A Learning Automata-based Algorithm for Energy-efﬁcient Elastic

Optical Networks

Georgia A. Beletsioti

1 a

, Georgios I. Papadimitriou

1 b

, Petros Nicopolitidis

1 c

Emmanouel Varvarigos

2 d

and Stathis Mavridopoulos

1 e

Department of Informatics, Aristotle University of Thessaloniki, Thessaloniki, GR-54124, Greece

School of Electrical and Computer Engineering, National Technical University of Athens, Athens, GR-15780, Greece

Keywords:

Adaptivity, Elastic Optical Network, Energy-efﬁciency, Learning Automata.

Abstract:

Efﬁcient use of available bandwidth plays an important role in performance enhancement due to the wide

penetration of high-bandwidth demanding services. The ﬂexible nature of elastic optical networks (EONs)

effectively uses spectral resources for communication by allocating the minimum required bandwidth to cus-

tomer requirements. Since the energy consumption of such networks scales with the magnitude of bandwidth

demand, many studies have addressed the issue of energy wastage in optical networks. Learning Automata

are Artiﬁcial Intelligence tools that have been used in networking algorithms where adaptivity to the charac-

teristics of the network environment can result in a signiﬁcant increase in network performance. This work

introduces a new adaptive power-aware algorithm, which selectively switches off bandwidth variable optical

transponders (BVTs) under low utilization scenarios supporting energy efﬁciency. A novel algorithm which

uses LA technology and signiﬁcantly reduces the total energy consumption, while maintaining low bandwidth

blocking probability (BBP), is proposed. LA mechanism applied in this work, aims to ﬁnd the best number of

BVTs to be switched off so as for the BBP not to be affected. Simulation results are presented, which indicate

that the proposed algorithm achieves a power saving of up to 50%, compared to non-adaptive solutions.

1 INTRODUCTION

The demand for bandwidth grows exponentially ev-

ery year, driven by a growing number of global in-

ternet users. Furthermore, high capacity-demanding

technologies, including autonomous vehicles, the in-

ternet of things, high bandwidth enhanced video, and

virtual reality, will also drive future needs. According

to Cisco, global IP trafﬁc stood at 122 Exabytes in

2017 and it is estimated that these numbers will triple

by 2022 (Cisco, 2019).

Elastic optical networks (EON), as a novel con-

cept of WDM networks, are considered the most

suitable architecture for backbone and next gen-

eration metropolitan networks as they are charac-

terized by high spectral efﬁciency and adaptability

(Jinno, 2017). EONs which are based on orthog-

https://orcid.org/0000-0002-1895-094X

https://orcid.org/0000-0001-9529-9380

https://orcid.org/0000-0002-5059-3145

https://orcid.org/0000-0002-4942-1362

https://orcid.org/0000-0002-7058-3147

onal frequency-division multiplexing (OFDM) (Dao

et al., 2018) support lightpaths with different bitrates,

exploit the ﬂexible grid technology where the spec-

trum is split into 25, 12.5 GHz or less slots compared

to coarser splitting of 50 GHz or 100 GHz of tradi-

tional WDM networks. Hence, the slots are com-

bined to create channels, which are not overlapping

due to OFDM’s orthogonality capacity, of the desired

size using bandwidth what is strictly necessary for the

transmission spectrum (Soumplis, 2017).

The energy consumed by ICT (Information and

Communication Technology) equipment, which is

rapidly expanding (Belkhir and Elmeligi, 2018),

(Beletsioti et al., 2016), causes a signiﬁcant economic

and environmental problem. According to European

Framework Initiative for Energy and Environmental

Efﬁciency in the ICT Sector, ICTs account for 8-10%

of the European electricity consumption and up to 4%

of its carbon emissions. Furthermore, the network in-

frastructure is becoming a large portion of the energy

footprint in ICT. Thus, the concept of energy efﬁcient

or green networking has been emerged as a research

topic. The issue of energy saving in IP Over WDM

Beletsioti, G., Papadimitriou, G., Nicopolitidis, P., Varvarigos, E. and Mavridopoulos, S.

A Learning Automata-based Algorithm for Energy-efﬁcient Elastic Optical Networks.

