or 2.0 Mb/s(32×64 kb/s channels; Eu-
rope) for telephone service. Transmission system bit-rates,
1892
IEICE TRANS. COMMUN., VOL.E90–B, NO.8 AUGUST 2007
Fig. 2 Hierarchical path layer structure in SDH.
however, are much larger than the service-access path band-
widths and now reach forty giga bits per second. Traffic
demands between any two nodes are not always sufficient to
fill the bandwidth. This has triggered the use of another path
stage, higher order paths (49 or 150 Mb/s), to accommodate
several service-access paths. The higher order paths are ac-
commodated within a higher capacity transmission link and
are cross-connected [16], [37] at transit nodes to construct
an economical transport network. The role of the higher or-
der paths is, therefore, called trans-access. The higher order
paths can be used to separate IP networks and traditional
telephone and data networks and to share a physical media
network [38] among the services. The higher order paths are
usually used for network protection/restoration against fail-
ures. The classification of service-access and trans-access
is not strict, since higher order paths can be directly used to
provide large bandwidth services such as video and leased
line services. The hierarchical bandwidth paths are thus
very useful for creating an efficient network and hence are
widely utilized in the SDH network. On the other hand,
the hierarchical path structure and the multiple stages of the
multi/demultiplexing, can complicate path accommodation
designs and may lead to poor link utilization. For exam-
ple, if the multiplexing efficiency of the lower stage path
into the next stage is 0.8 for each multiplexing stage, three
stage multiplexing results in the final link utilization of 0.5
(=0.83). This inefficiency can be removed with the logi-
cal realization of paths such as VPs in ATM and LSPs in
IP/MPLS. In LSP, a shim header is defined to be hierarchical
at the initial standardization stage [33] so that capabilities of
LSPs can be enhanced as needed.
2.2 Optical Path and the Hierarchical Architectures
Optical paths provide for the wavelength routing of signals,
which eliminates rather expensive OE/EO (Optical to Elec-
trical/Electrical to Optical) conversion, enables bit rate and
transfer mode (protocol) independent cross-connects, dra-
matically enhances node throughput, and greatly reduces
node system power consumption. These are discussed in
Fig. 3 Different paths to network throughput expansion.
detail in [26]. Optical paths are first utilized as trans-access
paths since the bandwidths are usually higher than 2.5 Gb/s;
they accommodate electrical lower order paths such as SDH
digital paths, and LSPs.
Routing functions in the optical domain with optical
paths were first put into commercial use with the employ-
ment of ROADMs. OXCs have been used for creating
nation-wide testbed networks. With regard to optical tech-
nologies, however, large scale optical switch technologies
are not mature and the restrictions are still substantial. When
the large scale deployment of OXCs will start depends on
the degree of both the traffic and switch technology matu-
rity. The OXCs will be deployed to interconnect multiple
ROADM rings or to create mesh-based networks.
One of the salient features of optical paths is that switch
complexity does not depend on the bit rate carried by an op-
tical path. With electrical technologies, switching becomes
more and more difficult and consumes more electrical power
as the bit rate increases. Thus, the wide deployment of op-
tical path technologies will be driven by the traffic increase,
and the limit in terms of power consumption and through-
put inherent in electrical switching. With regard to the node
throughput enhancement, different directions have been ex-
plored as shown in Fig. 3 [39]; introduction of higher-order
optical paths (wavebands) and the introduction of optical
burst switching. It will take much longer for the latter to
yield a practical application than the former, and so only the
former is discussed below.
Hierarchical optical path arrangement is a natural ap-
proach when traffic increases, and when optical layer ser-
vices such as OVPN (Optical Virtual Private Network) ser-
vices, lambda leased line services, and lambda switched ser-
vices for e-science emerge. The optical paths are used as
a service-access path, and the introduction of higher-order
optical paths is envisaged. Another important point is that
some optical switches support optical signals with a wide
range of wavelengths; this means that the same switches
can be used for switching multiple optical paths. Switch-
ing multiple optical paths or switching wavebands can re-
duce total switch size (necessary number of cross-connect
switch ports) substantially. This mitigates one of the ma-
jor present challenges in creating large scale optical cross-
connects. The hierarchical path structure can degrade link
utilization, and this effect may be enhanced by the addi-
tional complexity created by the waveband/wavelength as-
signment problem. This problem can be mitigated by the
SATO and HASEGAWA: PROSPECTS AND CHALLENGES OF MULTI-LAYER OPTICAL NETWORKS
1893
development of effective hierarchical optical path network
design algorithms. They minimize total network cost that
consists of node and link cost, or total number of optical
cross-connect ports. These merits and demerits are evalu-
ated in the following sections.
3. Hierarchical Photonic Network Enabling Technolo-
gies
In order to create hierarchical photonic networks that war-
rant introduction, substantial technology development is
still necessary. Figure 4 depicts some of the key issues
that need further R&D and are discussed in the following
sections. Recent development and future challenges are de-
scribed.
3.1 Hierarchical Optical Path Network Design
The waveband routing and waveband assignment problem
of multi-granular optical networks is a generalization of
the single-granular optical network design problem. The
problem aims at minimizing cost functions subject to wave-
band and wavelength continuity constraint. Analogous to
the single-granular case [40], the problem is inherently NP-
complete and can equivalently be formulated as a combina-
torial optimization problem that targets minimizing the total
number of optical ports [41]–[43] of cross-connect systems
or maximizing the utilization of fiber capacity [44]. The
number of binary variables in the combinatorial optimiza-
tion problem explosively increases with network size. This
characteristic makes the problem computationally impossi-
ble to accurately solve for large networks. Indeed, previous
publications that tackled the combinatorial optimization for-
mulation gave up on exact solutions and provided alternative
algorithms based on heuristics or relaxation instead. Several
such methods have been proposed [41]–[47]; they are cate-
gorized as follows.
Fig. 4 Key technologies necessary to develop hierarchical optical path
networks.
3.1.1 Grooming of Wavelength Paths Having Common
Source/Destination or Partially Shared Routes [46],
[47]
Wavelength paths with the same destination are first col-
lected and then accommodated within wavebands [46]. This
approach can also be applied to wavelength paths having the
same source. Since the requirement of having the same des-
tination is rather restrictive, an alternative [47] computes the
routes of all wavelength paths first and then locates sets of
wavelength paths that share common routes at least for some
significant distance. Wavebands are then constructed along
these shared routes. This kind of method is simple because it
is almost equivalent to a design method for single-layer opti-
cal path networks. However, further aggressive grooming is
necessary to improve waveband utilization ratio especially
when traffic demand is not large.
3.1.2 Waveband Tunnel Construction First [45]
The potential traffic loads of all fibers are estimated first
and the amounts of potential incoming and outgoing traf-
fic of each node are defined by using the potential loads.
