White Paper based on the IEEE Workshop SDN4FNS (Trento 11th – 13th November, 2013)

29th January 2014

Editor and contact: antonio.manzalini@telecomitalia.it

Software-Defined Networks
for Future Networks and Services
Main Technical Challenges and Business Implications

Antonio Manzalini, Telecom Italia, Italy Walter Cerroni, Franco Callegati, University of
Roberto Saracco, EIT ICT Labs, Italy Bologna, Italy
Cagatay Buyukkoc, AT&T Labs, USA Giovanni Schembra, Vincenzo Riccobene,
Prosper Chemouil, Orange, France University of Catania, Italy
Sławomir Kukliński, Orange Polska, Poland Carmen Mas Machuca, Technische Universität
Andreas Gladisch, Deutsche Telekom, Germany München, Germany
Masaki Fukui, Wenyu Shen, NTT, Japan Alex Galis, University College London, U.K.
Eliezer Dekel, IBM, Israel Julius Mueller, FhG FOKUS, Germany
David Soldani, Huawei, Germany
Mehmet Ulema, Manhattan College, USA

Abstract — In 2013, the IEEE Future Directions Committee automated operation processes (e.g. configuration of networks
(FDC) formed an SDN work group to explore the amount of and services systems and equipment) could limit human
interest in forming an IEEE Software-Defined Network (SDN) intervention, reducing also wrong operations; on the other
Community. To this end, a Workshop on “SDN for Future hand, concerning the CAPEX, a flexible and optimal
Networks and Services” (SDN4FNS’13) was organized in Trento, provisioning of network functions and services could reduce
Italy (Nov. 11th-13th 2013). Following the results of the workshop, systems and equipment costs and allows postponing
in this paper, we have further analyzed scenarios, prior-art, state investments.
of standardization, and further discussed the main technical
challenges and socio-economic aspects of SDN and virtualization Emerging paradigms such as Software-Defined Networks
in future networks and services. A number of research and (SDN) [1] and virtualization, for instance Network Functions
development directions have been identified in this white paper, Virtualization (NFV) [2], if properly designed and deployed,
along with a comprehensive analysis of the technical feasibility could help in fulfilling the above mentioned requirements: as a
and business availability of those fundamental technologies. A matter of fact, a deeper integration of networks and IT (e.g.
radical industry transition towards the “economy of information Cloud) domains, and the related Operations (now mainly
through softwarization” is expected in the near future. carried out separately), could allow huge savings, and the
acquisition of greater flexibility in services provisioning.
Keywords—Software-Defined Networks, SDN, Network
Functions Virtualization, NFV, Virtualization, Edge, In particular, according to a widely accepted definition,
Programmability, Cloud Computing. SDN concerns the decoupling of the software-based control
plane from the hardware-based data plane (e.g., packets
I. INTRODUCTION forwarding) of networking and switching pieces of equipment;
The ossification of Internet and Telecom networks is in principle, this would allow moving control logic (and states)
creating several difficulties for Service Providers (SP) and to logically centralized controllers. NFV is one of the most
Network Operators (NO) to develop and deploy, flexibly, any innovative expressions of virtualization, and specifically it
innovative network functionalities, services and management implies the virtualization of network functions and services
policies, which are essential to benefit from the increasing that could run on general purpose hardware; this would allow
dynamicity of the ICT markets. Launching new services, for dynamically placing and moving said functions in various
example, is still time-consuming and requires expensive locations of the networks and services infrastructures.
efforts: this is preventing any rapid roll-out of new businesses
Importantly, it should be noted that different developments
models and opportunities. A first requirement is thus making
and deployment scenarios of SDN and NFV could be
the innovation cycles of networks and services features faster
envisioned, depending on network segments (e.g., core or
and simpler. Moreover there is a need, for both SP and NO of
edge) and, consequently, on time horizon (e.g., medium-long
reducing OPerational EXpenditures (OPEX) and CAPital
term or short term). These deployment scenarios, in turn, could
EXpenditures (CAPEX): concerning the OPEX, for instance,

1
White Paper based on the IEEE Workshop SDN4FNS (Trento 11th – 13th November, 2013)
29th January 2014
Editor and contact: antonio.manzalini@telecomitalia.it

require different amounts of investments (e.g., less investments programming capabilities of other abstractions and even of
in the edge), expected revenues and impacts on the Operations physical or virtual resources.
of NO and SP. Nevertheless, we are witnessing an overall trend
of networks and services “softwarization” which is unstoppable
as it is mainly due and driven by the continuous IT technology
evolution and cost reductions.
In 2013, the IEEE Future Directions Committee (FDC)
formed an SDN work group to investigate and determine the
amount of interest in assembling an IEEE SDN community. In
order to achieve this goal, a Workshop on Software-Defined
Networks for Future Networks and Services (SDN4FNS’13)
was held in Trento, Italy, on Nov 11-13th 2013 [3]. Around 70
international experts from industry (e.g., NO, SP, Technology
Providers, etc), SMEs and academia gathered together to
present visions, pieces of results and to discuss and draw the
key challenges about the potential adoption of SDN and
virtualization. Three keynote speeches, thirty presentations and
three panel discussions provided a comprehensive overview of
current and future research and development work, socio-
economic aspects and impacts.
In this white paper, leveraging the results of the workshop, Figure 1 – SDN architectural model (Source: [4])
the authors have further analyzed scenarios, prior-art, state of This evolution aims to cope with some specific problems
standardization, and reviewed the main technical challenges and requirements such as scalability, performance,
and socio-economic aspects of SDN and virtualization in future management, robustness or ability to adapt to multi-operator
networks and services. environment in which multiple SDN controllers or SDN
The rest of the paper is organized as follows. Section II domains have to co-exist. As high level functional definition
reports some of the widely adopted terms and definitions of may determine the elaboration of different systems and
SDN and NFV, in order to align with the nomenclature; software architectures (with the related implementation for
Section III summarizes the main outcomes of the IEEE different networks’ areas) there are, currently, several
Workshop SDN4FNS. Section IV describes some thought- viewpoints in considering and assessing future deployments of
provoking scenarios and use-cases. Section V provides an SDN.
analysis of some of the more debated conceptual and functional It has also to be noted that in the past other attempt have
models for SDN and related challenges. A brief status report on been made to decouple software from hardware and to achieve
the main standardization activities is presented in Section VI. network programmability: an overview of such a concept was
Conclusions and next steps are drawn in Section VII. presented in [5] [6]. At present, several latest attempts, such as
II. TERMS AND DEFINITIONS PCE [7], ForCES [8], or i2rs [9] are also denoted as in line
with SDN principles and an effort to consolidate them under
According to the Open Networking Foundation (ONF)[1] the SDN umbrella can be perceived. Such integration would
[4] a SDN is a network where the control (software-based) and be justified by some similarities of these concepts, especially
data forwarding planes (hardware-based) are decoupled, so the programmability of some operations and the separation of
that, in principle, the network infrastructure could be control and data planes. However some concerns related to the
abstracted from functions and business applications [4]. The overall complexity and cost of such integrated solution could
proposed concept and architecture for SDN is illustrated in arise. Moreover between these concepts some substantial
Figure 1. differences also exist, as “distributed i2rs driven by real-time
In a SDN, for example, the decisions concerning flows management” versus “centralized SDN with fundamental role
switching and engineering are taken by a so-called SDN of control plane”, which make the integration of those
controller which interacts with SDN switches via the troublesome.
OpenFlow protocol (as defined by ONF). The protocol It should be noted that SDN should not be confused with
procedures are mostly related to data flows, queues and ports, NFV, which is about the virtualization of some network
while applications and functions, running on top of the functions that could be executed on standard of the shelf
controller, (not defined by ONF) may be developed by other hardware.
implementers.
In principle, this approach could allow introducing in
Nevertheless, above SDN definition is still evolving in network operations, those features which are today normally
order to enrich said networks with functionality, for example, carried out in Data Centers (DC), such as dynamic allocation,
to provide support for mobility, to add control and

