LIGHTPATH SURVIVABILITY WITH QOT GUARANTEES
Key Issues in Wavelength-Routed Photonic Networks
Luiz Gustavo Zuliani, Stefano Giordano, Michele Pagano
Dept. of Information Engineering, University of Pisa, Pisa, Italy
Davide Adami
CNIT Research Unit - Dept. of Information Engineering, University of Pisa, Pisa, Italy
Keywords: Photonic networks, Survivability, RWA, Physical impairments, QoT, Residual link capacity, Power
constraint.
Abstract: The Routing and Wavelength Assignment (RWA) problem has been extensively studied over the last
decade, but only in recent years algorithms were specifically developed (or adapted) to consider
transmission layer physical impairments in photonic (all-optical) networks, while providing lightpath
survivability. This work discuss the last contributions in the field, and presents a new impairment-aware
RWA algorithm for survivable photonic networks. As a work in progress, the new algorithm is presented
with open design decisions.
1 INTRODUCTION
Nowadays, the Internet backbone traffic is mainly
carried by Wavelength Routed Networks (WRN)
based on WDM lines due to the high capacity, low
Bit Error Rate (BER) and low per-bit monetary
transmission cost of these systems. In order to setup
a lightpath (LP), i.e., an optical circuit between two
nodes, a route must be calculated and available
wavelengths must be allocated in all links that
compose the path. This problem is NP-complete, and
is known as Route and Wavelength Assignment
(RWA). It has been extensively studied in the last
decade considering different aspects, mainly
minimization of resource allocation, online
operation and LP survivability. For online operation,
usually the RWA problem is divided in two sub-
problems that are solved separately: first, a shortest
path algorithm is used to compute the route, then an
heuristic is used for wavelength allocation.
With the advance of DWDM systems (carrying
hundreds of channels per fiber, each one operating at
rates up to 40 Gb/s), the electronic processing at the
nodes has became a bottleneck. To overcome this
problem, all-optical (or photonic) crossconnects
(PXCs) that do not realize optical-electrical-optical
(OEO) conversion were introduced, and meshed
WRNs based on PXCs without wavelength
conversion capabilities (all-optical wavelength
conversion is still an immature technology) are
slowly being deployed. As a side effect of OEO
absence, signals are not regenerated at each hop
anymore, and accumulate transmission impairments
that affect the QoT and consequently the BER at the
end node. Impairments-aware RWA (IA-RWA)
algorithms have being envisaged, using different
network impairment models, resulting in different
performance (Azodolmolky, et al., 2008). However,
specifically regarding network survivability, the IA-
RWA problem is even more complex, and a reduced
number of works is available (Zhai, et al., 2007).
In the first part of this work, the most recent IA-
RWA solutions for Survivable Photonic WRNs are
reviewed. Aspects such as key issues, design
decisions and performance evaluation metrics are
discussed. The second part presents the IA-RWA for
survivable networks being developed by the authors.
Comparisons between the algorithm under study and
the previously presented works are also depicted. At
the end, conclusions are shown.
165
Zuliani L., Giordano S., Pagano M. and Adami D. (2010).
LIGHTPATH SURVIVABILITY WITH QOT GUARANTEES - Key Issues in Wavelength-Routed Photonic Networks.
In Proceedings of the International Conference on Data Communication Networking and Optical Communication Systems, pages 165-168
DOI: 10.5220/0003033201650168
Copyright
c
SciTePress
2 SURVIVABLE IA-RWA
Generally speaking, the main goal of all survivable
IA-RWA algorithms is to provide LP resilience.
Despite that fact, they can be designed in very
different ways, depending on the constraints
considered for the WRN itself but also for the LPs.
Survivable IA-algorithms can be classified in
function of the network impairment model used, the
type of the combined IA-RWA process, the type of
resilience and the quality levels offered (evaluated
using ad-hoc performance metrics). In the following
subsections the most recent IA-RWA solutions
(Zhai, et al., 2007; Askarian, et al., 2008; Kim, et al.,
2008; Markidis & Tzanakaki, 2008; Jirattigalachote,
et al., 2009) are reviewed under the perspective of
each of these characteristics.
2.1 Network Impairment Models
Transmission in optical fibers is affected by a
number of physical impairments. The most relevant
are intersymbol interference (ISI), amplified
spontaneous emission (ASE), polarization mode
dispersion (PMD) and node and interchannel
crosstalking (Zhai, et al., 2007). The predominant
impairment depends on many factors, like the
quality of fibers and node components, the LP
optical signal power and bandwidth, and the
wavelength spacing between channels.
All of the cited works consider ISI, ASE and
both crosstalking forms as noise-like terms, and the
sum of their variances is accounted for the Q factor
calculation, which is a signal-to-noise ratio. The LP
BER is estimated in function of the Q factor with a
simple equation.
PMD was ignored in all works, because it is
relevant only at data rates of 40 Gb/s and beyond.