DOI: 10.5220/0009819400270034

In Proceedings of the 17th International Joint Conference on e-Business and Telecommunications (ICETE 2020) - DCNET, OPTICS, SIGMAP and WINSYS, pages 27-34

ISBN: 978-989-758-445-9

networks has been extensively studied during the pre-

vious years (Shen and Tucker, 2009), (Chabarek et al.,

2008), (Melidis et al., 2019), (Dharmaweera et al.,

2014).

Various power-efﬁcient algorithms considering

the design of IP over EON (Zhu et al., 2019) can be

found in the literature. A fairly common, yet effective

method of energy saving is the extensive application

of optical bypass, reducing thus the number of high

energy-consuming optical-electrical-optical (O-E-O)

conversions, as the signal can be transported, ampli-

ﬁed and switched directly in the optical domain. In

(Zhang et al., 2015), energy efﬁcient trafﬁc grooming

in IP-over-elastic optical networks taking into account

sliceable optical transponders is studied. MILP mod-

els among their corresponding heuristics are imple-

mented, for each of three different types of bandwidth

variable transponders, and investigated in terms of en-

ergy efﬁciency. Based on trafﬁc and optical groom-

ing methods, Selene heuristic (Kyriakopoulos et al.,

2018) is an online algorithm which exploits the in-

novative Signal Overlap technique for power savings

in EONs. The work in (Vizca

ıno et al., 2012) is dedi-

cated to the study of energy efﬁciency in optical trans-

port networks, comparing the performance of an in-

novative ﬂexible network grid based on Orthogonal

Frequency Division Multiplexing (OFDM) with that

of Wavelength Division Multiplexing (WDM) with

a Single Line Rate (SLR) and a Mixed Line Rate

(MLR) operation. Energy-aware heuristic algorithms

are proposed for resource allocation both in static

(ofﬂine) and dynamic (online) scenarios with time-

varying demands for the Elastic-bandwidth OFDM-

based network and WDM networks (with SLR and

MLR). Lopez et al. in (Vizcaino et al., 2012), pro-

vides an in depth energy efﬁcient comparison between

conventional path protection schemes for ﬁxed-grid

(WDM) and ﬂexible-grid (EON) networks.

Besides the above mentioned techniques, a con-

siderable number of published articles pertaining to

artiﬁcial intelligence (AI) approaches in conjunction

with energy efﬁciency issues in optical networks can

be found in the literature (Musumeci et al., 2018),

(Mata et al., 2018). Kyriakopoulos et al. in (Kyr-

iakopoulos et al., 2014) propose a heuristic method

based on ant colony optimization to reduce network

energy footprint by exploiting the basic principles

of swarm intelligence for ﬁnding the most energy-

efﬁcient routes from source to destination nodes. In

addition, a multi-objective genetic algorithm is pro-

posed by Fern

andez et al. in (Fern

andez et al., 2012)

to design virtual topologies in order to reduce both

energy consumption and network congestion.

Optical switching node

Line ampliﬁer

IP routers

IP Layer

Optical Layer

BVT

Figure 1: IP-Over-EON Architecture.

2 NETWORK MODEL

2.1 IP Over EON Architecture

A typical IP Over EON architecture, as shown in Fig.

1, is considered. The IP Over EON consist of two lay-

ers, the IP and the optical layer. In the IP layer, each

node is equipped with a central IP router, while the

optical layer consists of the optical switching nodes

connected with ﬁber optic cables. The optical layer

offers the link between the IP routers. In each node

multiple trafﬁc streams from access network enters

the IP router. Each IP router port is connected to the

optical switching node through BVTs. At the start-

ing point of the data transmission, BVTs are responsi-

ble to convert the electrical ﬂows from the IP layer to

optical ﬂows (E/O conversion), then the trafﬁc enters

the optical domain and is routed in all optical con-

nections over the optical network. When all optical

trafﬁc traveling along the lightpath reaches its desti-

nation, the BVTs converts the signal back to electrical

(O/E conversion) and ﬁnally reaches the end point at

the IP layer. Data are then forwarded and handled by

the corresponding IP router. Finally, to enable optical

signals to travel over long distances, erdium doped

ﬁber ampliﬁers (EDFAs) are used in ﬁber optic con-

nections.

2.2 Elastic Optical Transponder

Technologies

Two types of transponder technologies according to

their sliceability degree can be categorised as follows.