2
White Paper based on the IEEE Workshop SDN4FNS (Trento 11th – 13th November, 2013)
29th January 2014
Editor and contact: antonio.manzalini@telecomitalia.it

migration and cloning of virtual resources and functions (e.g. IEEE Workshop SDN4FNS. The Workshop was organized in
for server consolidation, load balancing, etc.). three tracks: 1) Telecom and Internet SDN Scenarios, 2)
In the IT context, virtualization is already well known and Hardware and Software 3) Regulatory, Biz, Techno-Economic
widely deployed in Data Centers for enabling the execution of Sustainability. The main takeaways are briefly summarized in
multiple isolated instances of a software entity on top of a the following sub-sections.
single physical server. IT Virtualization has several A. Track 1: Telecom and Internet SDN Scenarios
advantages, for example it increases resource utilization and
There an emerging understanding that the disruptive
improves state encapsulation.
potential of SDN and virtualization could provide a number of
These principles have not been fully extended to networks. new opportunities to NOs, SPs, software developers and
Network virtualization already exists in virtual private equipment vendors. However, all these new potential
networks (VPNs) which generally use the multi-protocol label opportunities carry also a new set of challenges to cope with
switching (MPLS) technology, operating on the link level this transformation. State-of-the-art SDN and virtualization
layer. Another form of virtualization is to segment the implementations have already shown many issues to be
physical local area networks into virtual local area networks addressed, and a significant number of those are related to
(VLANs). An overlay network is yet another form of network security aspects [10]. In general, SDN could be seen as a
virtualization which is typically implemented in the paradigm helping in overcoming ossification of current
application layer, though various implementations at lower layering and protocols stacks. For example, [11] argued that,
layers of the network stack are also being used. Extension of in the future, the TCP/IP layering itself may represent just one
IT virtualization principles to network equipment (such as of the ways to deployment of a truly flexible software-defined
routers and switches) could determine several advantages as networking environment.
well, i.e. optimizing the use of physical resources and Several interesting SDN applications scenarios have been
allowing a deeper integration of IT and network resources. presented and discussed, for both wired and wireless-radio and
NFV could bring the ability to co-locate multiple instances networks. For instance, SDN can be envisioned as a potential
of network functions on the same hardware – each running in solution for efficient and scalable implementation of control
one (or more) different Virtual Machine (VM). This could functions in extremely dense and heterogeneous wireless
provide NO and SP with the ability to dynamically instantiate, networks [12]. According to this vision, a novel network
activate, and re-allocate resources and functions, and even architecture, accounting for MAC control and Mobility
program those according to needs and policies. SDN and NFV Management, was proposed as part of the activities of the EU-
could be seen as mutually beneficial, but they are not funded FP7 CROWD project [13].
dependent on each other: e.g., network functions can be SDN and virtualization could also enable a flexible and
virtualized and deployed without an SDN being required and more efficient implementation of the LTE Evolved Packet
vice-versa. An example of functional architecture from ETSI Core by splitting its main functions between a virtualized
is depicted in Figure 2. cloud environment and an SDN-based transport infrastructure
[14]. SDN could provide key functions such as load balancing
over different wireless technologies and related flow
admission control in heterogeneous 5G mobile networks [15].
Finally, the issues to enable switch mobility, while
enforcing robustness against network attacks were addressed
in [16], by introducing an enhanced transport layer, based on
standard IP mobility techniques for OpenFlow controller-to-
node communications.
One of the key challenges that could be faced with SDN is
the optimization of (virtual) resource allocation and usage.
Relevant case studies include: inter-data center
communication for service brokerage over large scale
distributed and heterogeneous cloud environments [17];
performance evaluation of virtual network functions migration
across cloud-based edge networks [18]; dynamic traffic
engineering and adaptive network design to efficiently map
Figure 2 – Example of functional model for NFV (Source: ETSI). logical/virtual topologies on physical network infrastructures
[19], [20].
III. SUMMARY OF THE SDN4FNS’13 WORKSHOP
This section intends to provide a brief overview of the most
important topics covered by papers and discussions during the

3
White Paper based on the IEEE Workshop SDN4FNS (Trento 11th – 13th November, 2013)
29th January 2014
Editor and contact: antonio.manzalini@telecomitalia.it