After connecting pairs of nodes that have large incoming
and outgoing traffic loads by waveband tunnels, wavelength
paths are accommodated within the waveband tunnels based
on the shortest paths. Reference [45] focuses on dynamic
wavelength paths allocation that is necessary in realizing op-
tical circuit-switch type services.
3.1.3 On Demand Waveband Assignment [42], [43]
This approach assigns the routes of wavelength paths one by
one. For each route assignment, the optimal route is iden-
tified considering existing waveband or fiber tunnels. If no
proper tunnel is available, new tunnels are constructed. The
approach tries to find only a shortest path at each wavelength
path accommodation. The waveband tunnels will be satis-
factory filled whereas it lacks global optimization of wave-
band tunnel placement.
3.1.4 Relaxation-Based Methods [41], [44]
One approach [41] divides the original combinatorial opti-
mization problem into three easier sub-problems that consist
of other combinatorial optimization problems. An alterna-
tive approach [44] makes the original problem easier by us-
ing Lagrangean Relaxation. Both methods then resolve the
easier problems so as to derive an approximation of the op-
timal solution. This kind of approach is useful mainly for
small networks where the sub-optimal solutions can be de-
rived with a slight degree of relaxations.
3.1.5 Cluster-Search Method in a Source-Destination
Cartesian Product Space [21], [48]
Except for the relaxation-based method, the aim of the
1894
IEICE TRANS. COMMUN., VOL.E90–B, NO.8 AUGUST 2007
Fig. 5 Traffic demand and clusters in S-D Cartesian product space.
above heuristics is to improve utilization of waveband paths
as well as to reduce node cost incurred by carrying each
wavelength path by minimizing wavelength path cross-
connection at intermediate nodes. Suppose that we have
waveband paths between every node pairs where trafficde-
mand exist, then the node cost of carrying each wavelength
is generally reduced, however, waveband utilization may be
degraded.
A heuristic algorithm that is based on a different strat-
egy was recently proposed [21], details and a performance
analysis are given in [48]. To minimize the total cost, which
consists of node cost and link cost, the wavelength path
cross-connection at intermediate nodes must be reduced and
at the same time waveband utilization must be maximized.
We must therefore search for a set of wavelength paths that
are efficiently carried by a waveband path and can suffi-
ciently occupy the waveband path. A space, named “s-d
(source-destination) Cartesian product space,” is newly de-
fined to evaluate closeness among wavelength paths. The
space can be used to effectively search for waveband paths
that can reduce total cost (See Fig. 5). For each set of wave-
length paths, a waveband path is constructed so as to max-
imize the degree of cost reduction. The wavelength paths
that are not accommodated with the first step are finally ac-
commodated in wavebands by identifying the shortest paths
in a multi-layered graph considering waveband tunnels (See
Fig. 6).
The multi-layer optical network can reduce the num-
ber of optical ports, but will lower fiber utilization levels. It
is therefore important to clarify the conditions under which
the waveband approach is attractive. Figure 7 shows the cost
reduction possible with the multi-layer optical path network
in a comparison against a single layer one for a 9 ×9 poly-
grid network with randomly distributed wavelength path de-
mands. The vertical axis of the graph is relative cost normal-
ized by the cost of a single layer optical network designed
by locating the shortest paths with re-routing of wavelength
paths in sparsely used fibers. End-to-End represents the
method that accommodates wavelength paths within a wave-
band path that has the same source and destination as the
wavelength paths. Even this simple scheme, End-to-End,
achieves approximately 40% cost reduction when the aver-
age wavelength path demand between nodes equals wave-
Fig. 6 Multi-layered graph where each layer is assigned to each
wavelength.
Fig. 7 Comparison among proposed and conventional algorithms (end-
to-end and BPHT).
band width. With the method in [21], [48], the cost reduc-
tion is greatly enhanced even for smaller traffic demands.
The investigation provides not only performance compar-
isons of different methods, but also clarifies the applicability
of the multi-layer optical path network.
3.2 Multi-Layer Optical Path Cross-Connect Switch Ar-
chitectures
3.2.1 Basic Cross-Connect Node Architecture
In a single-layer OXC, wavelength paths are de-multiplexed
by DEMUX, routed by WXC (Wavelength path XC), and
then recombined by MUX (See Fig. 8(a)). Wavelength con-
verters can be inserted before or after WXC so as to avoid
wavelength collision. The number of wavelength convert-
ers may be limited because of their cost; collision can be
reduced by applying effective routing and wavelength as-
signment algorithms [49]–[52].
The switch architectures for multi-layer optical path
networks are determined to suit the multi-layer optical path
arrangements. Figure 8(b) depicts a generic configuration
of the hierarchical optical cross-connect switch architecture.
HOXC is divided into two parts as shown in Fig. 8(b). One
SATO and HASEGAWA: PROSPECTS AND CHALLENGES OF MULTI-LAYER OPTICAL NETWORKS
1895
(a) Single layer switch architecture. (b) Hierarchical switch architecture. (c) Non-hierarchical switch architecture.
Fig. 8 Different cross-connect switch architectures.
part consists of WB MUX/DEMUX and BXC (waveband
XC) for routing higher-order waveband paths, and the other
part consists of MUX/DEMUX and WXC for routing lower-
order wavelength paths. The routing of optical paths should
be processed by BXC as much as possible so as to make
the best use of the HOXC arrangement. In other words, the
WXC part should only be used when WB routing is not pos-
sible. This is possible due the network design algorithms,
which accommodate wavelength paths into wavebands and
wavebands into fibers as described in the previous section.
Several studies [53] proposed non-hierarchical OXCs
(see Fig. 8(c)) where wavelength paths and WBs are ac-
commodated simultaneously in a fiber. At a node, wave-
length paths and WBs are fed to a WXC and a BXC, re-
spectively. With this arrangement, they are treated inde-
pendently. Grooming of wavelength paths at the interme-
diate OXCs are not done and as a result no interworking be-
tween wavelength path cross-connects and waveband cross-
connects is required, which simplifies the switch architec-
tures. This arrangement can decrease the number of the total
cross-connect switch ports needed when the number of opti-
cal paths per waveband is within a certain range [53], how-
ever, if the traffic pattern or traffic volume changes, the opti-
mum wavelength path and waveband configuration changes
and as a result, the effectiveness decreases. Furthermore, the
flexibility of the cross-connect switch is very limited since
the ratio of the number of wavelength paths to that of WBs
accommodated within a fiber has to be fixed to optimize the
switch architecture. We, therefore, do not discuss this ap-
proach below.