Non-Sliceable BVT. This type of transponder is de-

signed to provide ﬂexible lightpaths. NS-BVT allows

any optical channel with any spectral width and cen-

tral frequency to be established. NS-BVT has only

one slice and it is exclusively used to serve one light-

path, and thus it is called non-sliceable. Due to its

high available bandwidth it is offered to serve future

DCNET 2020 - 11th International Conference on Data Communication Networking

demands (i.e 400 Gbps). However, it often suffers

from low utilization.

Sliceable BVT. To overcome the above inﬂexibility of

NS-BVT, sliceable BVT were proposed in the litera-

ture (Sambo et al., 2017), (Jinno et al., 2012). Unlike

NS-BVT, this type of transponder which is also de-

signed to provide ﬂexible lightpaths, allows more than

one lightpath to be established in the same transpon-

der. A physical transponder can be logically sliced

into multiple sub-transponders, each of which can

serve an independent lightpath between source and

destination nodes without electric processing at inter-

mediate nodes. As a result, various optical ﬂows can

be aggregated into one optical transponder in order

to improve its utilization. This feature of S-BVT en-

ables optical grooming (Zhang et al., 2015), which

can additionally, signiﬁcantly improve energy efﬁ-

ciency, since no new transponders required for new

connections to be accommodated.

3 POWER CONSUMPTION

ANALYSIS

The main components, used in this study, which can

inﬂuence the amount of power consumption on an IP

Over EON are the IP router ports, the S-BVTs and

the EDFAs. A 400 Gbps IP router port, which con-

nects the IP router to the BVT is considered. An IP

router port consumes 560 W (1) (Zhang et al., 2015).

The power consumption of a BVT can be expressed

as in (2) according to (Zhang et al., 2015). TR rep-

resents the transmission rate of the optical transpon-

der, where in case of a sliceable transponder indicates

the sum of transmission rates of all sub-transponders.

An additional 20% of power consumption is consid-

ered as an overhead contribution for each transponder.

Moreover, it is assumed that the energy consumption

of the transmitter and the receiver are identical and are

equal to half of the power consumption of a transpon-

der. Erbium Doped Fiber Ampliﬁers are considered

as ampliﬁers in this study. The power of the EDFA is

represented in Equation (3), in which X is the spec-

trum width for amplifying. An inline ampliﬁer is de-

ployed every 80km along the ﬁber, while a postampli-

ﬁer as well as a pre ampliﬁer are required at the ends

of the ﬁber link. The total power consumption is cal-

culated by adding the total energy consumption of the

BVTs, the EDFAs and the IP router ports (4).

= 560(Watt) (1)

BV T

= 1.683 ×T R(Gb/s) + 91.333(Watt) (2)

EDFA

= 0.0075 × X(GHz)(Watt) (3)

total

= PC

+ PC

BV T

+ PC

EDFA

(Wat t) (4)

4 THE LEARNING ENERGY

SAVING ALGORITHM (LESA)

4.1 Learning Automata Mechanism

Learning Automata (LA) are artiﬁcial intelligence

tools that can be applied to learn the characteristics

of a system’s environment. One major advantage of

LA is that they do not need to have any knowledge of

the environment they operate or any analytical knowl-

edge of the task to be optimized. A LA is a ﬁnite state

machine tool which improves its performance by in-

teracting with the random environment in which it op-

erates. The main purpose of a LA is to ﬁnd within

a set of actions the optimal one, that is the action

that causes the minimum average penalty received by

the environment (or the maximum average reward re-

ceived by the environment). The low computational

complexity that a LA exhibits enables it to rapidly

converge to the best action of the environment with

which it interacts.

Figure 2 illustrates the operation of a typical LA,

in which there is a set of possible actions a

, a

,.., a

as well as the corresponding probabilities p. P(n) =

(n), p

(n),..., p

(n) constitutes a vector which rep-

resents the probability distribution for M actions at

each instant n. It holds that

∑

i=1

(n) = 1. At ﬁrst,

the LA has no speciﬁc knowledge about the environ-

ment it operates and as a consequence all initial prob-

abilities are considered to be equal. At each instant

n, an action a

1 ≤ i ≤ M is selected with probability

(n). The action chosen by the automaton responds

with a stohastic reaction β

(n), which is used to up-

date the probability vector P. Upon completion of

this update, the LA selects the next action based on

the updated probability vector p

(n+1)

. This means that

the probabilities of some actions are increased or de-

creased according to the feedback received from the

environment.