B. Track 2: Hardware and Software telecommunications environment”, where most of the

Hardware performance advances and costs reductions are intelligence is moved to the network edges, eventually
creating the mass market favorable conditions to a large evolving into a fully interconnected Internet of Everything
adoption of the software-defined principles and virtualization. [29]. It is of course intended that virtualizing networks at any
Nevertheless this will create an high level of complexity in scale would require some level of orchestration: technical
future networks and services platforms. As such, SDN and challenges and potential architectural approaches to edge-to-
NFV will require to enhance current management systems edge virtualization, abstraction, control, and optimization of
with new capabilities, for example concerning the heterogeneous transport networks with packet- and circuit-
orchestration of virtualized functions and resources. switched technologies were discussed in [30].
Moreover, in order to tame the growing complexity and the Test-beds and proof-of-concepts of virtualized network
dynamism proper levels of software abstractions should be functions can provide an insight into the major feasibility
timely introduced, simplifying also the views of the issues of network virtualization and a viability check of NFV
architecture (e.g. slicing). Elasticity and flexibility could be performance. Good examples include: EmPOWER
achieved through systems and methods for VMs placement, experimental test-bed for wireless network virtualization [31];
move and traffic routing between VMs (i.e., solving double an OpenFlow-based prototype of routing function
constrained optimization problems in almost real time). virtualization [32]; CONTENT project approach [33] to
In general, there are several efforts in this avenue, which infrastructure virtualization over heterogeneous wireless and
appears to be very strategic. optical networks. Last, but not least, the internal structure of a
Inter-domain SDN controller integration could be enforced network node capable of function virtualization, which is also
by means of compatible east-west bound interfaces, based on a critical aspect of SDN, may be implemented through a plug-
either service-oriented architectures [21] or inter-platform in interface architecture as presented in [35].
signaling for distributed flow processing [22]. C. Track 3: Regulatory, Biz, Techno-Economic Sustainability
Most of the ongoing discussions about SDN typically focus
Although SDN is still considered to be in its infancy,
on control plane aspects. However, many unresolved issues
medium to long-term visions on how it will possibly evolve in
arise when considering the programmability of the data plane,
the future can set the grounds for strategic research and
such as the relative importance of data plane vs. control plane
investigations on the techno-economic sustainability.
services, the relevance of the underlying hardware platform,
Dynamic Network Service Chaining is, for example, one
and the need for standardized northbound and southbound
key research topic showing several challenges, considering the
interfaces in the data plane [33].
many aspects to be dealt with during the typical lifecycle of a
Another very important aspect of SDN concerns service
network service [36]. This is in alignment with the concept of
provisioning with guaranteed QoS: leveraging network
“Forwarding Graph”, which is used sometimes in preference
programmability features provided by SDN, specific platforms
to “Service Chain”, in order to account for the fact that end-to-
addressing the QoS monitoring and enforcement issues were
end forwarding within virtualized overlay service networks is
developed with managed [23] or autonomic [24] approaches.
not exclusively a one dimensional chain: instead they may,
Then, considering the increasing deployment of OpenFlow-
and often will, have branches.
enabled equipment, a smart platform capable of detecting
Form the techno-economic perspective, a wider and wider
OpenFlow rules interactions and determining possible
introduction of “software” in networks and services
inconsistencies could be extremely useful as a sort of
infrastructures [37] will accelerate the pace of innovation (as it
“debugger” for OpenFlow application development [25].
is doing continuously in the IT domain) and will reduce
SDN can be considered as a powerful enabler for many
operational costs (e.g., through optimizations exploiting big
emerging networking paradigms. It is indeed vital to deploy
data and automation). This trend will move “competition”
inter-cloud communication services over existing network
from hardware to software, lowering the threshold for several
infrastructures in a scalable and feasible way, as proposed
Players to enter into the ICT – Telco arena. It is likely that the
within the European Future Internet initiative [26]. It is also
so-called “softwarization” will to enter more and more in all
the key technology that can foster the implementation of
socio-economic processes.
multipath inter-data center communication architectures,
Finally, the vision of a future network where most of the
building on emerging standard protocols [27]. Finally, SDN
intelligence resides at the network edge was brought one step
may ease the deployment of Information-Centric Networking
further. This is especially true, considering the vision that
(ICN) in existing IP networks by effectively decoupling
SDN and NFV solutions will be able to create a sort of
forwarding information from object names [28].
distributed communication “fabric” around the users, covering
As mentioned, virtualization is one of the most important
all network equipment available at the edge (including any
technologies intertwined the growing interest in SDN: the
type of hyper connected devices, such us smartphones, robots,
capability of virtualizing practically any network function and
cars, drones, etc.), which can offer huge processing and
service and SDN principles are mutually beneficial. This is
also the rationale behind the so-called “hyperconnected

4
White Paper based on the IEEE Workshop SDN4FNS (Trento 11th – 13th November, 2013)
29th January 2014
Editor and contact: antonio.manzalini@telecomitalia.it

storage capabilities to execute and consume practically any A. Examples of core scenarios
virtualized function and service [38]. Core networks scenarios typically consider SDN as a
D. Main Takeaways paradigm providing incremental improvements (in terms of
flexibility, programmability, etc.) of current networking
Socio-economic drivers and technology progresses (with
concepts; practically, it is recognized that the concept of the
their down-spiraling costs) are steering the evolution of current
networks towards a highly dynamic and flexible environment separation of hardware from (control) software is not really
of virtual resources, interconnected by virtual links that are set new, but the point is that the decoupling is made possible
up and torn down to serve multiple applications. today thanks to the hardware technology advances.
In the context of these core scenarios, Dynamic Network
SDN and NFV are likely to represent a first step to this Service Chaining is one of the most mentioned classes of use
direction. In general, in the future, a growing number of cases, where IT and networks resources are integrated:
industries and small medium enterprises will rely more and network services are provided by “chaining” the executions of
more on “digitalization” and “software”. In this sense SDN and several service components.
NFV could be seen as powerful enablers to create and develop
SDN is often assessed as an opportunity reducing CAPEX
new ecosystems capable of aggregating and driving
and OPEX costs. As previously mentioned, savings may
investments, even outside the traditional Telco-ICT contexts.
This requires embracing this crucial industrial transformation derive from centralizing and, above all, automating processes
from a broader perspective. and postponing investments through optimized usage of
resources (provided that carriers’ class performances are still
This transformation is an unstoppable trend, because of the achieved by the adoption of general purpose hardware).
continuous technology evolution and cost reductions by On the other hand a deeper integration of networks and IT
lowering operating costs though simplified hardware, software, (e.g. Cloud) domains, and the related Operations requires also
and management will definitely enable new economic a deep “change of culture” in NOs and SPs, and maybe the
paradigms. For example, SDN and NFV are claiming network development of new skills for mastering “software”. This
cost reductions, due to the adoption (and consolidation) of might require some time, also to define new models of
standard hardware, capable of running virtualized network
business sustainability.
functions/applications. Nevertheless, we found wide consensus
Seamless integration with legacy equipment and the related
on the fact that this model will be successful only if SDN and
NFV are really based on open source software solutions. management systems might represent other critical issue,
(Closed software solutions, in fact, would move costs from mainly demanded to the standardization of interfaces, which
hardware CAPEX to software OPEX, probably erasing the might be delayed. In summary, it remains to be seen whether
claimed advantages of SDN and NFV in terms of cost saving.) SDN exploitation in core scenarios will be really followed,
and what the time horizon will be.
The analysis is obviously more complex as it should take
into account to which level the performance should be B. Examples of edge scenarios
managed as well, and how all these virtual network functions Edge scenarios concerns the exploitation of SDN principles
should be efficiently orchestrated. Moreover, it was argued by for creating very dynamic virtual networks out of a variety of
several people that this r-evolution would happen at the edge, aggregation nodes, devices, elements located at the edge of
first, as it would require fewer investments, it would scale current networks, up to around Users. Some of these elements
much more gracefully, and it would lead to immediate usually are not considered yet as network nodes: for example,
revenues. Evolution in Users’ devices, terminals, Customer cars, robots, drones, any Users’ devices, smart things with
Premises Equipment (CPE) or aggregation edge nodes is much
embedded communications, etc. In other words, this is about
faster than in core or WAN equipment, especially due to the
impact of SDN and NFV on traditional Operations Support developing a “fabric” made of an enormous number of nodes
Systems (OSS)/Business Support Systems (BSS). and elements aggregated in an application driven way, as
depicted in Figure 3.
In the past, the term “fabric” has been used to refer to a
Next sections will provide further analysis of scenarios, distributed computing system consisting of loosely-coupled
prior-art, state of standardization, and further discussions on storage, networking and processing capabilities interconnected
technical challenges and socio-economic aspects. by high-bandwidth links. It has also been used for describing
flat, simple intra data center networks optimized for horizontal
IV. SCENARIOS AND HIGH LEVEL REQUIREMENTS traffic flows, mainly based on a concept of “server-to-server
In this section, some scenarios (wired and radio, core and connectivity”. Based on these previous meanings, in this
access/distributions, edge) are provided from the points of context, the term fabrics is extended to indicate the edges of
view of NOs and SPs, as well as from other players’ angle, the metro networks, becoming like as distributed Data Centers
such as OTTs, Enterprise Networks Providers, Consumer consisting of loosely coupled processing and storage resources
Electronics Providers, etc. interconnected by pervasive high speed wired and wireless
links.

