Specialized Network Self-configuration: An Approach using
Self-Organizing Networks Architecture (SONAr)
Daniel Ricardo Cunha Oliveira
a
, Maur
´
ıcio Amaral Gonc¸alves
b
, Natal Vieira de Souza Neto
c
,
Fl
´
avio de Oliveira Silva
d
and Pedro Frosi Rosa
e
Faculty of Computing, Federal University of Uberl
ˆ
andia, Uberl
ˆ
andia, Brazil
Keywords:
Self-management, Self-configuration, Self-Organizing Networks, SDN, NFV, Cloud Computing.
Abstract:
The increasing complexity of computer networks is constantly requiring the proposal of new network au-
tomation models. These are mostly based on Software-Defined Networks (SDN) and Network Functions
Virtualization (NFV) with implementation of the features called self-* within the concept of Self-Organizing
Networks (SON), which includes self-configuration. Regarding this matter, some models have already been
proposed at a higher level of definition, but there is still a lot of practical work to be done. In this paper, we
present the modeling, implementation and performance of a network self-configuration system when a new
edge element, which demands a specialized type of configuration, has been inserted. This is a typical case
of a telecom network, in which edge network elements (NEs) usually require a very specific set of network
parameters to be configured. The system recognizes the element and implements this specific network seg-
ment parameterization, acting as a specialized ”plug-and-play” feature. The approach will be based on the
Self-Organized Networks Architecture (SONAr) framework, which was very suitable for this functionality.
1 INTRODUCTION
If there is an unquestionable fact in the recent his-
tory of technology, it is the fast growth of applica-
tions and the densification of devices connected to
the Internet, especially if we take into account ac-
cess on mobile devices. This growth tends to remain
at exponential levels, especially with technological
adventures such as 8K ultra-resolution TV, new vir-
tual/augmented reality devices, cloud gaming, mas-
sive Internet of Things (IoT) and other applications
that new access technologies, e.g. 5G, will support.
All this evolution brings intrinsic concerns with the
infrastructure that will support all this load.
In 2001, IBM demonstrated an apprehension with
the increase in the complexity of computer net-
works and the respective increase in cost that it im-
plies, in an article that defines the bases and con-
cepts for autonomic computing, including the con-
cepts of self-* (Ganek and Corbi, 2003), namely self-
a
https://orcid.org/0000-0003-4767-5518
b
https://orcid.org/0000-0002-6985-638X
c
https://orcid.org/0000-0001-5047-4106
d
https://orcid.org/0000-0001-7051-7396
e
https://orcid.org/0000-0001-8820-9113
configuration, self-healing, self-protection and self-
optimization, which defines the basis of “autonomic
computing”.
In this work, we aim at a Self-Organizing Net-
works (SON) implementation focused in the first
property, the self-configuration, which was imple-
mented in the context of a more complete frame-
work called Self-Organizing Networks Architecture
(SONAr), which provides the basis for implementing
the self-* functionalities that make up a SON system.
Furthermore, this proposal for a self-configuration
system will focus on network elements that require a
highly specialized network configuration to support
them (rather than the more usual generic network el-
ements), typically edge elements in a telecommuni-
cations network. It will present a model for describ-
ing this specialized configuration and the automated
process to implement it on a network without human
intervention.
2 BACKGROUND
In general, the term SON was popularized in the con-
text of automation of mobile network management,
Oliveira, D., Gonçalves, M., Neto, N., Silva, F. and Rosa, P.
Specialized Network Self-configuration: An Approach using Self-Organizing Networks Architecture (SONAr).
DOI: 10.5220/0010403601610168
In Proceedings of the 11th International Conference on Cloud Computing and Services Science (CLOSER 2021), pages 161-168
ISBN: 978-989-758-510-4
Copyright
c
2021 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved
161
having been standardized by 3GPP as a solution for
the administration of the increasing quantity and com-
plexity of network parameters since Release 8 (3GPP,
2018b) (3GPP, 2018a).