4.2 Adaptive Model Formulation using

Learning Automata

Regarding the use of LA, in the context of this study,

is the detection of an acceptable number of BVTs that

should be switched off so that one manages to achieve

important energy savings while maintaining the BBP

A Learning Automata-based Algorithm for Energy-efﬁcient Elastic Optical Networks

Random Environment

Learning Automaton

ai(n)

βi(n)

pi(n)

Figure 2: Operation of a Learning Automaton.

at low levels. In short, there are 2 actions that lead to

the next or previous state. This way the LA, based on

the corresponding probabilities, estimates where the

transition will take place. Equations (5) - (12) cor-

responds to the probability updating scheme of the

learning automaton that was described in the previous

section. At each cycle n, the basic choice probability

P of the selected action a is updated according to the

network feedback reaction. P

(t) refers to action

−1

(t) refers to action

−1

, whereas the term state (S)

refers to the number of BVTs switched off from net-

work nodes (i.e. S

corresponds to 50% of free dis-

abled BVTs per node), as it could be seen in Figure 3.

LA can then choose, based on P, whether to increase

the number of BVTs to be switched off (S

) per node

by action

, or decrease the number of BVTs to be

switched off (S

−1

) per node by action

−1

. At ﬁrst, the

ratio of energy savings to BBP is estimated for a spe-

ciﬁc state. Afterwards, the LA checks if this ratio of

the state is greater than the calculated ratio of the pre-

vious state. Should the ratio be greater or equal than

the previously estimated ratio the basic choice proba-

bility of a increases according to (5), (6), (9) and (10).

Otherwise, the basic choice probability of a decreases

according to (7), (8), (11) and (12). L is a parameter

that governs the speed of the automaton convergence.

Two L values are used in this study, L

= 0.01 and

= 0.05.

For example, it is assumed that at a certain cycle

n, the LA is found at state 3 (S

), the corresponding

actions are action

−1

with P

−1

= 0.45, action

with

= 0.55 and the ratio of energy savings to BBP is

r. At cycle n + 1, the LA chooses the action with the

greater probability, P

(action

) and the new state

is 4. Then the ratio r

for state 4 is estimated and

compared to previously estimated ratio r. Should the

ratio r

be greater than r, the LA receives a rewarding

response and as a consequence updates the probabil-

ity scheme using (5) and (6). Finally, the LA is now at

state 4 and the probabilities for action

−1

and action

are P

−1

= 0.4455 and P

= 0.5545 respectively. This

procedure is repeated until the LA converges to a cer-

tain state.

State

action+1action-1

P(+1)(t)

P(-1)(t)

Figure 3: LESA learning mechanism.

(+1)u

(t + 1) = P

(+1)

(t) + L

× (1 − P

(+1)

(t)) (5)

(−1)u

(t + 1) = 1 − P

(+1)

(t) (6)

(+1)d

(t + 1) = P

(+1)

(t) − L

× P

(+1)

(t) (7)

(−1)d

(t + 1) = 1 − P

(+1)

(t) (8)

(−1)u

(t + 1) = P

(−1)

(t) + L

× (1 − P

(−1)

(t)) (9)

(+1)u

(t + 1) = 1 − P

(−1)

(t) (10)

(−1)d

(t + 1) = P

(−1)

(t) − L

× P

(−1)

(t) (11)

(+1)d

(t + 1) = 1 − P

(−1)

(t) (12)

4.3 Algorithm Description

The main idea of the proposed algorithm, namely

LESA, is the design of an energy efﬁcient scheme

which manages to reduce the total energy con-

sumption during network’s operation, by adaptively

switching off a number of BVTs in low-use scenar-

ios without affecting the BBP. LESA algorithm con-

sists of two separate periods. The ﬁrst period involves

the observation phase of the algorithm, during which

calculations are made regarding the utilization of the

BVTs. The second period refers to the use of LA

for estimating the relation between the energy sav-

ings achieved and BBP under a different number of

excluded BVTs (learning phase). Finally, the value

that was indicated by the LA, constitutes the most pre-

ferred one between the energy savings achieved and

BBP. That is, the number of BVTs to be switched off

in order for the BBP not to be affected signiﬁcantly.

