Content uploaded by Michal Aibin
Author content
All content in this area was uploaded by Michal Aibin on Jul 23, 2020
Content may be subject to copyright.
Energy Consumption Reduction of the Survivable
Spectrally-Spatially Flexible Optical Networks
Michal Aibin, Membe r, IEEE, Joseph Gotengco, Justin Tran,
Jason Soukchamroeun, Connor Vinchoff, and Nathan Chung
British Columbia Institute of Technology, 555 Seymour Street, Vancouver V6B 3H6, Canada
Tel: +1 604.412.7494, e-mail: maibin@bcit.ca
ABSTRACT
Energy reduction is rapidly gaining interest due to several environmental changes in our society. In this paper,
we design an energy-efficient algorithm for the problem of Routing, Modulation, Core and Spectrum
Assignment in Spectrally-Spatially Flexible Optical Networks (SS-FON) with the network survivability.
We then evaluate it using existing, real-life networks in the CEONS simulator. Two primary metrics of the
evaluation are operational expenditure (OPEX) and Bandwidth Blocking Probability (BBP). The numerical
experiments confirm the efficiency of our solution compared to the ones existing in the literature.
Keywords: optical networks, optimization, energy, SDM, EON, FON, routing.
1. INTRODUCTION
Energy consumption research for optical networks gained interest in recent years. There is a growing consensus
on the need to reduce energy usage, thus, reduce the carbon footprint. It became one of the most important and
crucial policy concerns at the top of the research agenda. It is estimated that ICT technology contributes to about
2% of the worldwide global carbon footprint today. It is also expected that it will grow up to 3% by 2020 [1].
In addition, if no improvements will be achieved in the energy efficiency of the technology utilized
in communications network devices, the total power consumption of fixed communication networks is foreseen
to grow to 1.5 TW (i.e., 13140 TWh/year) in 2025 [2]. Just looking recently, the energy price increase, the
continuous growth of the customer population, the spreading of broadband access, and the expanding quantity of
services being given by Internet Service Providers (ISPs), brought the energy efficiency issue a high-priority
objective for optical and service infrastructure [1].
In this paper, we are mainly focusing on the energy consumption of optical networks while enabling network
survivability via path protection. Optical networks are widely used and currently, they constitute the necessary
physical network infrastructure in most parts of the world, thanks to their high speed, large capacity, and other
attractive properties. They are also a secure way of data transport [3]. We are going to use the Spectrally-
Spatially Flexible Optical Networks (SS-FONs), as they are expected to replace the current optical technology
[4]. Although fibre optic technology has grown in popularity over the past few years it is still considered
expensive to implement. The electricity consumed by the world's networks consume a large portion of the global
electricity supply and is expected to consume more as the population and Internet access grows. This problem
stems from the energy consumption used to maintain and operate networks. That is why it is ultimately vital to
optimize operations related to dynamic routing, both, in terms of energy and routing efficiency.
In this paper, we design an energy savings algorithm, named Energy Reduction Routing (ERG), and evaluate it
using SS-FONs to optimize the operators' network operational cost (OPEX). Network cost is defined as $ value
per 1 TB of data transmitted. The goal is to minimize it. The pricing models that we used are based on the
average North American [5] and European [6] average electricity prices. The rest of the paper is divided as
follows. In Section 2 we introduce the problem, and it is followed by the ERG algorithm description
in Section 3. We then discuss the simulation setup in Section 4. Finally, Section 5 contains results, and it is
followed by the conclusion.
2. PROBLEM DEFINITION
The world's population is roughly 8 billion people and is expected to grow in the future. As a result of more
advanced technology, increased data transfers and higher mobility, network energy consumption will likely
increase as well. The environmental impact of the by-products produced by the energy consumption of the
Internet, such as greenhouse gases, affects the future population and us.
In response to this problem, our study designs algorithms to tackle the problem of energy consumption with
respect to BBP through the CEONS simulator [7]. The energy model for SS-FON we will be using is similar to
the work done in [8]. The following table is used for the calculation of energy consumption, using various
modulation formats in SS-FONs (see Table I).
Once we know the energy consumption, we want to introduce the following ILP optimization model, with a
goal to minimize the networks’ energy consumption, defined in the Equation (1).