Different switch architectures and technologies have
been utilized to create optical switches. They include 2-
dimensional [54] and 3-dimensional (3D) MEMS [55], a
PLC (Planar Lightwave Circuit) [56] switch, and a mechan-
ical fiber switch. 3D MEMS switches are attractive due to
their high functionality and low optical insertion loss and
crosstalk [57]. The WSS (Wavelength Selective Switch)
[58], [59], which was originally proposed as the Delivery
and Coupling Switch (DC-SW) [60]–[62] in 1993 as de-
scribed below, uses 3D MEMS switches, and is going to be
utilized to develop ROADMs for applications in metropoli-
tan area networks. Since the micro mirrors must be con-
trolled in three dimensions and require extreme assembly
and operation precision; reliability of large scale 3D MEMS
switches in various conditions must be verified in more de-
tail. On the other hand, 2D MEMS and PLC switches (two
dimensional matrix-type switches) are attractive due to their
proven reliability [63] and high manufacturing yields. This
paper, therefore, focuses on BXC and WXC architectures
assuming the use of the latter switch technologies.
3.2.2 Flexibility for Expansion
To realize the economical introduction of optical cross-
connects and subsequent expansion to cope with traffic
growth, the system must offer modular growth capability.
The following attributes of expansion are important and are
explained in Fig. 9.
•modular growth regarding fiber
The number of input/output fiber ports is incrementally in-
creased as traffic demand increases. This minimizes the ini-
tial investment with regard to fiber ports.
•modular growth regarding WB
The number of WBs is incrementally increased as trafficde-
mand increases. This minimizes initial investment with re-
gard to the ports connecting WB MUX/DEMUX and BXC.
•modular growth regarding wavelength paths in WB
The number of wavelength paths in WB is incrementally
increased as traffic demand increases. This minimizes initial
investment with regard to WXC.
•modular growth regarding add/drop WB
The number of added WBs from WXC to BXC or dropped
WBs from BXC to WXC is incrementally increased as traf-
fic demand increases. This minimizes initial investment
with regard to the ports connecting BXC and WXC.
In the following discussion, Nis the number of in-
put/output fibers and mis the number of wavelength paths
per fiber. Variable xis introduced to stand for the ratio of
wavelength paths that are to be added/dropped for connec-
tions to electrical systems (digital cross-connect systems or
routers, etc.) at a node.
The first flexibility can be realized by using the Deliv-
ery and Coupling Switch (DC-SW) [60]–[62]. One example
for the single layer OXC is shown in Fig. 10(a). This OXC
utilizes N,m(1 +x)×N(1 +x) DC-SWs and one switch is
dedicated to each input fiber. If this architecture is used for
1896
IEICE TRANS. COMMUN., VOL.E90–B, NO.8 AUGUST 2007
(a) Modular growth: fiber.
(b) Modular growth:
waveband.
(c) Modular growth:
wavelength path.
(d) Modular growth: add/drop
waveband.
Fig. 9 Different levels of flexibility.
(a) Single layer OXC architecture that
uses DC-SWs.
(b) Matrix type m×NDC-SW. (c) Tree type m×NDC-SW.
Fig. 10 Single layer OXC using Delivery and Coupling switches.
BXC in HOXC, then the MUX in the figure should be WB
MUX. A matrix type arrangement of the m×NDC-SW is
shown in Fig. 10(b) and a tree type arrangement is shown in
Fig. 10(c). This DC-SW configuration was originally pro-
posed in 1993 [60]–[62] and has been used in the optical
cross-connect systems used to create nation-wide testbed
networks [7]. The DC-SW is now often called the WSS
(Wavelength Selective Switch). In Figs. 10(b) and 10(c), N
MUXs are used when the wavelength of each input signal is
fixed, however, an optical star coupler can be used instead
of the MUX when the wavelength of each input signal can
change. The DC-SW allows any of the mincoming opti-
cal signals to be connected to any of the Noutgoing ports.
These outgoing port signals are coupled to output fiber ports
by using a star coupler (SC). DC-SW is composed of 1 ×2
optical SW elements. This architecture makes the value of x
(Fig. 10(a)) equal among different fibers. The configuration
provides modular growth capability in terms of input and
output fiber pairs [26], [60].
The second flexibility can be realized by using a single-
layer OXC that has a N(1 +x)×N(1 +x)matrixSWfor
each optical path group with the same wavelength as shown
in Fig. 11. This architecture makes the value of xequal
among different wavelengths. The number of matrix SWs,
m, equals the number of wavelength paths per fiber. Each
N(1 +x)×N(1 +x) SW assigns an output fiber port to
each wavelength path. This configuration has modularity
in terms of wavelength path. HOXC architectures can re-
Fig. 11 Single layer OXC that consists of matrix switches dedicated to
the same wavelength paths.
alize this flexibility when the BXC utilizes the above type
of matrix switch for each waveband group in combination
with a WXC that has modular growth capability in terms of
wavebands.
The third flexibility is realized in the most simple case
at the sender and receiver side; corresponding transceivers
and receivers need to be added, which will be almost the
same procedure as the wavelength increase process for
WDM transmission channel increase. In regard to the
switch, if the switch architecture uses switches, each of
which correspond to a wavelength path group with the same
wavelength, then the addition of a new wavelength path
SATO and HASEGAWA: PROSPECTS AND CHALLENGES OF MULTI-LAYER OPTICAL NETWORKS
1897
Fig. 12 Single layer OXC with a large matrix switch.
group will allow modular growth of the switch.
The fourth flexibility is realized in HOXC when WXC
has modular growth capability in terms of wavebands.
3.2.3 Single-Layer OXC SW Architecture
First, we briefly discuss single-layer OXC switch (SW) ar-
chitectures. The simplest switch architecture for a single-
layer OXC is one that uses large mN(1 +x)×mN(1 +x)
matrix switches as shown in Fig. 12 or that uses the equiva-
lent Clos switching network [64]. It is possible to route an
optical path from any input port to any output port. The con-
figuration does not possess any of the flexibilities described
above.
For the single-layer OXC with large matrix switch,
the switch scale is about ten times bigger than that of any
other SW architecture considered here. DC-SW is the
best in terms of SW size, however, it requires many more
MUXs/DEMUXs than the other architectures; the number
is proportional to order N2. The SW architecture requires
NSCs which leads to 10 ×log NdB intrinsic loss, which
may be an issue when Nbecomes large. The single-layer
OXC that consists of matrix SWs dedicated to the same-
wavelength paths, shown in Fig.11, can be constructed with
relatively small scale switches and the necessary number of
MUXs/DEMUXs is low. As a result, the SW architecture
based on matrix SWs for each same-wavelength path group
is attractive for constructing single-layer OXCs with regard
to cost and loss, if modularity in terms of input/output fibers
is not necessary.
3.2.4 HOXC SW Architectures
Different HOXC switch architectures have been developed
and compared in terms of cost, loss and flexibility [65].
