126,99 €
Mehr erfahren.
- Herausgeber: John Wiley & Sons
- Kategorie: Wissenschaft und neue Technologien
- Sprache: Englisch
This book reviews the concept of Software-Defined Networking (SDN) by studying the SDN architecture. It provides a detailed analysis of state-of-the-art distributed SDN controller platforms by assessing their advantages and drawbacks and classifying them in novel ways according to various criteria. Additionally, a thorough examination of the major challenges of existing distributed SDN controllers is provided along with insights into emerging and future trends in that area. Decentralization challenges in large-scale networks are tackled using three novel approaches, applied to the SDN control plane presented in the book. The first approach addresses the SDN controller placement optimization problem in large-scale IoT-like networks by proposing novel scalability and reliability aware controller placement strategies. The second and third approaches tackle the knowledge sharing problem between the distributed controllers by suggesting adaptive multilevel consistency models following the concept of continuous Quorum-based consistency. These approaches have been validated using different SDN applications, developed from real-world SDN controllers.
Sie lesen das E-Book in den Legimi-Apps auf:
Seitenzahl: 244
Veröffentlichungsjahr: 2022
Ähnliche
Table of Contents
Cover
Dedication Page
Title Page
Copyright Page
Acronyms
Preface
Introduction
I.1. General context
I.2. Problem statement and motivations
I.3. Main contributions
I.4. Structure of the book
1 Toward a Decentralized SDN Control Architecture: Overview and Taxonomy
1.1. Introduction
1.2. Software-defined networking: a centralized control architecture
1.3. Physical classification of existing SDN control plane architectures
1.4. Logical classification of existing SDN control plane architectures
1.5. Conclusion
2 Decentralized SDN Control: Major Open Challenges
2.1. Introduction
2.2. Scalability
2.3. Reliability
2.4. Controller state consistency
2.5. Interoperability
2.6. Other challenges
2.7. Conclusion
3 Scalability and Reliability Aware SDN Controller Placement Strategies
3.1. Introduction
3.2. Related work
3.3. The SDN controller placement optimization problem
3.4. The proposed SDN controller placement scheme
3.5. Performance evaluation
3.6. Discussion
3.7. Conclusion
4 Adaptive and Continuous Consistency for Distributed SDN Controllers: Anti-Entropy Reconciliation Mechanism
4.1. Introduction
4.2. Related work
4.3. The consistency problem in SDN
4.4. Consistency models in ONOS
4.5. The proposed adaptive consistency for ONOS
4.6. Performance evaluation
4.7. Conclusion
5 Adaptive and Continuous Consistency for Distributed SDN Controllers: Quorum-Based Replication
5.1. Introduction
5.2. Background on eventual consistency in distributed data stores
5.3. The proposed adaptive Quorum-inspired consistency for ONOS
5.4. Implementation approach on ONOS
5.5. Performance evaluation
5.6. Conclusion
Conclusions and Perspectives
C.1. Summary of contributions
C.2. Perspectives and future work
References
Index
Other titles from iSTE in Networks and Telecommunications
End User License Agreement
List of Tables
Chapter 2
Table 2.1.
Main characteristics of the discussed SDN controllers
Chapter 3
Table 3.1.
NSGA-II parameters
Table 3.2.
The maximum number of objective function evaluations (MaxEvaluatio
...
Chapter 4
Table 4.1.
Test scenarios
Chapter 5
Table 5.1
Application SLA scenarios
Table 5.2.
Final Q-learning results of the constrained and controlled agents
...
List of Illustrations
Chapter 1
Figure 1.1
Conventional networking versus software-defined networking.
Figure 1.2
A three-layer distributed SDN architecture.
Figure 1.3
Physical classification of SDN control plane architectures.
Figure 1.4
Logical classification of distributed SDN control plane architect
...
Chapter 2
Figure 2.1
The main challenges of distributed SDN control.
Chapter 3
Figure 3.1
The controller placement problem.
Figure 3.2
Controller placement metrics.
Figure 3.3
Strategies 1–3: latency-based performance metrics.
Figure 3.4
Strategy 3: load imbalance.
Figure 3.5
Strategy 4: reliability metrics (maximum latencies in failure-fre
...
Figure 3.6
Strategy 4: performance metrics.
Figure 3.7
Computation time comparison between PAM-B and NSGA-II over the co
...
Chapter 4
Figure 4.1
The proposed adaptive consistency strategy.
Figure 4.2
Scenario no. 1: captured inter-controller traffic (in packets per
...