During the observation period, the algorithm starts

routing the trafﬁc demands which arrive dynamically

in the network. LESA calculates the shortest paths

between the node pairs, using the k-shortest path

method, and routes the demands according to the First

Fit algorithm, while ensuring the continuity and conti-

guity constraint. During this phase, the existing BVTs

on the physical topology, as well as the BBP are moni-

tored for a ﬁxed number of arrivals. Transmitters’ and

receivers’ utilization percentages for each node in the

physical topology have been calculated. Afterwards,

DCNET 2020 - 11th International Conference on Data Communication Networking

the mean BVT utilization per node is estimated. In the

ﬁnal step of this period, the power consumption, using

(1), (2) and (3), as well as the BBP of the initial phys-

ical topology are estimated. In addition, the algorithm

outputs the number of free BVTs and the total num-

ber of BVTs per node after a certain percentage of the

free BVTs have been removed. Observation’s phase

output is used as an input to phase two of the algo-

rithm, the decision making with Learning Automaton

phase (learning phase).

Algorithm 1 shows the pseudocode of the pro-

posed algorithm LESA during the learning phase with

a learning automata mechanism. Tran

[i][x], is an ar-

ray which constitutes the number of BVTs per node,

where x indicates the node on the physical topology,

when i% of the free BVTs have been removed, i.e.

i = 0%, 10%, 20%,.., 100%. This array corresponds

to the states that the learning automaton can be found.

In detail, S

corresponds to i = 20%, while S

corre-

sponds to i = 80%. LA may chose to either increase

the number of removed BVTs, action

(+1)

, from the

physical topology with P

(+1)

, or decrease the num-

ber of removed BVTs, action

(−1)

, from the physical

topology with P

(−1)

. Firstly, the algorithm chooses

randomly a state sr and calculates BBP

for this state,

as well as the energy gains (ES

) of this state in com-

parison to PC

total

given from phase 1. Then the al-

gorithm retrieves the action with the highest probabil-

ity from ActionVector (lines 10 and 21), action

(+1)

or action

(−1)

which corresponds to S

sr+1

or S

sr−1

re-

spectively, runs the simulation for a ﬁxed number of

arrivals and estimates the new BBP

for the current

state, as well as the energy gains (ES

) of this state

in comparison to PC

total

given from phase 1. Should

the ratio

BBP

be greater or equal to

BBP

, the learning

automaton updates the updating probability scheme

according to (5), (6), (9) and (10). Equations (5)

and (6) are applied when the LA rewards the incre-

ment of the switched off BVTs, while (9) and (10)

when the LA rewards the decrement of the switched

off BVTs. Otherwise, the LA updates the updating

probability scheme using (7), (8), (11) and (12). By

the end of this period, the algorithm ends up (conver-

gence of LA) with the estimated value of percentage

of switched off BVTs (S).

5 PERFORMANCE EVALUATION

A set of simulation experiments were conducted, in

order to evaluate the performance of the proposed al-

gorithm LESA. To estimate the overall power con-

sumption of different design solutions, the metropoli-

tan mesh network (Antoniades et al., 2004) of Figure

Algorithm 1: LESA, Learning Phase.

Input:

G(N, L): Physical Topology

N: Set of nodes in the network

L: Set of links in the network

total

: Total PC from Phase 1

Total BBP from Phase 1

Tran

[i][x]: Number of BVTs per node

x ∈ N according to i

1: i ← removedtransponders 0%, 10%, .. , 100%

2: S ← Tran

[i]

3: ActionVector ← [S,action

(−1)

,action

(+1)

]

4: action

(−1)

← P

(−1)

5: action

(+1)

← P

(+1)

6: Choose state (S) sr randomly

7: Calculate ES

compared to initial PC

total

for S

8: Calculate new BBP

for S

9: while y ≤ trainingLA,y = 0 do

10: if ActionVector[2] ≥ ActionVector[1] then

11: sr ← sr + 1

12: Run Simulation for S

13: Calculate new ES

and BBP

14: if

BBP

≥

BBP

then

15: Estimate P

(+1)u

(t + 1) using (5)

16: Estimate P

(−1)u

(t + 1) using (6)