The WXC SW architecture adopted for developing HOXC,
shown in Fig. 11, is attractive for single-layer OXCs in
terms of switch scale and MUX requirement, as discussed
above. Let Nbe the number of fibers, mthat of wave-
length paths per fiber, kthat of WBs per fiber, and lthat
of wavelength paths per WB. Variable yrepresents the ra-
tio of WBs dropped at a node to total WBs delivered to
the node (or the ratio of WBs added at a node to total WBs
launched from the node). Thus, the product of kand lequals
m,andyranges from 0 to 1. Among switches compared,
the most attractive switch architecture for BXC is the one
shown in Fig. 11 for the WXC case. The BXC consists of
(1 +y)N×(1 +y)Nmatrix SWs for each WB. The archi-
tecture makes the value of yequal among the different WB
groups. Each (1 +y)N×(1 +y)NSW assigns an output port
to each WB. This configuration has modular growth capa-
bility in terms of WBs, and can reduce WXC switch size
by setting the proper limit (y) to the number of add/drop
WBs. Namely, this architecture realizes the fourth flexibil-
ity described above, modular growth capability in terms of
add/drop WBs. This architecture, therefore, demonstrates
very attractive characteristics for developing BXCs in terms
of cost, loss, and flexibility [65].
As discussed for the single-layer OXC, the DC-SW
offers the smallest switch scale, but the number of WB
MUXs/DEMUXs needed increases with the second order of
power N. This architecture will be cost-effective when the
large scale integration of WB MUXs becomes possible and
a substantial cost reduction is attained.
3.2.5 Comparison of Switch Scale between Single-Layer
OXC and HOXC
For OXCs, some of the input and output ports of a switch
are used for intra-office interconnections between optical
switches and electrical systems such as electrical digital
cross-connect systems and routers. To minimize total op-
tical switch size, it is essential to realize add/drop ports in
an optimal manner. There are two ways to realize add/drop
ports [65]. One utilizes a part of the matrix SW ports as
add/drop ports. Because this system can restrict the num-
ber of add/drop ports to those needed, this system can min-
imize the number of add/drop ports. However, this system
increases matrix SW size. The other realizes add/drop ports
based on the use of 2 ×2SWs.
A2×2 SW is added to each output port of the ma-
trix SW. With this arrangement, the matrix SW size does
not depend on the existence of add/drop ports, that is, this
system can minimize the matrix SW size. However, this ar-
rangement cannot restrict the number of add/drop ports and
allows all input signals to the matrix switch to be dropped.
Here, the ratio of add/drop wavelength paths (between
optical switches and electrical systems) to all wavelength
paths delivered to the node is denoted as xas explained be-
fore. For HOXCs, the add/drop wavelength paths between
optical switches and electrical systems are restricted through
two stages of switches, BXC and WXC. BXC restricts the
ratio to y, as described before, and then WXC restricts it to
x, which results in the ratio of xy. In the following evalua-
tions for HOXCs, xis set to 1, in other words, the restriction
is done only in the BXC stage. Comparisons of switch scale
between single-layer OXC and HOXC have been done [65]
with the condition that xfor single-layer OXCs and yfor
HOXCs were equal. Figure 13 shows evaluated switch size
for single-layer OXC and HOXC, with the two variants de-
fined as using and not using 2 ×2SWs. InFig.13,Nis set
at 8, mis set at 96, and kis set at 8. In the single-layer OXC
that uses 2×2 SWs, the number of add/drop ports can not be
1898
IEICE TRANS. COMMUN., VOL.E90–B, NO.8 AUGUST 2007
Fig. 13 Comparisons of switch scale between single-layer OXC and
HOXC.
(a) Continuous WB arrangement.
(b) Discontinuous (Interleaved) WB arrangement.
Fig. 14 WB arrangement.
restricted and is constant, mN, regardless of x.Byusing2×2
SWs, switch size can be greatly reduced in both the single-
layer OXC and the HOXC. This advantage becomes more
prominent as x(y) becomes smaller. Figure 13 elucidates
the region and the conditions in which HOXCs are more ef-
fective. When x(y) is less than 0.5, HOXC switch size is at
least 55% smaller than that of the single-layer OXC.
3.3 Waveband Multiplexers/Demultiplexers
Waveband multiplexers/demultiplexers (WB MUX/
DEMUX) are one of the other key components in develop-
ing HOXCs. We first describe different WB arrangements
for the hierarchical optical path network. The first one,
a conventional arrangement, is the continuous wavelength
path arrangement shown in Fig. 14(a). The other, newly pro-
posed in [24], offers the discontinuous (interleaved) wave-
length path arrangement depicted in Fig. 14(b). From the
networking point of view, the WB arrangement has virtu-
ally no effect on network provisioning or OA&M (Opera-
tion, Administration and Maintenance).
WB MUX/DEMUX can be realized in different ways.
A thin-film filter has been reported that offers 8-skip-0 band
operation supporting a total of 32 channels with 100-GHz
Fig. 15 An example of WB MUX/DEMUX with continuous waveband
arrangement.
spacing [66]. Since it requires 409 dielectric thin-film lay-
ers [66], the manufacturing challenge is significant. Fur-
thermore, it demonstrates strong non-linear dispersion at
the band edges. Other band-filters based on arrayed wave-
guide gratings (AWG) have been reported. They realize
a 17-skip-3 band with a total of 100 channels and 100-
GHz spacing [67] and an 8-skip-0 band with a total of 40
channels and 100-GHz spacing [68]. The AWG band fil-
ters are susceptible to manufacturing errors that yield ad-
jacent crosstalk. A more recent proposal [24] utilizes two
concatenated cyclic AWGs. The point of the novel WB
MUX/DEMUX is that it retains multi/demultiplexing gran-
ularity at the individual wavelength channel level while out-
putting the WBs at different ports. This means that conven-
tional cyclic AWGs can be used. The salient feature of the
proposed WB MUX/DEMUX is that it can accommodate
multiple input fibers simultaneously and demultiplex each
band to different output fibers. Two arrangements of paired
AWGs have been identified. One eliminates crossings of any
waveguides that connect the two AWGs, which can be real-
ized with a continuous waveband arrangement. The other
arrangement requires the waveguide crossings, however, it
can reduce band crosstalk and maximize output port utiliza-
tion efficiency; it can be realized using the discontinuous
waveband arrangement [24].
Figure 15 depicts an example of the WB MUX/
DEMUX with continuous waveband arrangement [24]; it
supports a total of 32 100-GHz spaced channels in eight
bands. The WB MUX/DEMUX consists of two cyclic
AWGs [69], 16 ×16 and 20 ×20, and waveguides connect-
ing them. The device can accommodate four input fibers
(A-D) simultaneously, each of which carries 16 wavelength
channels. The device demultiplexes each band on each input
fiber and outputs each band separately from 16 (4 bands per
fiber ×4 input fibers) output ports out of the 20 ports. This
is made possible by the cyclic nature [69] of the AWGs; the
wavelength λiat input port #a(a=1∼N) is output from
output port #b(b=1∼N)wheni=(a+b−2) mod N+1.