The SON properties in mobile networks allow the
minimization of human intervention in certain aspects
of its management. However, SON functionalities
in mobile networks concern the parameterization of
access network configurations and not extend to net-
work core elements (TELUS and Huawei, 2016).
2.1 SDN and NFV Paradigm
SDN and NFV concepts were born out of the indus-
try requirements to evolve traditional network archi-
tectures into more flexible models that would allow a
greater degree of automation and freedom from ven-
dors. The first allows the infrastructure to be ab-
stracted into network applications and services, so
that it can be treated as a logical or virtual entity
(Open Networking Foundation, 2012) and the later
involves the implementation of network functions in
software that can run on a range of industry standard
server hardware (Chiosi et al., 2012).
In this work, it is assumed the adherence of both
paradigms in the target networks.
2.2 Related Work
An approach towards establishing a framework and a
reference architecture for SON combining the SDN
and NFV approaches was made in (Neves et al.,
2016), with the SELFNET project. It proposes high
level SON functions, focusing on self-healing, self-
protection and self-optimization, with more emphasis
on the latter, with self-configuration not being treated.
The article itself does not present practical implemen-
tations or results, however other later publications al-
ready show measured results, especially dealing with
self-optimization (Jiang et al., 2017).
Superfluidity is a project within Horizon 2020
program and aims to present a flexible and adaptive
network architecture, focusing mainly on important
aspects for 5G, in a context of NFV and SDN net-
works (Bianchi and et al., 2016). It defines logical
entities that perform a set of functionalities and the
objective is to obtain a formal and agnostic descrip-
tion of how these entities should work. Nevertheless,
the work does not present any practical implementa-
tion proposal other than the framework itself.
(Katiyar et al., 2015) proposes a self-configuration
mechanism for a new SDN switch inserted in hybrid
SDN and non-SDN networks. The system was imple-
mented and tested in a simple network model, but no
response time or convergence metrics were presented,
and it is also not inserted in a larger context of SON.
Finally, it does not address the self-configuration of
specialized NEs.
To the best of our knowledge, no work with a com-
plete range of the self-* properties specially with a
particular focus on self-configuration for specialized
elements – has been presented yet.
3 SELF-ORGANIZING
NETWORKS ARCHITECTURE
(SONAr)
A SON system must run on a well-defined archi-
tecture. A practical framework and implementa-
tion was proposed in our previous work (Gonc¸alves
et al., 2020), the Self-Organizing Networks Architec-
ture (SONAr), which was chosen to be the basis for
the specialized self-configuration features presented
in this work.
The SONAr framework is shown in the Figure 1.
It consists of five layers that have specific functions
and functionalities, which will be described below. In
the next subsections, a brief description of the lay-
ers and components of this framework will be pre-
sented and all component descriptions will be based
on (Gonc¸alves et al., 2020), unless otherwise men-
tioned.
3.1 Client Layer and SDN Network
Layer
In the Client Layer are placed the edge elements that
will be added to the network and that will require an
automatic configuration in order to allow full opera-
tion within it, a process which is the object of study
of this work (specialized self-configuration).
The SDN Network Layer can be logically divided
into two: the lower one represents an infrastructure
sublayer, composed of the network elements them-
selves, e.g. switches, routers and bridges, whether
they are physical or virtual elements , while the up-
per sub-layer, centralizing the control plane, hold the
primary “intelligence” of the network.
3.2 Interface Layer
This layer comprises the Collecting Entities (CoEs)
composed of several modules that together collect
metrics, topology information, alarms, samples, logs
and results during the network operation and other
components that will be presented below.
CLOSER 2021 - 11th International Conference on Cloud Computing and Services Science
162
Figure 1: Simplified layered diagram of the SONAr archi-
tecture.
Network Event Manager (NEM): it is a pub-
lisher/subscriber provider that manages events,
which, in turn, are categorized into topics. Network
Service Broker (NSB): it is used to configure services
based on intentions expressed in natural or formal lan-
guage by customers or system administrators. Control
Primitives Interceptor (CPI): It is located in the South-
bound Interface (SBI), acting as a proxy between the
network and the controller. It collects a large base of
information related to network management and store
it in the Network Database (NDB).