5
White Paper based on the IEEE Workshop SDN4FNS (Trento 11th – 13th November, 2013)
29th January 2014
Editor and contact: antonio.manzalini@telecomitalia.it

An example of use case is that of a SP that may want to appropriately addressed with the new supported and deployed
provide end-to-end ICT services to users who are attached to capabilities.
edge networks, even if belonging to different infrastructure This transformation is a big task, and the existing paradigm
providers or NOs. This could be achieved by operating an is not quite ready in achieving this objective, but the
overlay service platform capable of chaining, managing and capabilities at hand should allow the research community to
orchestrating virtualized resources and functions made work in this research direction.
available by the different edge networks. In the IEEE SDN4FNS Workshop, several contributions
were given towards this goal. Yet, some other key areas would
need to be addressed in the upcoming period by the research
community, including, but not limited to:
(a) Development of an end-to-end SDN framework for
Collaboration (Coordination, Cooperation, Communication
Orchestration) taking into account the current and new actors
(prosumers/functions/applications), in order to agree and
achieve global goals (i.e., utility maximization, resource
optimization QoE, Policy, Security…).There is an ongoing
work with this broad aim, but new capabilities are definitely
needed. As an example, Identity, Security, Mobility and QoS
Management are key high level objectives; providing these in a
modular way, across the network/s and all layers need
corresponding abstractions from all, as well as new concepts
and architectures of Access and Non Access Strata. In the
Figure 3 – Example of scenario where the “edge” is becoming a security context, perimeter defenses do not usually work,
“fabric” of resources to execute networks functions and store data. instead of firewalls at the perimeter, using internal modular
firewalls to define enclaves within the network, and using
Another use case is that where Data and ICT services (seen secure protocols [e.g., Secure Sockets Layer (SSL) or
as apps executed via chains of VMs) will follow the users Transport Layer Security (TLS)] and service validation at their
when they are moving from one network attachment point to inputs to prevent attacks might be a better approach. Similarly,
another one, even across different edge networks. E2E QoS and mobility objectives should be apportioned and
Management and orchestration capabilities should allow relegated to the corresponding domains and layers to provide
this “follow-me” service whereby personal data and ICT the necessary carrier grade performance.
services will be moved seamlessly with little or no impact on (b) Development of an SDN/NFV framework for Service
the Quality of Experience (QoE) of users. Moreover, data and Providers. The existing paradigm would need to be extended
services associated to users can be even federated to build beyond data centers and enterprise networks. This would
distributed virtual data center at the edge (ideal for example require longer geographic reach and diversity. All different
for universities, enterprises, etc) at costs which are a small and access, core, and link technologies (e.g., RAN, cellular, Wi-Fi,
a fraction of traditional clouds. optical, etc.) should be brought into the framework as well, to
create an end-to-end infrastructure, where these domains could
C. Other Expectations and Requirements be offered as a Service. It also requires fundamental rethinking
In general, NOs and SPs expectations related to the on how the current network may evolve (typically, delivery of
SDN/NFV frameworks are much broader than the currently new network-based services takes weeks and in some cases
addressed Telco’s requirements. One far-reaching months) and how new services may be provisioned using a
circumstance is that the carriers have made huge deployment much more dynamic business process, characterized by the
investments in legacy networks. Hence, many of the novel configuration and management of all virtual resources to
SDN capabilities need to enable the transformation of the provide even network services, as previously introduced.
current nodes to the new infrastructure vision the research Dynamic network provisioning and configuration needs to be
community is creating. coordinated and orchestrated, in order to efficiently direct the
assignment, creation and configuration of virtual resources,
This is a fundamental requirement: developing a radical
and, especially, to satisfy customer requests and manage
simplification on how carrier networks are designed, built,
customer experience per SLA. Network Orchestration will play
deployed, operated and managed. End users should have the a key role in implementing these requirements at various levels
desired QoE, new services should be easily introduced, and and will be responsible for coordinating the delivery of the
resources should be optimally used for meeting all resources and attributes of a customer request into an
performance requirements of the new complex applications. operational service that embodies the runtime requirements of
The networks will transform into a new “market” where all that request. Network Services Orchestration is expected to
actors’ (users, applications, networks) requirements are provide highly reliable and scalable capabilities to enable these

6
White Paper based on the IEEE Workshop SDN4FNS (Trento 11th – 13th November, 2013)
29th January 2014
Editor and contact: antonio.manzalini@telecomitalia.it