Please note that this specific connection arrangement pre-
SATO and HASEGAWA: PROSPECTS AND CHALLENGES OF MULTI-LAYER OPTICAL NETWORKS
1899
Fig. 16 An example of WB MUX/DEMUX with discontinuous WB
arrangement.
vents waveguide crossing, which is advantageous to the
monolithic realization of the device. The device feature is
generalized as follows. If the number of channels is m,and
m=(# of bands per fiber, k)×(# of channels in each band, j),
then AWG 1 is m×mand AWG 2 is (m+k)×(m+k). This de-
vice can support min {m/k,m/j}input fibers simultaneously.
The detailed formulation regarding the configurations of the
component AWGs and the connection patterns that create
WB MUX/DEMUX and prevent waveguide crossings are
analyzed in [25].
In the discontinuous WB arrangement, the WB
MUX/DEMUX is realized with two cyclic m×mAWG s .
This WB MUX/DEMUX can accommodate m/kinput fibers
in one device. Figure 16 depicts an example of the WB
MUX/DEMUX with the discontinuous waveband arrange-
ment, when m=16 and k=4. The WB MUX/DEMUX
consists of two cyclic 16 ×16 AWGs, and waveguides con-
necting them. This WB MUX/DEMUX has two major dif-
ferences from the continuous WB arrangement. One is that
no connection patterns between AWG 1 and AWG 2 are
possible that avoid crossings. The other is that this dis-
continuous WB arrangement can enhance AWG port usage.
Furthermore, this MUX/DEMUX permits the bi-directional
connection of multiple input fibers. This can maximize in-
put and output port utilization efficiency of the AWGs; it has
been proven to reach 100% [25].
This device will be realized monolithically using PLC
technologies. To confirm device feasibility, experiments us-
ing two separate uniform-loss and cyclic frequency (ULCF
[70]) 32 ×32 AWGs connected with 32 optical fibers have
been performed [24]. The device can support four input
fibers simultaneously. Each fiber accommodates 32 100-
GHz spaced channels (192.8+0.1×nTHz; n=0–31) on
an ITU-T grid. Each fiber carries four discontinuous bands.
Each band consists of eight wavelength channels. One of the
important characteristics of the MUX/DEMUX is crosstalk.
The output WB configurations from the second AWG differ
according to the input fiber connection patterns, which de-
termine the magnitude of crosstalk level at each band. The
crosstalk deteriorates when neighboring output ports carry
the same WB from different input fibers or carry neighbor-
ing WB on the same or different input fibers. The device
allows different input fiber connection patterns and the pat-
terns that minimize crosstalk have been identified and have
been experimentally confirmed [24].
3.4 Waveband Conversion Technologies
Waveband conversion resolves waveband collision when
routing wavebands and increases waveband accommodation
efficiency within each fiber. The technology is much more
difficult than single wavelength conversion. The require-
ments for waveband conversion are low crosstalk, small
guard band, and bit rate and modulation format indepen-
dency. Parametric wavelength conversion is one solution,
however, it usually requires wide guard bands around the
pump wavelength so as to minimize signal quality degrada-
tion due to the parametric crosstalk among the WDM sig-
nals. The degradation occurs in the conversion process and
is enhanced when the channel spacing narrows. Conversion
of 100 GHz-spaced 32 channels has been demonstrated with
a 12.8 nm guard band, using fiber four wave mixing [71].
The performance has been improved by using period-
ically poled LiNbO3(PPLN) waveguides based on quasi-
phase matching (QPM) [72], [73]. Due to the strict phase
matching condition possible, the efficiency can be raised,
and narrower-channel-spaced wavelength conversion with
low crosstalk and small guard band be expected. Us-
ing a high efficiency PPLN waveguide, inter-band wave-
length conversion of 25 GHz-spaced, 1.03 Tbit/s (103 ×
10 Gbit/s) DWDM signals (C-band to L-band) has been
reported [74]. The guard band was quite small at 8 nm
(4 nm ×2) and converted signal Q-factors exceeded 15.6dB.
The technology was also used to create a polarization-
independent waveband converter that supported 25 GHz-
spaced, 10 Gbit/s×64 channels. It was coupled with an
8×8 PLC optical switch to demonstrate waveband path
switching with a throughput of 5.12 Tbit/s across 32 km of
field installed fibers in JGN-II (Japan Gigabit Network-II)
testbed [75].
It has been shown that the two wavelength pump
method can convert an arbitrary wavelength to another ar-
bitrary wavelength [76], [77], and further, the simultane-
ous and arbitrary wavelength conversion of a waveband is
possible [78] by using a LN waveguide having multiple
QPM wavelengths [79], [80]. Waveband conversion tech-
nologies continue to advance, however, further improve-
ments in terms of conversion efficiency and cost are required
before their practical application can be considered.
4. Conclusions
This paper discussed the next step in photonic network evo-
lution, the creation of hierarchical optical path networks.
Key enabling technologies and recent advances were inves-
tigated. The introduction of wavebands was shown to reduce
optical switch size at cross-connects, which mitigates one of
the major barriers to the implementation of large throughput
1900
IEICE TRANS. COMMUN., VOL.E90–B, NO.8 AUGUST 2007
optical cross-connect systems. One of the other obstacles,
network design complexity, was shown to be effectively re-
solved by a newly developed hierarchical optical path net-
work design method that introduces a traffic demand expres-
sion in a Cartesian product space. Some of the key compo-
nent technologies were also discussed and a new waveband
MUX/DEMUX was demonstrated. As discussed in the pa-
per, the hierarchical path arrangement is an efficient and ef-
fective means to create reliable large-throughput networks.
The hierarchical optical path network will be implemented
in the not so distant future when traffic volumes warrant it.
Acknowledgments
Part of this work was supported by NICT (National Institute
of Information and Communications Technology) and JST
(Japan Science and Technology Agency), to which we are
deeply indebted. We are also grateful to Dr. Atushi Takada,
NTT Network Innovation Laboratories, Mr. Isao Yagyu and
Mr. Shoji Kakehashi, Nagoya University, for their useful
discussions.
References
[1] K. Sato, S. Okamoto, and H. Hadama, “Optical path layer technolo-
gies to enhance B-ISDN integrity,” Proc. ICC’93, pp.1300–1307,
Geneva, May 1993.
[2] K. Sato, S. Okamoto, and H. Hadama, “Network integrity enhance-
ment with optical path layer technologies,” IEEE J. Sel. Areas Com-
mun., vol.12, no.1, pp.159–170, Jan. 1994.
[3] K. Sato, S. Okamoto, and I. Tokizawa, Patent no.3221457 and
3247886, Optical Communications Network.
[4] I. Redpath, D. Cooperson, and R. Kline, “Metro WDM networks de-
velop an edge,” Proc. OFC/NFOFC 2006, NThC1, Anaheim, March
2006.
[5] G.K. Cambron, “The multimedia transformation,” Presented at OFC
‘06, Plenary Presentation, Anaheim, March 2006.
[6] H. Tsushima and Y. Fukahiro, “OTN-based optical cross-connect
systems to create reliable and transparent optical networks,”
Proc. SPIE, vol.6012, 601208-1–10, OpticsEast2005 (ITCom2005),
Boston, Oct. 2005.