3.3 Management Layer
The management layer can be divided into two sub-
layers: the first integrates the Self-Organizing Entities
(SOE) and the last combines the Self-Learning Enti-
ties (SLE). SOEs are responsible for performing real-
time algorithms related to the fundamentals of auto-
nomic computing. The components of the SOE sub-
layer are as follows.
Self-Configuration Entity (SCE) is the most rel-
evant to this work. It acts in two moments, in the
network bootstrapping process and when a new NE
is plugged into it, in the process called plug-and-play.
The last can be further divided into self-configuration
of general cases and specific cases applicable to spe-
cialized edge NE (Oliveira et al., 2020), which will be
the investigation topic of this work.
Self-Healing Entity (SHE) implements algorithms
that allow to predict or detect failures in all the ele-
ments shown in the Figure 1, providing solutions for
recovering them. Self-Optimization Entity (SOPE)
aims to intervene in the network in order to maximize
its performance, based on the collected and analyzed
statistics. Self-Protection Entity (SPE) guarantees
the protection of all layers presented in the Figure 1
against internal and external attacks. Self-Planning
Entity (SPLE) organizes, controls and schedules all
actions on the network coming from the other enti-
ties of this layer. Self-Orchestration Entity (SORE)
has the role of receiving all actions from other enti-
ties and organizing them in queues, prioritizing these
actions.
Above the SOE sub-layer is the SLE sub-layer.
The purpose of these components is to provide AI-
based mechanisms to overcome the absence of hu-
man intervention. Finally, the NDB implements the
databases that aggregate all information generated
and stored in SONAr.
3.4 Administration Layer
The Administration Layer makes up the last level of
the SONAr. It implements the Network Adminis-
tration Dashboard (NAD), where network operators
can access SONAr data and administrate the system,
Operations Support System (OSS), Business Support
System (BSS) and Service Level Agreement (SLA)
Monitoring Systems.
4 SPECIALIZED
SELF-CONFIGURATION
MODULE (SSCM)
The purpose of this work is the implementation of
a module within the SCE in the SONAr system, i.e.
Specialized Self-Configuration Module (SSCM).
The goal of SSCM is to allow for a specific con-
figuration of a segment of a network in response to
the insertion of a new network element that requires
a specialized level of configuration, such as an edge
element that requires specific network parameteriza-
tion for its functioning. The module must identify
this new NE (typically an edge element with tele-
com functions), check if there are specific configura-
tions described for it and, if any, implement them on
the network, acting as a specialized plug-and-play. It
should be flexible for use in any practical cases, that
is, support the configuration for any types of edge el-
ements and all topology settings.
Specialized Network Self-configuration: An Approach using Self-Organizing Networks Architecture (SONAr)
163
4.1 SSCM Domain and Architecture
The applicability of SSCM operation presupposes
that there are two well-defined points for the begin-
ning and ending of the specialized configuration: the
first point is the exit interface of the edge network el-
ement itself (edge NE), and the second the exit inter-
face of the element where the network converges, usu-
ally a gateway router or an aggregating switch, where
the configuration process will stop. The final desti-
nation elements of the packets sent by the edge ele-
ment can be inside or outside the SONAr domain and
can be one or more, but the edge element itself must
be unique (identifiable by its outgoing interface) and
must also be necessarily within the SONAr domain.
Both SONAr and SSCM domains are represented in
the Figure 2.
Figure 2: Topology requirements for SSCM applicability.
For the network to be configured correctly after the
insertion of an edge element, a detailed description
of this configuration must be informed to SONAr.
This is done by an specific entry in a keyspace on
NDB, named Specialized Self-Configuration Descrip-
tion (SSCD).
SSCM is a module within the SCE that acts only
when, during the plug-and-play process, the MAC ad-
dress of the element is matched to an SSCD. If not,
the plug-and-play process normally occurs, consid-
ering that the connected element is a standard NE.