expectations. SDN would require a new set of capabilities that fault management, configuration management, accounting
would work with the existing networks. Legacy networks management, performance management and security
should gracefully evolve to the future network through an management.
evolution. Key control and management functions should
interoperate with existing frameworks. This is a key The role of management (including fault, configuration,
requirement for SP that cannot afford a clean slate approach. accounting, performance and security areas) is especially
critical in large, production networks. Typically, in such
Looking at Flexible Service Control, the virtualization networks, the operators interact with the network via the
technologies for telecommunication infrastructures enable management system. After network deployment, the most
more opportunities in terms of flexibility, scalability and typical network operator’s actions are usually related to
efficiency for on demand driven dynamic network network reconfiguration, as a result of network enlargement or
dimensioning and traffic engineering. For instance, the flexible topology changes. It is commonly agreed that in order to cope
control of radio (Cloud-RAN/evolved-NodeBs) and core with future demands and the ever-growing number of managed
network (Evolved Packet Core, IP Multimedia Subsystem, devices, the network management should be automated.
Service Delivery Platform) resources enables novel lifecycle
management and control possibilities. Hence the network availability is expected to be improved
and OPEX related to the network management significantly
The main advantages for virtualization, with respect to reduced. This type of management lies on performing selected
today's telecommunications architectures are: management operations in “real-time”, which is often referred
to as autonomic.
(a) Data path optimization for signaling and user data plane.
The influence of SDN and NFV on the control and data plane The management of SDN can reuse some well-known
within Telco architectures allows a precise and particular network management concepts and it should incorporate the
optimization in overloaded parts of the network. Energy latest trends in network and service management as well. So far
consumption and carbon footprint are playing an important it is not the case. In facts, the management issues of SDN have
role in the ICT in general and telecommunication in particular, been as of today ignored. This situation can be explained by
since clouds, data center, network elements and base stations the experimental status of deployed networks.
are responsible for a large amount of the total energy In the current SDN approaches, the usage of existing IP
consumption. Therefore, the influence of virtualization on network management systems combined with OpenFlow is
Telco networks, as a method for enabling optimized network often assumed. The analysis of the OpenFlow protocol leads to
design and traffic engineering instead of over-provisioning the the conclusion that it lacks primitives that are able to cope with
Telco network, arise. High performance, flexible, elastic and network management operations. In order to support
demand-driven solutions are addressed in European research management functions, ONF has defined the OF-Config
projects (MCN [39], iJOIN [64], etc). protocol. This protocol is used for configuration of links
between the OpenFlow switches and the controller. In the short
(b) Network aware service enablement / QoS on demand. term, the SDN management operations could be supported by
Flexible QoS control is one of the key denominators in the existing network management protocols (SNMP [63],
emerging future telco environments. Features such as user NETCONF [62]).
demanded QoS, QoS level per subscriber base, and The openness of OpenFlow enables the implementation of
differentiation between multiple Mobile Virtual Network some network management functions, but the lack of
Operator (MVNO) on top a physical network infrastructure all standardization of a management interface makes impossible to
require flexible QoS control to ensure connectivity for use third-party management solutions in a way as they are used
dedicated service types. These service types might be in legacy management systems.
emergency calls, public safety services or critical SDN-based networks have some specific properties. Their
infrastructures such as smart metering. owner or operator has not only the ability to manage them but
V. CONCEPTUAL AND TECHNICAL CHALLENGES also the possibility of defining their functionality.
This Section describes a list of the most important technical There are several management issues that appear to be
challenges for the development and deployment of SDN. specific to SDN only. Most of them are related to the critical
role of the controller, which has to perform most of its
A. Management and Orchestration operations in a real-time manner. It means that the performance
The fundamental feature of SDN is the decoupling of the of the controller has to be monitored, and the controller
network control plane from the network forwarding (data) processes should be categorized and handled according to their
plane. Such separation is well-known in telecommunication real-time requirements. In case of controller failure, a hot swap
networks. However, network architectures always include an operation is necessary. Controller programmability requires
additional plane, namely the management plane. Such plane special care. In order to fully use such capability, the network
and their functions are defined and standardized by the operator should be able to remotely add new controller
TeleManagement Forum and ITU-T. The functions include functions and to update the existing ones. Such functionality

7
White Paper based on the IEEE Workshop SDN4FNS (Trento 11th – 13th November, 2013)
29th January 2014
Editor and contact: antonio.manzalini@telecomitalia.it

means on-the-fly re-programmability of the controller. The Specific optimization techniques are required for both the
management operations should be able to start, stop and placement of the VMs into the physical networks and the
monitor each software module. Moreover abnormal behavior traffic routing between VMs (e.g., this means solving double
of such modules has to be monitored and cure taken. Last but constrained optimization problem in “almost” real time).
not least, the security of the controller platforms is critical and
has to be handled accordingly. This includes also the In the case of wireless infrastructures, other characteristics
and capabilities have to be considered, e.g. limited bandwidth,
authentication of newly added switches. It has to be combined
with the security of the management platform. processing capabilities, storage, energy (battery), type of
interfaces supported of the mobile nodes and mobility,
It is important to determine a border line between the SDN conflicting requirements. As the mapping of virtual to physical
network operations (implemented in the controller) and SDN- resources should be transparent to higher control layers,
based applications implemented in application servers. Such a mechanisms have to be developed that allow the seamless
separation can help in the definition of the management system hand-off between different wireless devices. Additionally,
role. In order to cope with scalability and complexity, separate algorithms will be identified that optimize the coverage of
platforms for the control and the management of SDN wireless radio connections to provide access to enough
networks are highly recommended. A proper information physical resources while avoiding unnecessary energy
exchange between these platforms is necessary, and it could be consumption.
provided by programmable interfaces.
In the case of wireline networks, optimized allocation will
It is an open issue whether the management will be fully involve a wide variety of resources available from the
centralized, or implemented by distributed functionalities. underlying wireline network, including communication,
There is however no doubt that the added complexity should execution and storage capabilities. The mapping will take into
not significantly increase the cost of the SDN nodes and that account the top-level service/operational requirements such as
the distribution of functionalities should still give a centralized the demanded QoS requirement and resilience capability to be
view of the network. embedded into the resulting virtual network. By addressing this
challenge virtual networks will be customized with optimally
1) Adaptive and Autonomic Methods and Systems allocated capabilities such as virtual nodes (with computing
This research challenge deals with the critical nature of and storage capabilities), virtual links and paths for specific
developing the methods, enablers and systems for autonomic networked services.
management functions applied not only to the physical
resources, but also virtual resources located inside the physical As an example, the Cross Layer Optimization paradigm as
/ virtual network. outlined in [65] maps top-level service/operational QoS
requirements against underlying network capabilities and thus
In this avenue, even if the requirement of unification of all enhances network management and traffic engineering
autonomic functions might be highly desirable, it will be sustainable. Individual characteristics of user profiles, service
almost impossible to achieve it, as the number of functions is descriptions and network topologies have to be optimized
expected to be large. New approaches probably should be given a connectivity with certain requirement on the transport
developed, allowing separated autonomic managers to system in order to guarantee QoS. Algorithms and concepts for
coordinate and resolve conflicts in their activities realizing a network-awareness for services through Cross
autonomously with minimal disruption. Layer Optimization are required in the scope of SDN and NFV
One example is that where management and control for dynamic service chaining and optimal network function
functions would be distributed and located or hosted in or close placement.
to the managed network and service elements, enabling control 3) Energy Management
of CAPEX and significant operational costs reduction for Finally, looking at the management aspects of energy
physical (i.e. OPEX) and virtual (i.e. VPEX) systems. These consumption, we may relate this challenge to the critical nature
may include adaptive re-allocation of virtual resources of developing mechanisms for energy-aware networking,
according to changing network conditions or service demands. including the optimization of energy consumption within the
Additionally, this challenge deals with the critical nature of limits of a single domain and/or multiple domains. Such
developing autonomous actions that provide network stability energy-aware capabilities would include the optimal dynamic
and optimizations in absence of higher-level control. distribution of VMs across the set of networks and servers and
2) Optimised allocation and orchestration of resources providing stabilization of the local networks in response to
It has been mentioned several times that one major variable electricity demand-response requirements.
challenge will be the capability of dynamically instantiating, 4) Security and Safety
orchestrating and migrating multiple VMs across the networks Security in terms of privacy, integrity has to be guaranteed
and services infrastructures. Orchestration, in this context, in a virtual environment [41] in at least a similar or a higher
means also lifecycle management of physical and virtual level than in today's networks to foster the acceptance. Even
resources. more security challenges arise in a virtualized environment due
to shared hardware, processor, memory, system bus and control