2.2 RWA Combined Process
As stated by Azodolmolky, et al. (2008), the routing,
wavelength assignment and QoT evaluation
processes can be combined in many ways, with
different levels of complexity and performance. The
best (and most complex) IA-RWA algorithms
consider the physical impairments during the RWA
phase, and also estimate the BER of the candidate
LP.
Three of the reference works divide the IA-RWA
problem in two sub-problems. To calculate the work
and backup paths, it was used fixed-alternate routing
with Yen's algorithm (offline) and Dijkstra
algorithm (online). Non IA-routing used link length
as link cost metric, and IA-routing used the Q-
penalty metric (Markidis & Tzanakaki, 2008), that is
also calculated as noise-like terms. The wavelength
Assignment was realized using the following
algorithms: First Fit (FF), Last Fit (LF), Best Fit
(BF), Random Pick (RP) and Most Used (MU). It is
important to note that these heuristics present
different behavior in ideal networks and physical
impaired networks (He, et al., 2009). Zhai, et al.
(2007) and Markidis & Tzanakaki (2008) presented
single-phase RWA process, where the shortest path
for each wavelength plane is calculated.
All proposals evaluate the BER of candidate
LPs. If the BER is under a predefined value (usually
Q factor equal to 6 or 7), the request is blocked.
2.3 Protection and Restoration
LP resilience can be pre-configured or just pre-
planned. In both cases the backup LP is already
computed, but only in the former case the resources
are already allocated to the backup LP. If the backup
LP carries the same traffic as the working LP even
before failure, this kind of resilience is called 1+1
dedicated protection. If the backup LP is used for
Best Effort traffic or not used at all, it is called 1:1
dedicated protection. Protection is very efficient
(service disruption is inferior to 50 ms), but is also
the most expensive kind of resilience.
Pre-planned resilience is also called restoration,
and can be dedicated or shared. In both cases the
wavelength remains unused in the fiber links until
the restoration mechanisms are activated. Therefore,
the fiber remains “dark”, at least for that particular
channel. In the case of shared restoration, a
wavelength reserved for shared backup remains free
to be used in other shared backup path computations,
i.e., it can (and possibly will) be used to protect
more than one LP. Restoration is better for the
overall network QoT, because the backup LPs
remain dark and do not interfere with the QoT of the
working LPs. Also, shared restoration improves the
network resources utilization. On the other hand,
when a LP must be restored through a pre-planned
computation, there is no guarantee that a) it will
satisfy the required BER and b) it will not
compromise the QoT of other established LPs. That
situation is even worse in the case of shared
restoration. That happens because when a new LP
must be setup, the IA-RWA engine does not take
into account the physical impairments of dark
wavelengths used to restore LPs.
DCNET 2010 - International Conference on Data Communication Networking
166
All kind of resilience described in this section
are investigated in the referred works, with
interesting results.
2.4 Performance Evaluation Metrics
Usually, IA-RWA algorithms are evaluated through
simulations using either real-world topologies (like
the classic 14 nodes NSF topology) or mesh toroid
networks due to the high degree of symmetry and
connectivity. Results are in the form of the blocking
probability in function of the traffic load. Other
interesting metrics found in the cited works are:
Vulnerability Ratio or QoT-Vulnerability: the
probability that, in the case of a link failure, a
pre-planned backup LP cannot be restored due to
unacceptable QoT;
Cascading Failure Vulnerability: the probability
that a given LP become unusable due to physical
impairments induced by the activation of pre-
planned backup LPs;
Failure Ratio: it is defined as the ratio between
the number of connections that are not recovered
due to unacceptable QoT to the number of
working LPs affected by a link failure. It is
averaged over all single link failures;
Running time: The time needed to compute the
LP from the instance of request arrival.
Interesting values are the average and worst case
scenario.
2.5 LP Application
LPs are used by the clients of the optical layer to
transport two types of traffic, basically: mission
critical and non-critical applications traffic. The
former requires strict QoT guarantees and protection
mechanisms (downtime can be as low as 5 min per
year), usually remains active for very long periods
but do not require a short setup time. On the other
hand, non-critical applications require a short setup
time, but have loose requirements regarding QoT
and resilience. An IA-RWA must be designed to
satisfy one or another type of LPs, due to the
tradeoffs involved. Most of the IA-RWA algorithms
presented in the referred works have different
versions with protection and restoration.
3 PROPOSED ALGORITHM
The objective of the proposed algorithm is to
compute LPs with dedicated 1+1 or 1:1 protection,
with assured QoT and survivability. The main goal
is to satisfy these conditions while minimizing the
resource allocation.
Different from the referred works, the network
impairment model used in the proposed IA-RWA is
not based on the sum of noise-like terms (one for
each dominant impairment), but instead is based on
the minimum and maximum power constraints. The
minimum power constraint, which is best known as
sensitivity level, assures that optical signals can be
properly detected by all optical devices. Thus, the
survivable IA-RWA algorithm not only gives a path
and a wavelength for each of the LPs (work and
backup), but also the optical power that must be
injected at the ingress PXC in order to guarantee the
requested BER at the egress PXC. The maximum
power constraint imposes a limit to the optical
power on fiber links. This way, the fiber
nonlinearities (that are completely power-dependent,
such as channel crosstalking) can be indirectly
managed. The network impairment model and the
impairment validation process of the proposed
algorithm are based on the ASE noise and on the
desired Q factor, and use the analytical model
discussed by Pavani, et al. (2008).