But, if the physical address of the connected inter-
face belongs to an element contained in this NDB ta-
ble, a change in the SCE flow will be done, so that
the specific configuration is made. This SSCD is im-
plemented in JSON content-type format and contains
the configuration characteristics for the network that
will receive the element. The SSCD is added into the
NDB through the NAD module, in the Administra-
tion Layer, which provides a friendly interface with
the operator or system administrator.
The architecture of the SONAr system can be re-
vised, including the SSCM module and the table with
the SSCDs, as shown in Figure 3. This figure com-
plements the Figure 1, with the identification and lo-
cation of the new modules within the layers that com-
pose the SONAr.
Figure 3: Simplified layered diagram of SONAr architec-
ture showing SSCM module and SSCD database.
Note that SSCM is a module in SCE, together with
bootstrapping and plug-and-play and SSCD is within
NDB.
4.2 SSCM Operation
SSCM is a SONAr additional service for plug-and-
play service: once active, when it detects that the in-
serted element is described by an SSCD in NDB, thus
demanding a specific configuration, SSCM is started
to carry out this process. The key that triggers it is
the MAC address of the added edge element, which
is also the key field in the NDB’s specific keyspace,
which may correspond to an SSCD. While plug-and-
play continues to play its roles, configuring the natural
routes between elements enabling ARP and LLDP
routing amongst other default signaling forwarding
SSCM takes care of more specialized configurations,
such as packets with IEEE 802.1Q header (VLAN),
prioritization in layer 2 (PCP), layer 3 (DiffServ), spe-
cific routes and other actions, depending on the spec-
ification described in the SSCD. These configurations
are performed on all NEs in the path between the ini-
tial element (edge NE) and the final element identified
on the SSCD (which is not necessarily the final des-
tination element of the packets); additionally, specific
flow rules are implemented to all elements in the path
between these two NEs. Thus, SSCM assumes that
the bootstrapping has already been performed (which
is always true) and the basic plug-and-play configura-
tions have already been implemented, so that SSCM
is the last module to be executed within the SCE.
The insertion of SSCM in the context of self-
CLOSER 2021 - 11th International Conference on Cloud Computing and Services Science
164
configuration in SONAr leads to some changes in the
established flows of SCE. Essentially, two processes
are changed to fit the SSCM module. These changes
are shown in Figure 4.
The first change is in the Address Provider (AP),
which is SONAR’s customized DHCP server: when
receiving the DHCPDISCOVERY message, the AP
first queries the NDB to check if there is an SSCD
associated with the MAC address contained in the
broadcast message. If so, in the DHCPOFFER mes-
sage it offers the IP address and gateway address spec-
ified on the SSCD, instead of searching for a unre-
served IP on NDB. This ensures that the edge NE can
have its IP and gateway set by the engineer or network
operator even before it is plugged into the network.
The second change occurs in the SCE: after gen-
erating the flow rules with the default routes, the SCE
checks the MAC address of the inserted element’s in-
terface. If it is in the SSCD base in NDB, SCE starts
a specialized configuration process, generating spe-
cific flow rules. This process allows that, in addition
to generic actions, e.g. simple default signaling route
configurations, the SCE makes more elaborated con-
figurations in the topology, as specified in the SSCD.
After the generation of all flow rules is finished, they
are all sent to the controller through NBI.
Figure 4: Message flow exchanges in SONAr elements by
implementing SSCM.
4.3 SSCD Format and Functionalities
The entire process of self-configuration in SSCM be-
gins with the specification of the specialized config-
uration to be made on the network so that it is pre-
pared to connect a given edge element to its destina-
tion. This specification is performed through an ob-
ject in JSON format containing the attributes describ-
ing the specificities of the network in question, the
SSCD. This SSCD is a record in the NEBootConfig-
uration keyspace, on NDB. An example of an SSCD
entry is shown below:
{
"id": {
"mac_address": "00:00:00:00:00:01",
"description": "Edge NE",
"ip": "192.168.0.100",
"gateway": "192.168.0.200"
},
"config_end_node": {
"mac_address": "00:00:00:00:00:02",
"ip": "192.168.0.200",
"description": "Provider edge router"
},
"dst_node": [
{
"description": "First Destination",
"ip": "10.1.1.1",
"priority": 5000,
"vlan": 10,
"pcp_cos": 1.