Figure 4.3
Scenario no. 1: inter-controller overhead in ONOS and ONOS-WAC ac
...
Figure 4.4
Gain in anti-entropy overhead of ONOS-WAC with respect to ONOS ac
...
Chapter 5
Figure 5.1
Architectural overview of our adaptive Quorum-based consistency s
...
Figure 5.2
Reinforcement learning (RL) architecture.
Figure 5.3
The proposed adaptive consistency system
Figure 5.4
Quorum-inspired Write operations in our CDN-like application.
Figure 5.5
Quorum-inspired read operations in our CDN-like application.
Figure 5.6
Overview of the main tasks executed by our TCL-Expect scripts.
Figure 5.7
Workload 1: a read-intensive application scenario.
Figure 5.8
Workload 3: a write-intensive application scenario.
Figure 5.9
Scenario 1: a latency-sensitive application.
Figure 5.10
Scenario 2: a consistency/latency-balancing application.
Figure 5.11
Scenario 3: a consistency-favoring application.
Figure 5.12
Dynamic changes in the workload (Workload 2-Workload 1-Workload
...
Guide
Cover Page
Dedication Page
Title Page
Copyright Page
Acronyms
Preface
Introduction
Table of Contents
Begin Reading
Conclusions and Perspectives
References
Index
Other titles from iSTE in Networks and Telecommunications
Wiley End User License Agreement
Pages
ii
iii
iv
ix
x
xi
xiii
xiv
xv
xvii
xviii
xix
xx
xxi
xxii
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
115
116
117
118
119
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
139
140
141
To my parents, my sisters and my wonderful readersFetia Bannour
To my beloved family, my wife and my dear childrenSami Souihi
To my oldest, beloved and magnificent daughter Ikram on her birthday this year.Abdelhamid Mellouk
New Generation Networks Set
coordinated byAbdelhamid Mellouk
Volume 2
Software-Defined Networking
Extending SDN Control to Large-Scale Networks
Fetia BannourSami SouihiAbdelhamid Mellouk
First published 2022 in Great Britain and the United States by ISTE Ltd and John Wiley & Sons, Inc.
Apart from any fair dealing for the purposes of research or private study, or criticism or review, as permitted under the Copyright, Designs and Patents Act 1988, this publication may only be reproduced, stored or transmitted, in any form or by any means, with the prior permission in writing of the publishers, or in the case of reprographic reproduction in accordance with the terms and licenses issued by the CLA. Enquiries concerning reproduction outside these terms should be sent to the publishers at the undermentioned address:
ISTE Ltd27-37 St George’s RoadLondon SW19 4EUUK
John Wiley & Sons, Inc.111 River StreetHoboken, NJ 07030USA
www.iste.co.uk
www.wiley.com
© ISTE Ltd 2022The rights of Fetia Bannour, Sami Souihi and Abdelhamid Mellouk to be identified as the authors of this work have been asserted by them in accordance with the Copyright, Designs and Patents Act 1988.
Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s), contributor(s) or editor(s) and do not necessarily reflect the views of ISTE Group.
Library of Congress Control Number: 2022941880
British Library Cataloguing-in-Publication DataA CIP record for this book is available from the British LibraryISBN 978-1-78630-849-8
Acronyms
API
Application Programming Interface
AS
Autonomous System
CAP
Consistency Availability Performance
CDN
Content Delivery Network
CLARA
Clustering LARge Applications
CPP
Controller Placement Problem
DDBS
Distributed DataBase System
DHT
Distributed Hash Table
DoS
Denial-of-Service
ForCES
Forwarding and Control Element Separation
FSM
Finite-State Machine
IETF
Internet Engineering Task Force
IoT
Internet of Things
IXP
Internet eXchange Point
MD-SAL
Model-Driven Service Abstraction Layer
ML
Machine Learning
MOCO
Multi-Objective Combinatorial Optimization
NIB
Network Information Base
NSGA-II
Non-dominated Sorting Genetic Algorithm II
ODL
OpenDayLight
OF
OpenFlow
ONF
Open Networking Foundation
ONOS
Open Network Operating System
PACELC
Partition, tradeoff Availability and Consistency, Else, tradeoff Latency and Consistency
PAM
Partitioning Around Medoids
POCO
Pareto-Optimal COntroller
PSA
Pareto Simulated Annealing
QL
Q-Learning
QoE
Quality of Experience
QoS
Quality of Service
RL
Reinforcement Learning
RSM
Replicated State Machine
SDN
Software-Defined Networking
SDX
Software-Defined eXchange
SLA
Service-Level Agreement
SPOF
Single Point of Failure
TE
Traffic Engineering
UDP
User Datagram Protocol
WAN
Wide Area Network
XFSM
eXtended Finite-State Machine
Preface
Due to the emergence of new kinds of communication and networking technologies (e.g. the Internet of Things (IoT), mobile trends, network virtualization) and the rise of many advanced services (e.g. real-time services, e-health, multimedia, smart cities, gaming) supported by these technologies, today’s networks – considered relatively static, “ossified” and “challenging to manage”– are no longer suitable to handle the complexity and diversity of network information being disseminated in today’s modern and dynamic networking environments.