17: else

18: Estimate P

(+1)d

(t + 1) using (7)

19: Estimate P

(−1)d

(t + 1) using (8)

20: end if

21: else

22: sr ← sr − 1

23: Run Simulation for S

24: Calculate new ES

and BBP

25: if

BBP

≥

BBP

then

26: Estimate P

(−1)u

(t + 1) using (9)

27: Estimate P

(+1)u

(t + 1) using (10)

28: else

29: Estimate P

(−1)d

(t + 1) using (11)

30: Estimate P

(+1)d

(t + 1) using (12)

31: end if

32: end if

33: y ← y + 1

34: end while

Output:

Convergence of LA

4, which consists of 29 nodes and 41 links was con-

sidered.

5.1 Simulation Parameters and

Assumptions

An elastic optical network simulator has been im-

plemented, using Python 3.7 on Spyder. The num-

ber of frequency slots (FS) on a link equals to 160.

A Learning Automata-based Algorithm for Energy-efﬁcient Elastic Optical Networks

Figure 4: Mesh based Metropolitan network.

The granularity of FS is 25 GHz, while the modula-

tion format used in every connection is assumed to

be the same during the whole simulation. Also, one

FS as a guard-band associated with each of the con-

nections is considered. Connection requests follow

a Poisson process with an average connection’s inter

arrival time (IAT) equals to 1 (λ), while their hold-

ing time follows a negative exponential distribution

with mean value (µ) and the offered load is deter-

mined by λ / µ (Erlangs). The latter is tuned to achieve

the desired trafﬁc load (Comellas and Junyent, 2015).

The number of FSs per connection corresponds to the

uniform distribution, while each new coming connec-

tion can take any value from 1 to 9 with a uniformly

distributed probability (Comellas and Junyent, 2015).

The source and destination nodes of a request are ran-

domly and independently selected from the network

topology. K-shortest path, with k=3, and the First Fit

scheme, are used for solving the RSA problem. 400

Gbps slicable-BVTs which can launch 10 sub-carriers

(a sub-carrier is associated to a FS) enabling optical

grooming, and each sub-carrier can carry a 40-Gbps

signal are considered in this study. The number of S-

BVTs per node is assumed to be 15. Results presented

below are averaged over 3 × 10

connection requests

per simulation and the LA training number is assumed

to be 100.

5.2 Simulation Results

In order to measure the energy-saving potential

of LESA, a simple non-energy aware routing and

spectrum assignment approach, namely Elastic case

(Comellas and Junyent, 2015), has been implemented.

Figure 5 depicts the total energy consumption versus

the offered load (Erlangs) between LESA and Elastic

case algorithm. Each point in the graph (concerning

the LESA algorithm) corresponds to the energy con-

sumption when using the value (S) obtained by LA

mechanism after a training time (x symbol indicates

the state S LA ﬁnds). The energy consumption of

Figure 5: Energy consumption (in Watt) of LESA.

Figure 6: BBP performance of LESA.

(a) Percentage of energy savings (%) and BBP for

50 Erlang under different states in LESA opera-

tion.

(b) Convergence of basic probabilities for different

states in LESA operation.

Figure 7: LESA performance for 50 Erlang.

DCNET 2020 - 11th International Conference on Data Communication Networking

(a) Percentage of energy savings (%) and BBP for

250 Erlang under different states in LESA opera-

tion.

(b) Convergence of basic probabilities for different

states in LESA operation.

Figure 8: LESA performance for 250 Erlang.

each compared methods rises in a common way as

the offered load increases. It is worth noticing that

LESA always outperforms the reference Elastic case

algorithm. Corresponding results obtained in terms

of power savings are summarized in Figures 7a and

8a for 50 and 250 Erlangs, respectively. These results

are translated into proﬁt by up to 50% and 33% Er-

langs, respectively.

Figures 7 and 8 illustrate the obtained results of

the proposed algorithm under different trafﬁc loads.

More speciﬁcally, Figures 7 and 8 show the perfor-

mance evaluation of LESA for 50 and 250 Erlangs,

respectively. For each trafﬁc load, two subﬁgures are

presented, with the ﬁrst (7a and 8a) representing the

percentage of energy savings under the different states

of the algorithm versus the BBP, whereas the sec-

ond (7b and 8b) representing the convergence of basic

choice probabilities of LA towards different levels of

energy-saving. The arrows shown in the line graphs

(7a and 8a) correspond to the increase or decrease in

the ratio

BBP

compared to the previous state.