[7] S. Okamoto, et al., “Nationwide GMPLS field trial using different
types (MPLS/TDM/Lambda) of switching capable equipment from
multiple vendors,” OFC 2005, Post deadline paper, PDP40, Ana-
heim, March 2005.
[8] L. Smarr, J. Ford, P. Papadopoulos, S. Fainman, T. DeFanti, M.
Brown, and J. Leigh, “The OptIPuter, quartzite, and Starlight
projects: A campus to global-scale testbed for optical technologies
enabling LambdaGrid computing,” OFC 2005, OWG7, Anaheim,
March 2005.
[9] L. Gommans, “Generic AAA based optical networking,” SURFnet
Research on Networks Workshop, Utrecht, April 2004.
[10] M. Brown, ed., “Blueprint for the future of high-performance net-
working,” Commun. ACM, vol.46, no.11, pp.30–77, Nov. 2003.
[11] G. Karmous-Edwards, “Today’s optical network research infrastruc-
tures for Escience applications,” Proc. OFC/NFOFC 2006, OWU3,
Anaheim, March 2006.
[12] K. Harada, K. Shimizu, T. Kudou, and T. Ozeki, “Hierarchical op-
tical path cross-connect systems for large scale WDM networks,”
Proc. OFC 1999, pp.356–358, Feb. 1999.
[13] L. Noirie, C. Blaizot, and E. Dotaro, “Multi-granularity optical
cross-connect,” Proc. Eur. Conf. Optical Communication (ECOC),
pp.269–270, Oct. 2000.
[14] J. Yamawaku, A. Takada, W. Imajuku, and T. Morioka, “Evaluation
of amount of equipment on single-layer optical paths,” J. Lightwave
Technol., vol.23, no.6, pp.1971–1978, June 2005.
[15] Y. Okano, T. Kawata, and T. Miki, “Designing digital paths in trans-
mission networks,” Proc. GLOBECOM 86, 25.2.1–25.2.5, Houston,
1986.
[16] Y. Okano, S. Ohta, and T. Kawata, “Assessment of cross-connect
systems in transmission networks,” Proc. GLOBECOM 87, pp.15–
18, Tokyo, Japan, Nov. 1987.
[17] H. Kobrinski, “Crossconnection of wavelength-division-multiplexed
high-speed channels,” Electron. Lett., vol.23, no.18, pp.974–976,
Aug. 1987.
[18] G. Hill, “Wavelength routing approach to optical communication
networks,” Br. Telecom Technical J., vol.6, no.3, pp.24–30, July
1988.
[19] G. Hill, P.J. Chidgy, and J. Davidson, “Wavelength routing for long
haul networks,” Proc. ICC’89, 23.3.1.–23.3.5, Boston, June 1989.
[20] X. Cao, V. Anand, and C. Qiao, “Waveband switching in optical
networks,” IEEE Commun. Mag., vol.41, no.4, pp.105–112, April
2003.
[21] I. Yagyu, H. Hasegawa, and K. Sato, “An efficient optical path
network design algorithm based on a traffic demand expression
in a Cartesian product space,” Eur. Conf. Optical Communication
(ECOC), Sept. 2006.
[22] Y. Suemura, I. Nishioka, Y. Maeno, S. Araki, R. Izmailov, and S.
Ganguly, “Hierarchical routing in layered ring and mesh optical net-
works,” ICC 2002, vol.5, pp.2727–2733, April-May 2002.
[23] R. Izmailov, S. Ganguly, V. Klepsyn, and A. Varsou, “Non-uniform
waveband hierarchy in hybrid optical networks,” Proc. IEEE INFO-
COM’03, vol.2, pp.1344–1354, 2003.
[24] S. Kakehashi, H. Hasegawa, K. Sato, and O. Moriwaki, “Wave-
band MUX/DEMUX using concatenated arrayed-waveguide grat-
ings,” ECOC 2006, Cannu, Sept. 2006.
[25] S. Kakehashi, H. Hasegawa, K. Sato, and O. Moriwaki, “Formu-
lation of waveguide connection for waveband MUX/DEMUX us-
ing concatenated arrayed-waveguide gratings,” to appear in IEICE
Trans. Commun.
[26] K. Sato, Advances in Transport Network Technologies -Photonic
Networks, ATM, and SDH, Artech House, Norwood, MA, 1996.
(ISBN 0-8906-851-8).
[27] ITU-T Recommendation G.803, Architecture of transport networks
based on the synchronous digital hierarchy (SDH), March 2000.
[28] ITU-T Recommendation G.805, Generic functional architecture of
transport networks, March 2000.
[29] M. Sexton and A. Reid, Broadband Networking: ATM, SDH, and
SONET, Artech House, Norwood, MA, 1997 (ISBN 0-89006-578-
0).
[30] K. Sato, S. Ohta, and I. Tokizawa, “Broadband transport network
architecture based on virtual paths,” IEEE Trans. Commun., vol.38,
no.8, pp.1212–1222, Aug. 1990.
[31] K. Sato and I. Tokizawa, “Flexible asynchronous transfer mode net-
works utilizing virtual paths,” Proc. ICC’90, 318.4.1.–318.4.8., At-
lanta, USA, April 1990.
[32] K. Sato, et al., Packet Multiplexed Communication Systems, Patent
No.2670613 and 2825156.
[33] “Multiprotocol label switching architecture,” IETF RFC 3031, Jan.
2001.
[34] B. Davie and Y. Rekhter, MPLS: Technology and Applications, Mor-
gan Kaufmann, 2000 (ISBN 1-55860-656-4).
[35] E. Mannie, ed., Generalized Multi-Protocol Label Switching (GM-
PLS) Architecture, IETF REC3945, Oct. 2004.
[36] N. Yamanaka, K. Shiomoto, and E. Oki, eds., GMPLS
Technologies—Broadband Backbone Networks and Systems, CRC
Press, 2005 (ISBN: 0-8247-2781-9).
[37] B.W. Stuck, “The digital cross-connect: Cornerstones of future net-
works?,” Data Communications, pp.165–178, Aug. 1987.
[38] A. Gladisch, R.-P. Braun, D. Breuer, A. Ehrhardt, H.-M. Foisel, M.
SATO and HASEGAWA: PROSPECTS AND CHALLENGES OF MULTI-LAYER OPTICAL NETWORKS
1901
Jaeger, R. Leppla, M. Schneiders, S. Vorbeck, W. Weiershausen,
and F.-J. Westphal, “Evolution of terrestrial optical system and core
network architecture,” Proc. IEEE, vol.94, no.5, pp.869–891, May
2006.
[39] K. Sato, “Recent developments in and challenges of photonic net-
working technologies,” IEICE Trans. Commun., vol.E90-B, no.3,
pp.454–467, March 2007.
[40] I. Chlamtac, A. Ganz, and G. Karmi, “Lightpath communications:
An approach to high bandwidth optical WAN’s,” IEEE Trans. Com-
mun., vol.40, no.7, pp.1171–1182, July 1992.