Sets
G(V,E,B,C,L) graph
V nodes
E links
K(e) set of cores on link e
B slices
D requests
C(d, p) candidate channels
P(d) candidate paths
M modulations
J(m) power consumed by the
modulation m per single
subcarrier
Constants
ωedp = 1, if link e belongs to the path p
realizing
the request d; 0, otherwise
ndp the number of required slices to serve the
request d on the path p
gdpcbk 1, if channel c associated with the request
d
on the path p uses slice b and core k;
0, otherwise
Variables
xdpc = 1, if a channel c on a candidate path p is
used to realize the request d; 0, otherwise
yebk = 1, if a slice b is occupied using core k on
a link e; 0, otherwise
yeb = 1, if a slice b is occupied on a network
link e; 0, otherwise
yb = 1, if a slice b is occupied on any network
link; 0, otherwise
ϸ(m) bit-rate per single subcarrier of modulation m
ϐ(d) volume of a request d
Objective function
(1)
Subject to
The optimization is performed subject to four constraints:
● for each request d precisely one candidate path p(d) and one candidate channel c(d, p) are selected (2)
● slice b on a link e can be allocated to one lightpath on a core k (3)
● slice b is used only when it is allocated on at least one link e (4)
● slice b is used only if at least one core k on link e uses it (5)
3. SURVIVABLE ENERGY REDUCTION ROUTING ALGORITHM (ERG)
In order to design an energy reduction based algorithm and not decrease the performance of routing, we
introduced a metric named Balance Energy and Resource Utilization (BERU). In short, each modulation format
is ranked based on the current effect on resource utilization and energy usage per particular request to be
allocated. Details of metrics are presented in Table II. We introduce three ranges for network resources and
energy consumption. When resources are utilized in less than 30% (links and regenerators), then we talk about
low resource utilization; medium is when the resource utilization is between 31% – 65% and high if it is more
than 65%. For the second metric, we calculate the historical average network energy consumption after the
network is warmed-up (we skip the first 10k requests). Since requests arrive in batches (more details in the next
section), we can estimate the effect of particular modulation choice on energy usage. If energy consumption will
increase more than 10%, we mark this scenario as high energy consumption for the next batch of requests;
if energy will be reduced by more than 10%, then we mark the next batch of requests as low energy
consumption. Finally, if the change is less than 10% (both directions), then we keep the medium energy
consumption scenario. Please note that medium energy consumption is also a default scenario for the first batch
of requests. Since we enable path protection, the calculations are made for both, the primary and backup paths.
4. SIMULATION SETUP
We assume that the considered network uses a flexible grid of 12.5 GHz frequency slot granularity and is
equipped with bit-rate variable transponders (BVTs). We consider a European network topology with 28 nodes,
82 links, and an average link length of 625 km, as well as a United Stated network with 26 nodes, 84 links and
an average link length of 755 km. Each link comprises 7 spatial modes, and each spatial mode provides 4 THz of
bandwidth. Transponders are capable of carrying different bit-rates by modifying the channel bandwidth using
a variable number of subcarriers (SCs), and modulation changes along the lightpath. We assume that six
modulation formats are available, namely: BPSK, QPSK, 8-QAM, 16-QAM, 32-QAM, and 64-QAM. We also
assume that the transponder is sliceable in subcarriers with a limit of 6 SCs per BVT. Requests arrive in batches
of 20, following the Poisson process with arrival rate ƛ and a negative exponential distributed holding time with
1/µ. Therefore, the traffic load could be defined as ƛ/µ Erlangs. For each value of traffic load, the first 10000
requests were not considered, due to the network not being in steady-state. After that, we evaluated 250k
requests.
Focusing on the optical network survivability, the following methods are used. The first technique that we
incorporate is Dedicated Path Protection (DPP). It implies that each demand is served by a working (primary)
path and a backup path, which is link-disjoint with the working path. The DDP method requires a large amount
of extra capacity for protection purposes, keeping protection resources idle when there is no failure. The second
method of path protection that we use is the Shared Backup Path Protection (SBPP). It allows different backup
paths to share spectrum resources on the overlapping portion if the corresponding working paths are link-
disjoint. SBPP utilizes capacity more efficiently than DPP, but in some cases may not provide 100% protection.
For e.g., when there are multiple-link failures in the network, which concurrently affect several demands that
share the same resources on the backup paths [9]. We compare our ERG algorithm with three other approaches,
modified to work with survivable SS-FONs: SPF [10], CP [11] and AMRA [12].
5. RESULTS
First, we analyze the BBP for various approaches based on the simulations performed using the Euro28 network
(Fig. 1a) and the US26 network (Fig. 1b).