There is a strong need to shift the current network architecture to a new model that adapts to such changes and leverages new control strategies to ease network management and automation, leading to better network performance and lower operating costs. In this context, software-defined networking (SDN) has emerged as a new networking paradigm that decouples network control and forwarding functions, enabling the network control to become directly programmable and the underlying infrastructure to be abstracted for applications and services.
SDN attempts to centralize the network control, thus offering improved visibility and flexibility to manage the network, optimize its performance and reduce its operating costs. However, centralized SDN designs, in which the control plane logic is physically centralized in a single software component called the SDN controller, present numerous challenges including the issues of control plane reliability, scalability and performance. To meet these challenges, it is necessary for the SDN control architecture to evolve toward a physically decentralized system. However, such physically distributed but logically centralized SDN platforms bring additional challenges.
In this book, we aim to provide a thorough exploration of the SDN technology and, more importantly, we deal with the SDN decentralization problem in the context of large-scale networks. We propose novel approaches to decentralize the SDN control plane without forgoing the centralization benefits of SDN. Part of this book was initially based on the work conducted within the framework of Fetia Bannour’s PhD thesis. This work was subsequently developed into a book to facilitate understanding of the decentralized SDN control plane. The latter may indeed be implemented using the existing distributed SDN controllers. However, their significant number, along with their particular pros and cons, made the choice extremely difficult for those who attempted to adopt a distributed SDN architecture in large-scale deployments.
To provide useful guidelines for such SDN research and deployment initiatives, this book reviews the SDN concept by studying the SDN architecture compared to the traditional one and provides a detailed analysis of state-of-the-art distributed SDN controller platforms by assessing their advantages and drawbacks, classifying them in novel ways (physical and logical classifications) and comparing them with respect to various criteria. Additionally, a thorough discussion on the major challenges of existing distributed SDN controller platforms is provided along with insights into emerging and future trends in that area. Furthermore, to tackle some of the most prominent challenges related to the decentralization of the SDN control plane in large-scale networks, three novel approaches are proposed in this book.
The first approach addresses the SDN controller placement problem by proposing scalability and reliability aware strategies for the placement of the SDN controllers at scale, with respect to multiple reliability and performance criteria and according to different uses and contexts. These strategies use different types of multi-criteria optimization algorithms, which are compared in terms of computation time, and the quality of final controller placement configurations.
The second and third approaches investigate the knowledge sharing problem in a distributed SDN cluster by proposing adaptive and continuous consistency models. The main aim of these two approaches is to achieve a consistency adaptation strategy that provides balanced trade-offs at runtime between the application’s continuous performance and consistency requirements. These real-time trade-offs should provide minimal application inter-controller overhead while satisfying the application-defined thresholds specified in the application’s service-level agreements (SLAs). These models primarily focus on the anti-entropy reconciliation mechanisms. Then, they address the replication mechanisms by proposing an intelligent Quorum replication strategy. These approaches were validated using two SDN applications with eventual consistency needs that are developed on top of the open-source Open Network Operating System (ONOS) controllers: a source routing application and a CDN-like application.
When writing this book, we were mainly driven by our belief that SDN, together with network virtualization, will play a significant role in enabling a “full network softwarization” and reshaping the next generation of computer networks. We were also motivated by the lack of research work on decentralized SDN, which extends SDN control to large-scale networks. Our purpose is to provide useful guidelines and lessons learned for dealing with the decentralization problem in SDN for academic and industrial research purposes.
This book is a start but also leaves many questions unanswered. We hope that it will inspire the new generation of researchers. It would not have been possible without the valuable support of our students and colleagues, whom the authors would like to thank warmly.
Finally, the authors hope the readers will enjoy reading this book and learn many useful ideas and overviews for their own work and studies.