An energy saving from 6% to 90% is achieved un-

der the offered load of 50 Erlangs, while the BBP

takes values from 0% to 66%, for states S

= 10%

to S

= 100%, respectively (Figures 7a and 7b). In

detail, for the ﬁrst six states (S

- S

) of the simulation

the BBP remains at zero levels, while the percentage

of power gain rises progressively as the number of

the switched off BVTs increases, since a signiﬁcant

number of BVTs as well as IP router ports is reduced.

For the rest of the states, the graph’s curve changes

signiﬁcantly, as the BBP shifts at a faster rate than

the energy savings. As a result, the most acceptable

value of the free BVTs that should be switched off is

60% or S

. Should 60% of free BVTs per node will

be switched off from the physical topology, a power

saving of about 50% is achieved, while the BBP still

remains at zero levels. Observations are veriﬁed on

Figure 7b which report the convergence of the LA.

As it could be clearly seen, the LA chooses the pre-

ferred value of S

most of times with a percentage of

49%. Similar results, pertaining to 250 Erlangs can

be seen in Figure 8.

Finally, BBP, in linear scale, versus the increas-

ing offered load is depicted in Figure 6. BBP of

both algorithms increases when the trafﬁc load in-

creases. As it could be seen, BBP remains the same

as long as the offered load is up to 100 Erlangs for

both LESA and Elastic case algorithm. Although, as

expected in higher offered load values the Elastic case

algorithm results in lower BBP, as the lightpaths have

more chances to be accommodated in a network with

a greater number of BVTs. However, the proposed

algorithm manages to save important amounts of en-

ergy without signiﬁcantly increasing the BBP.

6 CONCLUSIONS

A novel algorithm which makes use of Learning Au-

tomata (LA) in a mechanism that selectively switches

off BVTs in low-load scenarios to achieve energy sav-

ings, is presented in this work. LA, based on BBP ob-

servations, aims at ﬁnding the most acceptable num-

ber of BVTs that should be switched off so that there

is a noticeable increase in terms of energy gains with-

out affecting the BBP. Simulation results veriﬁed that

the proposed algorithm can achieve by up to 50% of

energy savings while keeping the BBP at low levels.

ACKNOWLEDGEMENTS

This research has been coﬁnanced by the Euro-

pean Union and Greek national funds through the

Operational Program Competitiveness, Entrepreneur-

ship and Innovation, under the call RESEARCH-

CREATE-INNOVATE (project code:T1EDK-05061).

A Learning Automata-based Algorithm for Energy-efﬁcient Elastic Optical Networks

REFERENCES

Antoniades, N., Roudas, I., Ellinas, G., and Amin, J. (2004).

Transport metropolitan optical networking: evolving

trends in the architecture design and computer model-

ing. Journal of lightwave technology, 22(11):2653.

Beletsioti, G. A., Papadimitriou, G. I., and Nicopolitidis,

P. (2016). Energy-aware algorithms for ip over wdm

optical networks. Journal of Lightwave Technology,

34(11):2856–2866.

Belkhir, L. and Elmeligi, A. (2018). Assessing ict global

emissions footprint: Trends to 2040 & recommenda-

tions. Journal of Cleaner Production, 177:448–463.

Chabarek, J., Sommers, J., Barford, P., Estan, C., Tsiang,

D., and Wright, S. (2008). Power awareness in net-

work design and routing. In IEEE INFOCOM 2008-

The 27th Conference on Computer Communications,

pages 457–465. IEEE.

Cisco (2019). Cisco visual networking index: Forecast and

trends, 2017 - 2022. Technical report.

Comellas, J. and Junyent, G. (2015). Improving link

spectrum utilization in ﬂexgrid optical networks.

IEEE/OSA Journal of Optical Communications and

Networking, 7(7):618–627.

Dao, H., Morvan, M., and Gravey, P. (2018). An efﬁ-

cient network-side path protection scheme in ofdm-

based elastic optical networks. International Journal

of Communication Systems, 31(1):e3410.

Dharmaweera, M. N., Parthiban, R., and S¸ekercio

glu, Y. A.