(a)
(b)
Figure 1. BBP for various approaches based on the simulations performed using:
(a) Euro28 network, (b) US26 network.
We set our SLA level to a maximum of 1% (industry standard). Let us now discuss the performance of
algorithms for the Euro28 network. As we can observe the worst algorithm is SPF, in both DPP and SBPP
protection scenarios – we then use it as a baseline reference method. The second worst is CP. Even its
performance is not ideal, it is still much better than SPF and allows acceptable levels of BBP up to the level of
400 Erlangs (ER) and 500 ER, for DPP and SBPP methods, respectively. The best two approaches are AMRA
and ERG, with a slight preference towards the AMRA approach. ERG allows us to keep the acceptable levels of
SLA up to the traffic loads of 550 ER, while AMRA allows around 50 ER more. The more significant benefits
are visible in the energy savings, and we will discuss that later. The trends for the US26 network are similar – the
best two approaches are AMRA and ERG. The main difference is that their performance deteriorates faster and
allows them to maintain the SLA for low and moderate traffic loads.
Since AMRA and ERG perform almost similarly, let us compare them with respect to their energy efficiency in
Table III. Since energy-efficiency is also related to OPEX, we do gather it in relation to the networking cost.
As we can observe, the benefits of using ERG are significant. We were able to reduce the energy costs by 60%
while maintaining almost similar SLA for BBP. The trends are similar in both networks, but due to the lack of
space, we present only a comparison of Euro28 network. To summarize what we discovered in two networks, the
costs in Europe are higher, also due to the higher electricity prices, but they are maintaining stable trends. On the
other hand, we can achieve much lower costs in North American networks, but those raise rapidly and for high
traffic loads are higher than in Euro28.
6. CONCLUSION
In this paper, we proposed the energy efficient algorithm to solve the problem of resource allocation in
Spectrally- Spatially Flexible Optical Networks. We compared its performance with various state-of-art methods
under two realistic network scenarios. For the future, we plan to investigate the possibility of adding the traffic
prediction feature into our newly designed algorithm.
ACKNOWLEDGEMENTS
This work was supported by statutory funds of the British Columbia Institute of Technology.
REFERENCES
[1] R. Bolla, R. Bruschi, F. Davoli, and F. Cucchietti, "Energy efficiency in the future Internet: A survey of
existing approaches and trends in energy-aware fixed network infrastructures," IEEE Communications
Surveys and Tutorials, vol. 13, no. 2, pp. 223-244, Second Quarter 2011.
[2] R. S. Tucker, "Green optical communications—Part II: Energy limitations in networks," IEEE Journal of
Selected Topics in Quantum Electronics, vol. 17, no. 2, pp. 261-274, Mar.-Apr. 2011.
[3] M. Furdek et al., "An overview of security challenges in communication networks," in RNDM, Halmstad,
2016.
[4] M. Klinkowski, P. Lechowicz, and K. Walkowiak, "Survey of resource allocation schemes and algorithms
in spectrally-spatially flexible optical networking," Optical Switching and Networking 27 (2018): 58-78.
[5] Short, Thomas Allen. Electric power distribution equipment and systems. CRC Press, 2018.
[6] G. Gissey et al., “Wholesale cost reflectivity of GB and European electricity prices,” Imperial College
London 2018.
[7] M. Aibin and M. Blazejewski, "Complex elastic optical network simulator (CEONS)," in 2015 17th
International Conference on Transparent Optical Networks (ICTON), Budapest, 2015, pp. 1-4.
[8] K. Walkowiak and M. Klinkowski, "Energy efficiency of dynamic routing in elastic optical networks with
back-to-back regeneration," in 2018 ICNC, Maui, HI, 2018, pp. 399-403.
[9] J. Rak, "Resilient Routing in Communication Networks, " Berlin: Springer, 2015.
[10] K. Ahuja et al., "Faster algorithms for the shortest path problem," Journal of the ACM 37.2 (1990).
[11] K. Walkowiak and M. Klinkowski, "Joint anycast and unicast routing for elastic optical networks:
Modeling and optimization," in 2013 IEEE ICC, Budapest, 2013, pp. 3909-3914.
[12] M. Aibin and K. Walkowiak, "Adaptive modulation and regenerator-aware dynamic routing algorithm in
elastic optical networks," in 2015 IEEE ICC, London, 2015, pp. 5138-5143.