8
White Paper based on the IEEE Workshop SDN4FNS (Trento 11th – 13th November, 2013)
29th January 2014
Editor and contact: antonio.manzalini@telecomitalia.it

interface. Virtualization enables new charging models, which environment is to provide generic means to manage access and
require precise monitoring, policy control and charging from resource control on the node-level. While execution
the virtualized system. Existing Operation and Management environments support the installation, instantiation, and
system need to be adapted or renewed for facing the challenges configuration of services code in various ways, the virtual
of virtualized environments. The safety of SDN systems environment puts a uniform management layer on top. This
requires redundancy and reliability of the controller, the real allows external clients to interact with services through the
and virtual resources and other ad-hoc engineering solutions interface of the virtual environment in a generic way, and the
[42]. The flow paradigm offers interesting opportunities for interactions will be mapped to specific interfaces of the
security processing: one example is traffic steering for execution environments. Several execution environments can
automated malware quarantine. Another area to be explored is be attached to a virtual environment, just the same way as other
how SDN could also bring to network security the ability for resources. This leads to another aspect of virtual environments:
security policies to follow logically specific network the partitioning of resources. The network provider can set up
applications or VM. virtual environments on selected network nodes, and assign
them to a particular service provider, in order to offer a virtual
B. Interoperability and federation network. Access to the virtual environments will be made
SDN solutions count already on numerous Open Source available to the respective service provider so that it can
and commercial OF software and hardware switches and manage its own virtual network. The resource partitioning
controller. Already the large heterogeneity of OF Switches, OF implemented among virtual environments will prevent
Controller and OF versions challenges interoperability. interference with other service providers and, additionally,
Evolution of OF protocol and the use of other protocols than allow an accounting per service provider.
OF will require to make sure that interoperability is
guaranteed. 2) Programmability in Future Networks and Services
This challenge refers to solutions for the fast, flexible, and
In general interoperability between the different NOs and dynamic deployment of new network services though
different domains with the same NO is extremely important. programmable enablers and primitives for all planes of SDN
This challenge would include: environments (e.g. application, operation, management, control
 interfaces that will allow a servicing /networking virtual and forwarding planes). This is also aimed to provide easy
introduction of new network services by realizing the dynamic
function to federate. Using this interface, the networking
programmability of the network and its devices such as routers,
function should be able to cooperate in order to provide
switches, and applications servers. Dynamic programming
inter-domain communication;

refers to executable code that is injected into the network
authentication for other NOs, and the different NOs to element in order to create the new functionality at run time.
confirm with each other the identity of the different users The basic idea is to enable third parties (operators, service
of a particular service; providers and other authorized users) to inject application-

specific services (in the form of code) into the network.
mechanisms for communication and programmability of
Applications may utilize this network support in terms of
service modules deployed by different operators for the
optimized network resources and, as such, becoming network
same service;

aware. Hence, network programming provides unprecedented
mechanisms for end-to-end service management, flexibility in telecommunications. However, viable
monitoring, and accounting. architectures for programmable networks must be carefully
engineered to achieve suitable trade-offs between flexibility,
performance, security, and manageability.
C. Networked and Computing Services Operations
Service Operation deals with keeping the service The exploitation of such flexibility for the benefit of both
infrastructures (and the services it provides) up and running the operator and the end user would require guarantees against
smoothly. It includes, for example, monitoring the services jeopardizing the integrity and stability of the network based on
execution environments to spot problems, provisioning of solutions for
services, ensuring programmability, etc. 

Rapid deployment of new services;
1) Performant and Safe Network Execution Environments Customization of existing service features; optimization


This challenge refers to the network hosting virtual of network resources;
environments and virtual machines to overcome the problem of Scalability and cost reduction in network and service


having several execution environments implemented in various management;


technologies, and providing different abstractions, interfaces, Independence of network equipment manufacturer;
and so on. This challenge could be faced by developing a

Information network and service integration;
unified network execution virtual environment and by having

Guarantees for Quality of Service;
groups of virtual machines which are managed (creation, Diversification of services and business opportunities in
change/update, deployment, migration, orchestration, deletion)
particular for virtual environments and clouds.
as one. The advantage of having an explicit notion of a virtual

9
White Paper based on the IEEE Workshop SDN4FNS (Trento 11th – 13th November, 2013)
29th January 2014
Editor and contact: antonio.manzalini@telecomitalia.it

D. Architectural and Functional Models Advances in programmable and virtual networks have been
The current developments in SDNs and NFVs are driven by the adoption of Open-Flow, which has led to refined
highlighting new and critical research topics related to what high level architectural model, as in Figure 1.
and how create the conditions to effectively and continuously We are now witnessing a growing interest moving from the
update and change the networking functions (e.g., centralized control and “monolithic” approaches, where
softwarization of future networks and services without systems are vertically integrated, towards a component-based
reinventing every time the network architectures or network approach, where systems are made of multiple components
layering). from different manufacturers, interacting with each other
This means the use of software to program individual through open interfaces to form new services [45].
network devices, network systems and services dynamically As depicted in Figure,, the expected results would make it
and therefore control, manage and operate programmatically possible to achieve a level playing field, where different
the behaviour of the network as a whole. Key software features stakeholders, such as infrastructure and services providers,
of the future networks and services are already identified [43], would compete with each other, while users may select and
[44] and include: service diversity, functional flexibility, customize services according to their needs [46],[47][48].
programmability, ease of new services introduction,
virtualisation of resources, energy consumption, service
universalization, network management, mobility, optimisation,
identification, reliability and security would need to be realised
as part of the future network services and continuously
updated.
The future networks and services need to move from being
merely defined by software to be programmable by software
and must be capable of supporting a multitude of providers of
services that exploit an environment in which services are
dynamically deployed and quickly adapted to heterogeneous
physical wire, wireless and smart object infrastructure(s),
according to evolving and sometimes conflicting customer
requirements.
Programmability is a key property that SDN framework
enables, but it does not make it “easier”, as the proper
abstractions and a set of layering have not been completely
defined yet.
1) Architectural Models
Different abstractions, layering, conceptual models and
architectural approaches have been proposed in the research
literature. Some of them are discussed in the following.
Programmability of network elements (switches, routers, and
so forth) was introduced about a decade ago, this set the basis
for rapid deployment and customization of new services, as
illustrated in Figure 4.. Figure 5 – Example of model of a unified environment integrating
connectivity, computational power and storage

A fundamental characteristic of the architectural model of

Figure 5 is the evolution towards a unified environment
integrating connectivity, computational power and storage.
This is requiring proper enhancement of current control and
management planes and the introduction of orchestration
capabilities.
Before that, there is a need of defining a meaningful
functional model capable of abstracting and representing all
Figure 4 – Example of model of a programmable router network
network functions and capabilities which will appear in SDN
and NFV (which means including processing and storage).