(2014). Toward a power-efﬁcient backbone network:

The state of research. IEEE Communications Surveys

& Tutorials, 17(1):198–227.

Fern

andez, N., Dur

an, R. J., de Miguel, I., Merayo, N.,

Aguado, J. C., Fern

andez, P., Jim

enez, T., Rodr

ıguez,

I., S

anchez, D., Lorenzo, R. M., et al. (2012). Surviv-

able and impairment-aware virtual topologies for re-

conﬁgurable optical networks: A cognitive approach.

In 2012 IV International Congress on Ultra Modern

Telecommunications and Control Systems, pages 793–

799. IEEE.

Jinno, M. (2017). Elastic optical networking: Roles and

beneﬁts in beyond 100-gb/s era. Journal of Lightwave

Technology, 35(5):1116–1124.

Jinno, M., Takara, H., Sone, Y., Yonenaga, K., and Hirano,

A. (2012). Multiﬂow optical transponder for efﬁcient

multilayer optical networking. IEEE Communications

Magazine, 50(5):56–65.

Kyriakopoulos, C. A., Papadimitriou, G. I., and Nicopoli-

tidis, P. (2018). Exploiting the signal overlap tech-

nique for energy efﬁciency in elastic optical networks.

In 2018 International Conference on Computer, Infor-

mation and Telecommunication Systems (CITS), pages

1–5. IEEE.

Kyriakopoulos, C. A., Papadimitriou, G. I., Nicopolitidis,

P., and Varvarigos, E. M. (2014). Energy-aware light-

path routing in optical networks based on adaptive

heuristics. In 2014 6th International Congress on Ul-

tra Modern Telecommunications and Control Systems

and Workshops (ICUMT), pages 15–22. IEEE.

Mata, J., De Miguel, I., Duran, R. J., Merayo, N., Singh,

S. K., Jukan, A., and Chamania, M. (2018). Artiﬁcial

intelligence (ai) methods in optical networks: A com-

prehensive survey. Optical switching and networking,

28:43–57.

Melidis, P., Nicopolitidis, P., and Papadimitriou, G. (2019).

Reserved energy-aware virtual topology management

for ip-over-wdm optical networks. Optical Switching

and Networking, 31:72–85.

Musumeci, F., Rottondi, C., Nag, A., Macaluso, I., Zibar,

D., Rufﬁni, M., and Tornatore, M. (2018). An

overview on application of machine learning tech-

niques in optical networks. IEEE Communications

Surveys & Tutorials, 21(2):1383–1408.

Sambo, N., Giorgetti, A., Cugini, F., and Castoldi, P. (2017).

Sliceable transponders: pre-programmed oam, con-

trol, and management. Journal of Lightwave Technol-

ogy, 36(7):1403–1410.

Shen, G. and Tucker, R. S. (2009). Energy-minimized de-

sign for ip over wdm networks. Journal of Optical

Communications and Networking, 1(1):176–186.

Soumplis, P. (2017). Routing and spectrum allocation al-

gorithms in elastic optical networks. PhD thesis.

Vizcaino, J. L., Ye, Y., L

opez, V., Jim

enez, F., Duque, R.,

and Krummrich, P. (2012). On the energy efﬁciency

of survivable optical transport networks with ﬂexible-

grid. In European Conference and Exhibition on Opti-

cal Communication, pages P5–05. Optical Society of

America.

Vizca

ıno, J. L., Ye, Y., and Monroy, I. T. (2012). Energy ef-

ﬁciency analysis for ﬂexible-grid ofdm-based optical

networks. Computer Networks, 56(10):2400–2419.

Zhang, J., Zhao, Y., Yu, X., Zhang, J., Song, M., Ji, Y., and

Mukherjee, B. (2015). Energy-efﬁcient trafﬁc groom-

ing in sliceable-transponder-equipped ip-over-elastic

optical networks. IEEE/OSA Journal of Optical Com-

munications and Networking, 7(1):A142–A152.

Zhu, M., Sun, Q., Zhang, S., Gao, P., Chen, B., and Gu,

J. (2019). Energy-aware virtual optical network em-

bedding in sliceable-transponder-enabled elastic opti-

cal networks. IEEE Access, 7:41897–41912.

DCNET 2020 - 11th International Conference on Data Communication Networking