August 2022
Fetia BANNOUR
École Nationale Supérieure d’Informatique pour l’Industrie et l’Entreprise (ENSIIE)
Sami SOUIHI
Université Paris-Est Créteil (UPEC)
Abdelhamid MELLOUK
Université Paris-Est Créteil (UPEC)
Introduction
I.1. General context
The unprecedented growth in demands and data traffic and the emergence of network virtualization, along with the ever-expanding use of mobile equipment in the modern network environment, have highlighted major problems that are essentially inherent to the Internet’s conventional architecture. This made the task of managing and controlling the information coming from a growing number of connected devices increasingly complex and specialized.
Indeed, the traditional networking infrastructure is considered highly rigid and static as it was initially conceived for a particular type of traffic, namely monotonous text-based contents, which makes it poorly suited to today’s interactive and dynamic multimedia streams, generated by increasingly demanding users. Along with multimedia trends, the recent emergence of the Internet of Things (IoT) has allowed for the creation of new advanced services with more stringent communication requirements in order to support its innovative use cases. In particular, e-health is a typical IoT use case where the healthcare services delivered to remote patients (e.g. diagnosis, surgery, medical records) are highly intolerant of delay, quality and privacy. Such sensitive data and life-critical traffic are barely supported by traditional networks.
Furthermore, in the traditional architecture where the control logic is purely distributed and localized, solving a specific networking problem or adjusting a particular network policy requires acting separately on the affected devices and manually changing their configuration. In this context, the current growth in devices and data has exacerbated scalability concerns by making such human interventions and network operations harder and more error prone.
Overall, it has become particularly challenging for today’s networks to deliver the required level of quality of service (QoS), let alone the quality of experience (QoE) that introduces additional user-centric requirements. To be more specific, relying solely on the traditional QoS, based on technical performance parameters (e.g. bandwidth and latency), turns out to be insufficient for today’s advanced and expanding networks. Additionally, meeting this growing number of performance metrics is a complex optimization task, which can be treated as an NP-complete problem. Alternatively, network operators are increasingly realizing that the end-user’s overall experience and subjective perception of the delivered services are as important as QoS-based mechanisms. As a result, current trends in network management are heading toward this new concept, commonly referred to as the QoE, to represent the overall quality of a network service from an end-user perspective.
That said, this huge gap between, on the one hand, the advances achieved in both computer and software technologies and, on the other hand, the traditional non-evolving and hard to manage (Kreutz et al. 2015; Bannour et al. 2018b) underlying network infrastructure supporting these changes has stressed the need for an automated networking platform (Samaan and Karmouch 2009) that facilitates network operations and matches today’s network requirements such as the IoT needs (Ren et al. 2019; Montaño et al. 2021). In this context, several research strategies have been proposed to integrate automatic and adaptive approaches into the current infrastructure for the purpose of meeting the challenges of scalability, reliability and availability for real-time traffic, and therefore guaranteeing the user’s QoE.
While radical alternatives argue that a brand new network architecture should be built from scratch by breaking with the conventional network architecture and bringing fundamental changes to keep up with current and future requirements, other realistic alternatives are favored for introducing slight changes tailored to specific needs and for making a gradual network architecture transition, without causing costly disruptions to existing network operations.
In particular, the early overlay network alternative introduces an application layer overlay on top of the conventional routing substrate to facilitate the implementation of new network control approaches. However, the obvious disadvantage of overlay networks is that they depend on several aspects (e.g. selected overlay nodes) to achieve the required performance. Moreover, such networks can be criticized for compounding the complexity of existing networks due to the additional virtual layers.
On the other hand, the recent software-defined networking (SDN) paradigm (Feamster et al. 2014) offers the potential to program the network and thus facilitates the introduction of automatic and adaptive control approaches by separating hardware (data plane) and software (control plane), enabling their independent evolution. SDN aims for the centralization of the network control, offering improved visibility and better flexibility to manage the network and optimize its performance. When compared to the overlay network alternative, SDN has the ability to control the entire network, not just a selected set of nodes, and use a public network for transporting data. Moreover, SDN spares network operators the tedious task of temporarily creating the appropriate overlay network for a specific use case. Instead, it provides an inherent programmatic framework for hosting control and security applications that are developed in a centralized way while taking into consideration the IoT requirements (Li et al. 2016; Ren et al. 2019; Montaño et al. 2021) to guarantee the user’s QoE.