...
},
{
"description": "Second Destination",
...
}
]
}
The attributes of this SSCD object are interpreted
as follows:
id: It is the identifier of the edge element to be au-
tomatically inserted in the network and the one that
marks the starting point of the network configuration.
It specifies the MAC address, which is the main iden-
tifier of the element (and which will be used to trigger
the SSCM process) and other complementary infor-
mation, such as the IP and gateway. If the IP and
gateway identifiers do not exist, they will be defined
and assigned by the AP, otherwise the AP will read
these identifiers and send them in the DHCPOFFER
message to the edge NE. The description and respon-
sible identifiers only fulfill documentation roles.
config end node: It identifies the final element of
the network’s self-configuration scope, that is, the
self-configuration process is carried out in the path
between the edge element and the element indicated
in this field. This element is typically a gateway or
an aggregator element. Its main identifier is also the
MAC address of the outgoing port for the next ele-
ment (which will not be automatically configured).
dst node: It is an array containing the possible
destinations of the packets sent by the edge element
and the specification of what to do with them. The
dst node must be beyond config end node and may
Specialized Network Self-configuration: An Approach using Self-Organizing Networks Architecture (SONAr)
165
be outside the SONAr domain. This is where the ac-
tions to be implemented in the network settings are
characterized. For example, a packet sent from the
edge element to a server (identified by the IP address)
must travel on a specific VLAN with a given Class
of Service (CoS). The priority field refers to the pri-
ority of the flow within the SDN context and not to
the prioritization of the packet on the network, and
it is recommended that it should be higher than those
of the generic flow rules inserted by the plug-and-play
(4,000 by default). The list of possible matching crite-
ria and actions to be implemented is given by (ONOS
Project, 2019). This array can contain as many target
elements with specific actions as required, depending
on the type of application.
4.4 Typical Application
The typical application for specialized self-
configuration is in telecommunications networks,
where it is compulsory to specify exactly the configu-
ration domain and the parameters to be implemented.
We can take as an example the integration of an
eNodeB (4G cellular station) to a network of an
MNO, as shown in Figure 5.
Figure 5: Example of a mobile network topology using
SSCM for specialized configuration.
The eNodeB in Figure 5 is the start node identified by
the id attribute on the SSCD, and is connected through
an access network, usually an MPLS or a metro ether-
net ring. This ring is integrated into the carrier’s trans-
port network by a provider edge router, and the net-
work core elements are in a third network, accessed
by a provider router. In the case of a cellular network,
the target elements are the Mobility Management En-
tity (MME), the System Architecture Evolution Gate-
way (SAE-GW), the OSS server and the clock (or
sync) server. These four core elements are the des-
tination nodes, specified by the dst node attribute on
the SSCD.
For each of the four destinations, the first switch
(S1) must, in order, insert an IEEE 802.1Q header
containing the VLAN tag in the package, mark the
package with a specific CoS, assign a priority to this
flow rule (recommended to be higher than the prior-
ities of the default flow rules which are 4,000 by de-
fault) and forward the packet on the correct outgo-
ing interface. The other switches on the path to the
provider edge router must forward the packets cor-
rectly and, finally, the last switch (access switch) must
remove the IEEE 802.1Q header (and with it, the CoS
mark) and forward the packet to the router. The same
configuration procedure must be performed for pack-
ets coming from the core of the network to eNodeB:
the access switch makes the markings while switch
S1 removes them.
5 EXPERIMENTS
In this section, we present the experimentation per-
formed in order to evaluate the overall performance
of SSCM and discuss the results we have obtained.
5.1 Experimentation Environment
For emulation of the network environment, GNS3
software version 2.1.16 was chosen for several rea-
sons. SONAr Server consists of all components of
the SONAr system and the SDN Controller (ONOS
2.1.0); it was implemented in VM VirtualBox 6.0, OS
Ubuntu Server 16.04 and developed using JDK 1.8.0-
151. NDB was implemented with Cassandra 3.11.4
and NEM with RAbbitMQ 3.8.0. The core and edge
elements were implemented using Docker 18.06.2-ce
containers, with OS Ubuntu 19.04. The switches and
routers were based on Open vSwitch (OvS) 2.10.1
and implemented as containers on Docker Engine
18.06.2-ce, with OS Linux Alpine 3.9.