[41] P.-H. Ho, H.T. Mouftah, and J. Wu, “A scalable design of multi-
granularity optical cross-connects for the next- generation optical
Internet,” IEEE J. Sel. Areas Commun., vol.21, no.7, pp.1133–1142,
Sept. 2003.
[42] A. Kolarov, T. Wang, B. Sengupta, and M. Cvijetic, “Impact of
waveband switching on dimensioning multi-granular hybrid optical
networks,” Proc. Conference on Optical Network Design and Mod-
eling 2005, pp.371–381, Feb. 2005.
[43] Y. Shun, C. Ou, and B. Mukherjee, “Design of hybrid optical net-
works with waveband and electrical TDM switching,” Proc. IEEE
Global Telecommunications Conference 2003 (GLOBECOM’03),
vol.5, pp.2803–2808, Dec. 2003.
[44] S.S.W. Lee, M.C. Yuang, and P.-L. Tien, “A Lagrangean relax-
ation approach to routing and wavelength assignment for multi-
granularity optical WDM networks,” Proc. IEEE Global Telecom-
munications Conference 2004 (GLOBECOM’04), vol.3, pp.1936–
1942, Nov. 2004.
[45] P.-H. Ho and H.T. Mouftah, “Routing and wavelength assignment
with multigranularity traffic in optical networks,” IEEE J. Lightwave
Technol., vol.20, no.8, pp.1292–1303, Aug. 2002.
[46] M. Lee, J. Yu, Y. Kim, C. Kang, and J. Park, “Design of hierarchical
crossconnect WDM networks employing a two-stage multiplexing
scheme of waveband and wavelength,” IEEE J. Sel. Areas Commun.,
vol.20, no.1, pp.166–171, Jan. 2002.
[47] X. Cao, V. Anand, Y. Xiong, and C. Qiao, “A study of wave-
band switching with multilayer multigranular optical crosscon-
nects,” IEEE J. Sel. Areas Commun., vol.21, no.7, pp.1081–1094,
Sept. 2003.
[48] I. Yagyu, H. Hasegawa, and K. Sato, “An efficient waveband path as-
signment algorithm for hierarchical optical path network based on a
traffic demand expression in a cartesian product space,” IEICE Tech-
nical Report, PN 2006-12, Aug. 2006.
[49] K. Zhu and B. Mukherjee, “A review of traffic grooming in WDM
optical networks: Architectures and challenges,” Optical Network
Magazine, vol.4, no.2, pp.55–64, March/April 2003.
[50] W. Yao and B. Ramamurthy, “Survivable traffic grooming with
path protection at the connection level in WDM mesh networks,”
IEEE/OSA J. Lightwave Technol., vol.23, no.10, pp.2846–2853,
Oct. 2005.
[51] M. Kodialam and T.V. Lakshman, “Dynamic routing of locally
restorable bandwidth guaranteed tunnels using aggregated link us-
age information,” Proc. IEEE INFOCOM 2001, pp.376–385, April
2001.
[52] H. Zhu, H. Zang, K. Zhu, and B. Mukherjee, “A novel generic graph
model for traffic grooming in heterogeneous WDM mesh networks,”
IEEE/ACM Trans. Netw., vol.11, no.2, pp.285–299, April 2003.
[53] J. Yamawaku, A. Takada, W. Imajuku, and T. Morioka, “Evaluation
of amount of equipment on single-layer optical paths,” J. Lightwave
Technol., vol.23, no.6, pp.1971–1978, June 2005.
[54] L.Y. Lin, E.L. Goldstein, J.M. Simmons, and R.W. Tkach, “Free-
space micromachined optical switches with submillisecond switch-
ing time for large-scale optical crossconnects,” IEEE Photonics
Technol. Lett., vol.10, no.4, pp.525–527, 1998.
[55] V. Kaman, R. Anderson, R. Helkey, O. Jerphagnon, A. Keating, B.
Liu, H. Poulsen, C. Pasarla, D. Xu, S. Yuan, and X. Zheng, “Optical
performance of a 288 ×288 photonic cross-connect system,” Proc.
of PS2002, paper-PS. TuA4, pp.59–61, 2002.
[56] M. Okuno, A. Sugita, T. Matsunaga, M. Kawachi, Y. Ohmori, and K.
Katoh, “8 ×8 optical matrix switch using silica-based planar light-
wave circuit,” IEICE Trans. Electron., vol.E76-C, no.7, pp.1215–
1223, July 1993.
[57] M. Yano, F. Yamagishi, and T. Tsuda, “Optical MEMS for photonic
switching-compact and stable optical crossconnect switches for sim-
ple, fast, and flexible wavelength applications in recent photonic net-
works,” IEEE J. Sel. Top. Quantum Electron., vol.11, no.2, pp.383–
394, March/April, 2005.
[58] D.M. Marom, D.T. Neilson, D.S. Greywall, N.R. Basavanhally, P.R.
Kolodner, Y.L. Low, F. Pardo, C.A. Bolle, S. Chandrasekhar, L.
Buhl, C.R. Giles, S.-H. Oh, C.S. Pai, K. Werder, H.T. Soh, G.R.
Bogard,E.Ferry,F.P.Klemens,K.Teffeau, J.F. Miner, S. Rogers,
H.E. Bower, R.C. Keller, and W. Manfield, “Wavelength-selective
1×4 switch for 128 WDM channels at 50 GHz spacing,” OFC 2002,
Postdeadline Papers, FB7-1–FB7-3, Anaheim, CA, March 2002.
[59] J. Tsai, L. Fan, C. Chi, D. Hah, and M.C. Wu, “A large-count 1 ×
32 wavelength-selective switch using a large scan-angle, high fill-
factor, two-axis analog micromirror array,” ECOC 2004, Tu1.5.2,
Stockholm, Sept. 2004.
[60] S. Okamoto, A. Watanabe, and K. Sato, “A new optical path cross-
connect system architecture utilizing delivery and coupling matrix
switch,” IEICE Trans. Commun., vol.E77-B, no.10, pp.1272–1274,
Oct. 1994.
[61] A. Watanabe, S. Okamoto, and K. Sato, “Optical switch using
WDM,” Patent no.3416895.
[62] A. Watanabe, S. Okamoto, K. Sato, and M. Okuno, “Optical switch,”
Patent no.3444548.
[63] M. Okuno, T. Goh, S. Sohma, and T. Shibata, “Recent advances in
optical switches using silica-based PLC technology,” NTT Technical
Review, vol.1, no.7, pp.20–30, Oct. 2003.
[64] C. Clos, “A study of non-blocking switching networks,” Bell Syst.
Tech. J., vol.32, no.2, pp.406–424, 1953.
[65] S. Kakehashi, H. Hasegawa, and K. Sato, “Optical switch archi-
tectures for hierarchical optical path networks,” OpticsEast 2006,
Boston, MA, Oct. 2006.