I.2. Problem statement and motivations
Despite the considerable interest in SDN, its deployment in the industrial context is still in the relatively early stages. Indeed, there may be a long road ahead before technology matures and standardization efforts pay off so that the full potential of SDN can be achieved.
In fact, along with the hype and excitement, there have been several concerns and questions regarding the widespread adoption of SDN networks. For instance, research studies on the feasibility of the SDN deployment have revealed that the physical centralization of the control plane in a single programmable software component, called the controller, is constrained by several limitations such as the issues of scalability, availability and reliability. Gradually, it became unavoidable to think about the control plane as a distributed system (Canini et al. 2014; Sarmiento et al. 2021), in which several SDN controllers are in charge of handling the whole network while maintaining a logically centralized network view.
In this respect, networking communities argued about the best way to implement distributed SDN architectures while taking into account the new challenges brought by such distributed systems. Consequently, several SDN solutions have been explored and many SDN projects have emerged. Each proposed SDN controller platform adopted a specific architectural design approach based on various factors such as the aspects of interest, the performance goals, the deployed SDN use case and also the trade-offs involved in the presence of multiple conflicting and competing challenges.
Here, we underline the importance of conducting a thorough analysis of the proposed SDN solutions to envisage the potential trends that may drive future research in the area. In particular, we place a special focus on distributed SDN control designs with the aim of solving some of the major challenges encountered in the decentralization of the SDN control planes in the context of large-scale deployments.
The main motivations for this work are as follows:
– ensuring a thorough understanding of existing state-of-the-art distributed SDN controller platforms, and developing a critical awareness of the ongoing and future key research and operational challenges facing the design and deployment of such platforms;
– proposing novel approaches for decentralizing the SDN control plane in large-scale networks. Such a decentralized SDN control plane should be efficient (i.e. scalable, high-performance and robust) as it should meet the SDN controller application requirements (e.g. scalability, reliability and consistency);
– paving the way for the emergence of a new common standard for the distributed SDN control plane. This standard should also ensure the inter-controller communication between different vendor-specific controller technologies (i.e. the interoperability challenge).
I.3. Main contributions
In this section, we outline the main contributions of this work. More specifically, we propose novel approaches for decentralizing the software-defined networking (SDN) control plane in large-scale networks, while tackling some of the major associated challenges:
1) scalability and reliability aware strategies for the placement of distributed SDN controllers at scale using different types of multi-criteria optimization algorithms (see Chapter 3); 2) an adaptive and continuous consistency model for the distributed SDN
controllers: a novel anti-entropy reconciliation mechanism for applications (with eventual consistency needs) on top of the ONOS controllers (see Chapter 4);
3) an adaptive and continuous consistency model for the distributed SDN controllers: a novel Quorum-based replication strategy for applications (with eventual consistency needs) on top of the ONOS controllers (see Chapter 5).
Additionally, given the lack of available literature on the subject of decentralized SDN control and given its relevance today, our work also provides the following:
– a survey on distributed control in SDN: an overview and taxonomy of current SDN controller platforms (i.e. a physical classification and a logical classification) (see Chapter 1);
– a thorough analysis of the challenges encountered by the discussed state-of-the-art distributed SDN controller platforms, and the different approaches adopted for solving these challenges (see Chapter 2).
I.4. Structure of the book
Chapter 1 presents a survey on SDN with a special focus on distributed SDN control solutions. In addition to explaining the fundamental elements of the SDN architecture, this chapter proposes a taxonomy of the most prominent state-of-the-art SDN controller platforms by classifying them in two different ways: a physical classification and a logical classification.
Chapter 2 provides a thorough analysis of the major open challenges faced by the existing distributed SDN controller platforms discussed in the previous chapter. These challenges include the issues of scalability, reliability, consistency and interoperability of the SDN control plane. Furthermore, this chapter explores potential approaches to tackle these challenges for an optimal SDN deployment and provides some useful insights into the emerging and future trends in the design of efficient distributed SDN control planes.
Chapter 3 addresses the distributed SDN control problem by tackling the SDN controller placement problem in large-scale IoT-like networks. It puts forward novel scalability and reliability aware controller placement strategies that deal with several aspects of the controller placement optimization problem with respect to multiple reliability and performance criteria and according to different uses and contexts. These strategies use two different types of heuristic-based algorithms: a clustering algorithm based on PAM and a modified genetic algorithm called NSGA-II. These multi-criteria algorithms are compared in terms of computation time, as well as the quality of final controller placement configurations.