The topology used was as shown in Figure 6.
In the access network (SSCM domain), different
amounts of in-line switches were used to assess per-
formance in each scenario.
5.2 Performance Evaluation
In this section, the experiments that evaluated the be-
havior of SSCM in networks with different amounts
CLOSER 2021 - 11th International Conference on Cloud Computing and Services Science
166
Figure 6: GNS3 topology adopted on experiments.
of elements in the access network will be dis-
cussed. Specifically, the scenarios for specialized
self-configuration for 3, 6, 9, 12, 15 and 18 in-
line switches were evaluated. All of these scenarios
were simulated starting from the network bootstrap-
ping to assess the impact of SSCM on the total self-
configuration convergence time, starting from discov-
ery until the final configuration of the network. There-
fore, we have three stages:
The discovery stage: when SONAr finds all net-
work elements in its domain through the TCoE
module, it establishes a topology in the form of a
graph, records this topology in the NDB and noti-
fies the other components of the completion of the
process by publishing an event in the NEM.
The generic configuration stage: it begins when
the SCE receives the discovery completion event
and the plug-and-play starts the configuration pro-
cess, which involves calculating the shortest path
between the SONAr server and the others net-
work components and the generation of flow rules
that will allow the communication of the elements
through the control and signaling protocols (ARP,
DHCP, LLDP, etc).
The specialized configuration stage: an SSCD,
corresponding to one of the network elements, is
matched in the NDB. In this case, SSCM performs
the calculation and generation of specialized flow
rules for the network, after which all these flow
rules (including the generic ones generated by the
plug-and-play module) are sent to SDNC via NBI.
For this experiment, each scenario was simulated five
times from the bootstrap and the times for each stage
mentioned above were measured in milliseconds by a
specific class, which recorded the intervals between
the events received at the NEM. In this way, we have
a very accurate measurement of the duration of the
three stages of the network self-configuration. Af-
ter measuring the times, the averages for each inter-
val and for each scenario were calculated, and then
the 95% confidence intervals were determined. All
graphs show the average along with the lower and up-
per limits of that range.
The results obtained for network discovery,
generic configuration and specialized configuration
are respectively shown in Figures 7, 8 and 9.
Figure 7: Time (in milliseconds) for network discovery for
each in-line amount of switches.
Figure 8: Time (in milliseconds) for network generic con-
figuration for each in-line amount of switches.
Figure 9: Time (in milliseconds) for network specialized
configuration for each in-line amount of switches.
Looking only at the Figure 9, we notice that the
SSCM convergence time does not vary substantially
with respect to the number of switches that are in
the path between the edge element and the provider
edge router. This is explained by the fact that, when
Specialized Network Self-configuration: An Approach using Self-Organizing Networks Architecture (SONAr)
167
the specialized configuration stage starts, all elements
have already been identified, all routes have already
been calculated and all generic configurations have
been carried out, so that SONAr already has full
knowledge of the network structure. There is, of
course, an increase in the time involved in the calcu-
lation of the specific flow rules (the greater the num-
ber of elements, the longer the time of this calcu-
lation), however the duration of this process is very
short when compared to the total self-configuration
time, remaining at tens of milliseconds.
In discovery and generic network configuration,
this relationship is much more obvious. In both cases,
discovery and the configuration are done element by
element in sequence, that is, it is necessary that one
is done for another to start. This means that there is
a direct and visually evident relationship between the
convergence time of both stages and the number of
elements in the path between the beginning and the
end of the network.
In general, the time for the specialized configura-
tion process – which is the one being evaluated in this
study was between 4 and 7 seconds for all samples
obtained. It is noticeable that SONAr’s response to an
edge element addition in the network is fast enough
for use in practice, exceeding by several orders of
magnitude the time of manual configuration of a net-
work even using semi-automated processes.