10
White Paper based on the IEEE Workshop SDN4FNS (Trento 11th – 13th November, 2013)
29th January 2014
Editor and contact: antonio.manzalini@telecomitalia.it

Realizations and instantiations (with proper adaptations) of coordination/collaboration proxy, that needs to receive policy
the above model to Cloud [49] and in RAN [50] and cellular- directives from “above”, i.e., from nonfunctional requirements
wireline [51] domains (figure 6). (i.e., business, technology and quality) and translates them
One of the main problems with the earlier models is that the into the domain of applicability is a key function that is
richness and diversity of applications and technologies missing to achieve a global goal.
prevents a simple model to be representative for all approaches An overarching architectural model should include shared
(just a simple northbound API might not be enough). virtualised resources, and all the corresponding abstractions,
including those in wired, wireless and resource-constrained
A recursive approach is needed, where a new set of mobile devices and smart objects. Such an model would need
abstractions are developed, depending on the area or domain to be engineered to facilitate the integration and delivery of a
and the set of layers as necessary. variety of ICT services, computing and network Clouds and to
enhance integration of the key enabling technologies:
programmability, networks, network virtualization and network
function virtualisation and self-management.
2) Functional Modeling
Starting from the OSI Layering [52] integration of the
software infrastructures and traditional communication /
telecommunication technologies has been always a challenge
for network and service operators, as far as service deployment
and management are concerned [53][54]. This is due mainly
the fact that OSI layers were designed to address IP packet
delivery paradigm of early Internet days.
Interestingly, to overcome this problem, ITU-T developed a
functional formalism capable of modelling transport networks
in a technology independent way. ITU-T Recommendation
Figure 6 – Example of instantiations of SDN in Cloud, RAN and G.805 describes the modeling of connection oriented (point to
Cellular/Wireline domains (SDN4FNS Workshop) point) transport networks; then G.809 has been developed to
model connectionless network and, eventually, G.800 reports a
In addition as a result of users shifting to mobile model for unified connection oriented/connection less
environments and devices, the need to create denser networks architecture.
and more efficient usage of wireless resources are vital. (i.e.,
usage is rapidly growing on the demand side, on the resources
side). Generalizing a set of abstractions to mobile networks and
applying the SDN framework, would presumable bring
additional benefits in management, converged resources
control and agile deployment areas of increasingly dense
networks.
In these scenarios, there is a need for efficient resource
control and management across various wireless protocols,
ability to tie and coordinate these capabilities with other core
network functions, creating a common and programmable data
plane, and creation of a wireless network operating system.
Figure 7 – G.805 components for functional modeling
This will be only possible by rethinking the 3GPP
architectures, starting from the definition of the logical network
domains; session, mobility and identity management concepts The main characteristics of this functional modeling has
and architectures, as well as Access and Non-Access Stratum been the capability of describing recursively any transport
protocols. For instance, it remains to be seen how mobility will networks in terms of information transfer capability, in a
be ultimately handled in SDN without tunnels, and whether the technology-independent manner, and by using a small set of
protocols implementing the bearer service (3GPP layer) will components. Moreover the model is capable of relating the
converge with the protocols in the transport network layer, equipment/logical resource/management views.
which have been evolving separately. This approach has been very effective for the design and
In summary, as previously pointed out, SP’s networks the management of networks. It is argued that a similar
require a much more complex set of controls and layered approach, obviously with proper enhancements, should be
abstractions, which can be iteratively realized. A lot of details adopted for the functional modeling of SDN.
are still in need of development but a key piece, a

11
White Paper based on the IEEE Workshop SDN4FNS (Trento 11th – 13th November, 2013)
29th January 2014
Editor and contact: antonio.manzalini@telecomitalia.it

The key point will be definition of a small set of functional The standardization activities of the Object Management
components capable of modeling recursively a SDN in all its Group (OMG) aim to set a vendor-neutral global standard for
aspects (i.e., including also processing and storage features). SDN through the investigation of opportunities to foster the
development of open SDN specifications.
VI. STANDARDIZATION AND OPEN SOURCE ACTIVITIES
B. Open Source Software Initiatives
A. Standardization Bodies and other Fora
A great number of Open Source communities are working
The theoretical concept of SDN has been realized by on SDN, NFV, and Network Virtualization projects. As a
multiple standardization organizations (ONF, IETF/IRTF, matter of fact, the number of open-source projects is rapidly
ETSI, ITU-T, IEEE, etc.) into practical solutions (ONF’s growing. SDNCentral aims to capture a list of the main Open
OpenFlow, CISCO's OpenPK, IBM's DOVE, NEC's Source SDN projects [58].
Programmable Flow and more).
OpenDaylight is an example of forum created through the
Several working and discussion groups of the Open combination of open community developers and open source
Networking Foundation (ONF) are covering various SDN code and project governance that guarantees an open,
related topics. These are currently Active Working Groups, community decision making process on business and technical
Architecture and Framework, Configuration and Management, issues. The main goal is to help accelerate the development of
Forwarding Abstraction, Market Education, Migration technology available to users and enable widespread adoption
Discussion Group, Northbound Interface, Optical Transport, of Software-Defined Networking [59].
Testing and Interoperability, Discussion Groups, Wireless &
Mobile. The OpenFlow protocol in version 1.4.0 and Currently we are already witnessing a growing number of
OpenFlow Config protocol version 1.1.0 have been Industries (big but also SME) which are starting adopting Open
standardized by ONF. Both have the largest market share and Source (e.g. OpenStack, Avana, etc) - properly customized -
influence at this point in time, since most vendors and for real production environments. It is a sort of bottom-up
manufacturers support at least OpenFlow as part of their move that could change the rules of the game, as it is lowering
portfolio. the thresholds to new Players (even with little investments) to
enter the market. Arguably this will happen first at the edge,
IRTF Software Driven Networks focuses within SDNRG thus creating a fertile environment for creating and developing
on areas of interest covering the classification of SDN models, new ecosystems. Also this will create new roles and new ways
relationship to work ongoing in the IETF and other SDOs, of cooperating and competing. "Competition" will move more
SDN model scalability and applicability, multi-layer and more at the software level, so there is also a great value in
programmability and feedback control systems, system testing, assessing and certifying Open Source component,
complexity network description as well as security [55]. systems or platforms for security, trust, performances and
The IETF working group Forwarding and Control Element interoperability.
Separation (ForCES) defines an architectural framework and C. Need for Certification
associated protocols to standardize information exchange
It seems that there are very limited certification related
between the control- and data plane in a ForCES Network
activities on SDN. The Cisco Certified Network Professional
Element (ForCES NE). ForCES defines Network Elements
(CCNP) Service Provider certification program, which is
(NE), Control Element (CE) and Forwarding Element (FE). In
aimed at developing the skills and knowledge of IT
comparison to the SDN concept, which strictly separates
professionals to deploy and manage next-generation networks;
forwarding from control in different functional elements,
however, this is not specifically tailored for SDN, There is also
ForCES allows each NE to consist of multiple NE's and FE's.
Indiana University’s InCNTRE Lab, which is a sort of
A NE is therefore more complex and is controlled through a
certification lab for Open Flow.
CE Manager and FE Manager -each managing the referring
Control or Forwarding Element(s) [56]. Certification can be provided in the following three areas:
ITU-T Study Group 13 ITU (Future networks including  Professional Certification (SDN-P): This may consist in the
cloud computing, mobile and NGN) is standardizing FNs with establishment of Body of SDN knowledge (BoSK);
the objectives of service, data, environmental and socio- creation of BoSK Learning Tools that should include the
economic awareness as part of the topic Software-defined curriculum, books, newsletters, classes, as well as the
Networking (SDN) [57]. Standardization efforts include support. In addition, development of BoSK Examination
support network virtualization, energy saving features for FNs, Tools including SDN-P Certification Exam and Renewal is
and an identification framework. Future plans are to develop needed.
 Components Certification (SDN-C): This will require that
different facets of the smart ubiquitous network, requirements
of network virtualization for FNs, framework of telecom SDN
and requirements of formal specification and verification components requirements, application areas, and
methods for SDN. benchmarks need to be developed by solicitations and
interactions with the industry. Revisions need to be
managed. Also, evaluation tools for SDN Certification test-