Chapter 4 focuses on the distributed SDN control problem by tackling the knowledge sharing problem between the distributed SDN controllers. It proposes an adaptive multi-level consistency model following the concept of continuous consistency for the distributed SDN controllers. This approach is implemented for a source routing application on top of the open-source ONOS controllers. It involves turning ONOS’s eventual consistency model into an adaptive consistency model using the anti-entropy reconciliation period as a control knob for an adaptive fine-grained tuning of consistency levels. Our proposed consistency strategy is aimed at ensuring the application’s continuous consistency requirements (i.e. numerical error bounds), as specified in the given application service-level agreement (SLA). Its purpose is also to minimize the anti-entropy reconciliation overhead as compared to ONOS’s static consistency scheme at scale.
Chapter 5 further addresses the knowledge sharing problem in the distributed SDN control by proposing an adaptive and continuous consistency model for the distributed ONOS controllers. The approach is implemented for a CDN-like application on top of ONOS. It consists of changing ONOS’s eventual consistency model to an adaptive consistency model by turning ONOS’s optimistic replication technique into a more scalable replication strategy following Quorum-replicated consistency. The main focus is placed on improving ONOS’s replication mechanism: it uses the read and write Quorum parameters as adjustable control knobs for a fine-grained consistency tuning rather than relying on anti-entropy reconciliation mechanisms (Chapter 4). The main objective is to find optimal partial Quorum configurations at runtime that achieve, under changing network and workload conditions, balanced trade-offs between the application’s continuous performance (latency) and consistency (staleness) requirements. These real-time trade-offs should provide minimal application inter-controller overhead while satisfying the application-defined thresholds specified in the given application SLA.
Conclusions and Perspectives is the final chapter of the book and gives an insight into our ongoing and future work, and perspectives in the area of distributed SDN control.
1Toward a Decentralized SDN Control Architecture: Overview and Taxonomy
1.1. Introduction
In contrast to the decentralized control logic that underpins the construction of the Internet as a complex bundle of box-centric protocols and vertically integrated solutions, the software-defined networking (SDN) paradigm advocates the separation of the control logic from hardware and its centralization in software-based controllers. These key tenets offer new opportunities to introduce innovative applications and incorporate automatic and adaptive control aspects, thereby easing network management and guaranteeing the user’s QoE.
However, despite the interest surrounding SDN, adoption raises many challenges, including the scalability and reliability issues of centralized designs that can be addressed with the physical decentralization of the control plane. However, such physically distributed but logically centralized systems bring an additional set of challenges.
This chapter presents a survey on SDN with a special focus on distributed SDN control. In section 1.2, we start by describing the promises and solutions offered by SDN compared to conventional networking. We also elaborate on the fundamental elements of the SDN architecture.
Then, we expand our knowledge of the different approaches to SDN by exploring the wide variety of existing SDN controller platforms. In doing so, we intend to place a special emphasis on distributed SDN solutions and classify them in two different ways. In section 1.3, we propose a physical classification of state-of-the-art SDN control plane architectures into centralized and distributed (flat or hierarchical) in order to highlight the SDN performance, scalability and reliability challenges. In section 1.4, we put forward a logical classification of distributed SDN control plane architectures, logically centralized and logically distributed, while tackling the associated state consistency and knowledge dissemination issues.
1.2. Software-defined networking: a centralized control architecture
1.2.1. Conventional networking and the SDN paradigm
Over the past few years, the need for a new approach to networking has been expressed to overcome the many issues associated with current networks. In particular, the main vision of the SDN approach is to simplify networking operations, optimize network management and introduce innovation and flexibility compared to legacy networking architectures.
In this context, and in line with the vision of Kim and Feamster (2013), four key reasons for the problems encountered in the management of existing networks can be identified:
1) Complex and low-level network configuration: network configuration is a complex distributed task in which each device is typically configured in a low-level vendor-specific manner. Additionally, the rapid growth of the network, together with the changing networking conditions, have resulted in network operators constantly performing manual changes to network configurations, thereby compounding the complexity of the configuration process and introducing additional configuration errors.
2) Dynamic network state: networks are growing dramatically in size, complexity and consequently in dynamicity. Furthermore, with the rise of mobile computing trends as well as the advent of network virtualization (Bari et al. 2013; Alam et al. 2020) and cloud computing (Zhang et al. 2010; Sharkh et al. 2013; Shamshirband et al. 2020), the networking environment becomes even more dynamic as hosts are continually moving, arriving and departing due to the flexibility offered by VM migration, thus making traffic patterns and network conditions change in a more rapid and significant way.