6 CONCLUSIONS
The Specialized Self-Configuration Module (SSCM),
was proposed as a solution for the demand of net-
work configuration automation. The main objective
of the research was to investigate the alternatives for
a network self-configuration system for elements with
specific requirements, together with a model to de-
scribe these requirements. During the investigation,
several approaches to this problem were studied, and
the SONAr framework was well suited to the purpose.
Within this architecture, a new module was designed
and implemented (the SSCM) and a way to describe
the specifics of the network configuration was pro-
posed (the SSCD format).
In the experiments carried out, it was visible
that the SSCM’s performance far exceeds the esti-
mated time for manual configuration, being a suit-
able approach to real life applications, specially in the
telecommunications industry.
REFERENCES
3GPP (2018a). Telecommunication management; Self-
configuration of network elements; Concepts and re-
quirements. Technical Specification (TS) 32.501, 3rd
Generation Partnership Project (3GPP).
3GPP (2018b). Telecommunication management; Self-
Organizing Networks (SON); Concepts and require-
ments. Technical Specification (TS) 32.500, 3rd Gen-
eration Partnership Project (3GPP).
Bianchi, G. and et al. (2016). Superfluidity: a flexible func-
tional architecture for 5G networks: G. Bianchi et al.
Transactions on Emerging Telecommunications Tech-
nologies, 27(9):1178–1186.
Chiosi, M., Clarke, D., Willis, P., Reid, A., Ferger, J., Bu-
genhagen, M., Khan, W., Fargano, M., Cui, C., Deng,
H., Benitez, J., and et all (2012). Network Func-
tions Virtualisation, An Introduction, Benefits, En-
ablers, Challenges & Call for Action.
Ganek, A. G. and Corbi, T. A. (2003). The dawning of
the autonomic computing era. IBM Systems Journal,
42(1):5–18.
Gonc¸alves, M., Souza Neto, N., Oliveira, D., Silva, F., and
Frosi, P. (2020). Bootstrapping and plug-and-play op-
erations on software defined networks: A case study
on self-configuration using the sonar architecture. In
Proceedings of the 10th International Conference on
Cloud Computing and Services Science - Volume 1:
CLOSER,, pages 103–114. INSTICC, SciTePress.
Jiang, W., Strufe, M., and Schotten, H. D. (2017). A SON
decision-making framework for intelligent manage-
ment in 5G mobile networks. In 2017 3rd IEEE In-
ternational Conference on Computer and Communi-
cations (ICCC), pages 1158–1162, Chengdu. IEEE.
Katiyar, R., Pawar, P., Gupta, A., and Kataoka, K. (2015).
Auto-Configuration of SDN Switches in SDN/Non-
SDN Hybrid Network. In Proceedings of the Asian
Internet Engineering Conference on - AINTEC ’15,
pages 48–53, Bangkok, Thailand. ACM Press.
Neves, P., Cal
´
e, R., Costa, M. R., Parada, C., Parreira,
B., Alcaraz-Calero, J., Wang, Q., Nightingale, J.,
Chirivella-Perez, E., Jiang, W., Schotten, H. D., Kout-
sopoulos, K., Gavras, A., and Barros, M. J. (2016).
The SELFNET Approach for Autonomic Manage-
ment in an NFV/SDN Networking Paradigm. In-
ternational Journal of Distributed Sensor Networks,
12(2):2897479.
Oliveira, D., Souza Neto, N., Gonc¸alves, M., Silva, F., and
Frosi, P. (2020). Network self-configuration for edge
elements using self-organizing networks architecture
(sonar). In Proceedings of the 10th International Con-
ference on Cloud Computing and Services Science -
Volume 1: CLOSER,, pages 408–414.
ONOS Project (2019). Flow Rules - ONOS. Available at
wiki.onosproject.org/display/ONOS/Flow+Rules.
Open Networking Foundation (2012). Software-defined
networking: The new norm for networks. White pa-
per.
TELUS and Huawei (2016). Next generation son for 5g.
CLOSER 2021 - 11th International Conference on Cloud Computing and Services Science
168