The first part of the algorithm solves the RWA
problem in a combined way. For each wavelength
available to use at the source node (i.e., the
wavelengths that the available transponders at
ingress PXC can tune), a couple a disjointed shortest
paths is calculated using the Suurballe algorithm.
Physical impairments are considered during the
RWA phase as the link cost metric. The cost of a
given fiber link is a function of the residual
wavelength and optical power capacity. Thus, the
lower is the number of used wavelengths and the
total optical power that traverses a link, the lower
will be its cost. Another strategy introduced by the
proposed algorithm in order to minimize the
blocking probability is the Critical Link Avoidance
(CLA). The “altruist” idea of avoiding using
particular links to save them for future requests was
introduced by the Asynchronous Criticality
Avoidance (ACA) protocol. Except for sharing this
concept, CLA technique is completely different from
ACA by any perspective. All links that are labeled
as critical are initially pruned from the physical
topology. If, after the first attempt, no path is found
in any of the wavelength planes, the process is
repeated again, but this time considering all critical
links. To define a link as critical, upper bounds for
the wavelength use and total optical power are being
considered.
LIGHTPATH SURVIVABILITY WITH QOT GUARANTEES - Key Issues in Wavelength-Routed Photonic Networks
167
After the IA-RWA phase, the BER is verified in
all work and backup LP candidates. All work-
backup LP pairs which the BER is too high or whose
links can not accommodate their share of optical
power (due to the maximum power constraint) are
removed from the list of candidates. It is still to be
decided if a work-backup LP pair is to be removed if
only the backup LP do not met with the BER
requirements. After this step, all candidate LPs are
known to satisfy the BER requirements. At this
point, the final LP pair selection must take place. To
effectively choose the best candidates, a number of
simple heuristics are being considered. Simulations
must be performed to determine which ones are
relevant, and in which order. When more than one
candidate matches the criterion, the matched ones
are compared on the basis of the next criterion.
When only a pair of LP candidates are left (not
necessarily with the same wavelength), these are the
work and backup LP to be established. The
heuristics to be considered (so far) are:
lowest number of critical links;
LP pair that uses the same wavelength, if the cost
difference between the work and backup LPs (in
terms of number of hops and/or transmission
power) is bellow a given upper bound;
lowest transmission power;
lowest number of hops;
LP whose wavelength is the FF, MU or other
wavelength assignment heuristics.
4 CONCLUSIONS
The RWA problem is a well studied area, but only in
the last years works were conducted considering
fully transparent (photonic) WRNs, physical
impairments and LP survivability. In this work,
some of the last contributions in the survivable IA-
RWA field were reviewed. It was discussed the
network impairment model used, the mechanisms
for routing, wavelength assignment and physical
impairments validation, and the proposed
performance evaluation metrics. It was also
presented a new survivable IA-RWA algorithm to
compute LPs with dedicated 1+1 or 1:1 protection,
with assured QoT and survivability. The proposed
algorithm uses a different impairment model based
on sensitivity and maximum power constraints, and
uses the wavelength load and total optical power as
metrics to define the link costs and the criticality
levels. The introduced CLA saves critical links in
order to minimize the blocking probability.
Simulations are needed to be carried out to verify
if the added complexity minimizes the blocking
probability of setup requests. The performance
evaluation metrics and topologies introduced by the
referred works must be considered.
REFERENCES
Azodolmolky, S., et al., 2008. A survey on physical layer
impairments aware routing and wavelength
assignment algorithms in optical networks. Computer
Networks, 53, pp. 926-944.
Zhai, Y., et al., 2007. Performance of dedicated path
protection in transmission-impaired DWDM networks.
In: IEEE ICC 2007.
Askarian, A., et al., 2008. Protection and restoration from
link failures in DWDM networks: a cross-layer study.
In: IEEE ICC 2008.
Kim, S., et al., 2008. QoT-guaranteed protection:
survivability under physical layer impairments. In:
ICST BROADNETS 2008.
Markidis, G., Tzanakaki, A., 2008. Routing and
wavelength assignment algorithms in survivable
WDM networks under physical layer constraints. In:
ICST BROADNETS 2008.
Jirattigalachote, A., et al., 2009. Impairment Constraint
Based Routing (ICBR) with service differentiation in
survivable WDM networks. In: ECOC 2009.
He, J., et al., 2009. QoS-aware wavelength assignment
with BER and latency constraints for all-optical
networks. Journal of Lightwave Technology, 27 (5),
pp. 462-474.
Pavani, G. S., et al., 2008. Distributed approaches for
impairment-aware routing and wavelength assignment
algorithms in GMPLS networks. Computer Networks,
52, pp. 1905-1915.
DCNET 2010 - International Conference on Data Communication Networking
168