[66] G.J. Ockenfuss, “Ultra-low stress coating process: An enabling tech-
nology for extreme performance thin film interference filters,” Proc.
OFC, Postdeadline Paper, FA8-1–FA8-4, 2002.
[67] C.R. Doerr, R. Pafchek, and L.W. Stulz, “Integer band demultiplexer
using waveguide grating routers,” IEEE Photonics Technol. Lett.,
vol.15, no.8, pp.1088–1090, Aug. 2003.
[68] S. Chandrasekhar, C.R. Doerr, and L.L. Buhl, “Flexible waveband
optical networking without guard bands using novel 8-skip-0 band-
ing filters,” IEEE Photonics Technol. Lett., vol.17, no.3, pp.579–
581, March 2005.
[69] C. Dragone, C.A. Edwards, and R.C. Kistler, “Integrated optics
N×Nmultiplexer on silicon,” IEEE Photonics Technol. Lett., vol.3,
no.10, pp.896–899, Oct. 1991.
[70] K. Okamoto, T. Hasegawa, O. Ishida, A. Himeno, and Y. Ohmori,
“32 ×32 arrayed-waveguide grating multiplexer with uniform
loss and cyclic frequency characteristics,” Electron. Lett., vol.33,
pp.1865–1866, 1997.
[71] S. Watanabe, S. Takeda, and T. Chikama, “Interband wavelength
conversion of 320 Gbit/s(32×10 Gbit/s) WDM signal using a
polarization-insensitive fiber four-wave mixer,” ECOC’98, Post-
deadline paper, pp.85–86, 1998.
[72] M.H. Chou, I. Brener, M.M. Fejer, E.E. Chaban, and S.B.
Christman, “1.5-µ-band wavelength conversion based on cascaded
second-order nonlinearity in LiNbO waveguides,” IEEE Photonics
Technol. Lett., vol.11, no.6, pp.653–655, June 1999.
[73] O. Tadanaga, M. Asobe, H. Miyazawa, Y. Nishida, and H. Suzuki,
“A 1-THz optical frequency shifter using quasi-phase-matched—
LiNbO3wavelength converters,” Proc. OFC, ThDD1, pp.592–593,
March 2002.
[74] J. Yamawaku, H. Takara, T. Ohara, K. Sato, A. Takada, T. Morioka,
O. Tadanaga, H. Miyazawa, and M. Asobe, “Inter-band wavelength
1902
IEICE TRANS. COMMUN., VOL.E90–B, NO.8 AUGUST 2007
conversion of 25 GHz-spaced 1.03 Tbit/s (103 ×10 Gbit/s) DWDM
signals with small guard band and low crosstalk in PPLN wave-
guide,” Proc. CLEO 2003, paper CThPDB2, Baltimore, June 2003.
[75] J. Yamawaku, E. Yamazaki, A. Takada, T. Morioka, and K. Suzuki,
“Virtual grouped-wavelength-path switching based on QPM-LN
waveband converter and supercontinuum wavelength-bank source,”
Proc. OFC2005, paper OFE2, Anaheim, March 2005.
[76] R. Osellame, R. Ramponi, M. Marangoni, G. Tartarini, and P. Bassi,
“Integrated all-optical nonlinear device for reconfigurable add/drop
and wavelength shifting of WDM signals,” Appl. Phys. B, vol.73,
no.5–6, p.505, 2001.
[77] J. Yamawaku, A. Takada, E. Yamazaki, O. Tadanaga, H. Miyazawa,
and M. Asobe, “Selective wavelength conversion using PPLN wave-
guide with two pump configuration,” CLEO 2003, CWB5, Balti-
more, 2003.
[78] E. Yamazaki, A. Takada, J. Yamawaku, and T. Morioka, “Simul-
taneous and arbitrary wavelength conversion of WDM signals us-
ing multiple wavelength quasi phase matched LiNbO3waveguide,”
Proc. OFC 2004, paper FL6, Los Angels, March 2004.
[79] M.H. Chou, K.R. Parameswaran, M.M. Fejer, and I. Brener,
“Multiple-channel wavelength conversion by use of engineered
quasi-phase-matching structures in LiNbO3waveguides,” Opt. Lett.,
vol.24, no.16, pp.1157–1159, Aug. 1999.
[80] M. Asobe. O. Tadanaga, H. Miyazawa, Y. Nishida, and H. Suzuki,
“Highly flexible and robust multiple quasi-phase matched LiNbO3
wavelength converter,” CLEO 2003, paper CWB3, pp.1129–1132,
Baltimore, 2003.
Ken-ichi Sato received the B.S., M.S., and
Ph.D. degrees in Electronics Engineering from
the University of Tokyo, in 1976, 1978, and
1986, respectively. He is currently a professor
at the Graduate School of Engineering, Nagoya
University, and he is an NTT R&D Fellow. Be-
fore joining the university in April 2004, he was
an executive manager of the Photonic Transport
Network Laboratory at NTT. His R&D activi-
ties cover future transport network architectures,
network design, OA&M (operation administra-
tion and maintenance) systems, photonic network systems including optical
cross-connect/ADM and photonic IP routers, and optical transmission tech-
nologies. He has authored/co-authored more than 200 research publications
in international journals and conferences. He holds 35 granted patents and
more than 100 pending patents. He received the Young Engineer Award
in 1984, the Excellent Paper Award in 1991, and the Achievement Award
in 2000 from the Institute of Electronics, Information and Communication
Engineers (IEICE) of JAPAN. He was also the recipient of the distinguished
achievement Award of the Ministry of Education, Science and Culture in
2002. His contributions to ATM (Asynchronous Transfer Mode) and opti-
cal network technology development extend to co-editing three IEEE JSAC
special issues and the IEEE JLT special issue once, organizing several
Workshops and Conference technical sessions, serving on numerous com-
mittees of international conferences including OFC and ECOC, authoring a
book, Advances in Transport Network Technologies (Artech House, 1996),
and co-authoring thirteen other books. He is a Fellow of the IEEE.
Hiroshi Hasegawa received the B.E., M.E.,
and D.E. degrees all in Electrical and Electronic
Engineering from Tokyo Institute of Technol-
ogy, Tokyo, Japan, in 1995, 1997, and 2000,
respectively. From 2000 to 2005, he was a re-
search associate of Dept. of Communications
and Integrated Systems, Tokyo Institute of Tech-
nology. Currently he is an associate professor of
Nagoya University. His current research inter-
ests include Photonic Networks, Image Process-
ing (especially Superresolution), Multidimen-
sional Digital Signal Processing and Time-Frequency Analysis. He re-
ceived the Young Researcher’s Awards from SITA (Society of Informa-
tion Theory and its Applications) and IEICE (Institute of Electronics, In-
formation and Communication Engineers) in 2003 and 2005, respectively.
Dr. Hasegawa is a member of SITA and IEEE.