12
White Paper based on the IEEE Workshop SDN4FNS (Trento 11th – 13th November, 2013)
29th January 2014
Editor and contact: antonio.manzalini@telecomitalia.it

bed, SDN-C Benchmark tests, and SDN-C Certification core/edge functional split require fundamental rethinking in the
need to be established. The process for certification fabric designs. This implies impactful aspects such as
renewals needs to be established as well. addressing the virtualization of L4-L7 middle-boxes and the
 Internal SDN-test bed development: Certification
RAN evolution. NOs and SP’s spend 60-80% of CAPEX on
RAN technologies that are not keeping up with the changing
procedures should be developed based on the initial test bed application requirements. In SDN, it remains to be seen
evaluations. Implementation of a test bed via Internet 2.0 whether we need the classical core network functionalities at
and local data centers/supercomputer centers is necessary to all, as most of the intelligence will be placed in terminals for
utilize state-of-the-art resources. Education activities will handling mobility and at the Edge (data centers) for hosting
be included in the SDN-test-bed for education/curriculum any type of service, from basic connectivity to Internet and
activities. Third party requirements which require SDN applications.
test-bed specs and heavy Industry relations need to be
established. Finally, an evaluation board for test-bed Functional modeling and architecture. A proper functional
certification need to formed. architecture would accelerate the development and the
standardization of SDN. (In facts, there is an urgent need to
VII. CONCLUSIONS standardize interfaces.) The main characteristics of the
Socio-economic drivers, IT technology progresses, functional architecture should make it possible to model
hardware down-spiraling costs and availability of open source recursively all SDN features, in a technology-independent
software are steering the evolution of future networks and manner, using a small set of components. The model should be
services infrastructures. It is likely that a wider and wider capable of relating the equipment/logical
adoption of “software” in said infrastructures will accelerate resource/management-control views. It might be advisable to
the pace of innovation of processes (as it is doing continuously follow the effective philosophy of the ITU-T G-805 functional
in the IT domains) and will reduce operational costs (e.g., modeling, with proper enhancements to take into account the
through optimizations exploiting big data) for NOs and SPs. nature of a SDN (i.e., processing, storing and transferring
information).
Emerging paradigms as SDN and NFV represent a first
concrete step to this direction, catalyzing the idea of Management and Orchestration. Complexity and
decoupling software defined control plane from hardware dynamism of SDN will require enhancing current management
driven data plane and the virtualization of network functions on systems (OSS/BSS) to cope with of a sheer number of real and
general purpose hardware. This will influence significantly the virtual ICT resources. The border between IT and network
future developments of 5G technologies and architectures. resources will blur, requiring the integration of the operations
of the two domains. Another major requirement will be the
As a matter of fact, we are witnessing a growing number of automation of business management processes (e.g.,
other industries moving to the same direction. In this sense, introducing autonomic and learning features exploiting
SDN and NFV could be seen as powerful enablers for new networks and services big data) and adopting decentralized
ecosystems capable of aggregating and driving investments approaches (at the current centralized model will not scale
beyond the traditional Telco-ICT contexts. It is argued that this properly). Moreover, the SDN management needs to be
transformation is unstoppable, because of the continuous complemented with new capabilities such as programmability
hardware technology evolution and cost reductions, which will and orchestration of the life-cycles of virtual network functions
enable new economic paradigms. In facts, this will move the and services. Security is one of the key areas of management
competition from hardware to software, creating the favorable which should be deeply investigated in SDN.
conditions for a sustainable “economy of information”.
Distributed computing. SDN and virtualization will offer
This will require different business rules, and different the opportunity to enhance and adopt in the network
kinds of jobs, workers and skills than the economy of the 20th infrastructures means and methods today widely used in Data
century, mostly based on industrial factories, manufacturing Center. At the same time Data Center concepts need to be
and manual work. Economic and cultural values in the evolved reliably and efficiently, especially looking at the
economy of information are, and will be, placed on integration with networks. Achieving scale by using collections
information, knowledge, creativity and intelligence to cope of distributed components (e.g., out of the shelf) and devices
with the fast-changing socio-economic environment. To this will enable reliability and create a more competitive
end, the development of new skills, mindsets and education are Technology Providers ecosystems. Major issues on consistency
required to face this transformation. and reliable distribution for various network tasks remains to
In synthesis, a number of research and development be tackled. Distributed sets of data representing states of
avenues are envisaged: network resources and functions demand for carriers’ class
solutions in compliance with the limits imposed by the CAP
Core-Edge split. SDN deployment strategies for Core and (Consistency, Availability, Partition tolerance) theorem.
Edge networks should be distinguished in terms of technical
approaches, business models and time horizons [60]. Details Taming heterogeneity, geographic distribution and scale.
and control frameworks for end-to-end networks with NOs’ and SPs’ networks will become far more distributed and
capillary, geographically and technologically. The way of

13
White Paper based on the IEEE Workshop SDN4FNS (Trento 11th – 13th November, 2013)
29th January 2014
Editor and contact: antonio.manzalini@telecomitalia.it

putting all this diversity under a common end-to-end software applications)

framework remains a major challenge, especially to cope with ASIC Application Specific Integrated Circuit
the rigid performance requirements in terms of QoS and QoE,
BNG Broadband Network Gateway
for instance. SDN could be a powerful enabler to meet the
above requirements but proper levels of abstractions and BRAS Broadband Remote Access Server
recursive control and management planes needs to be achieved BSS Business Support System
Pursuing Open Source Software. There will be a great CDN Content Delivery Network
value in testing, assessing and certifying Open Source
CP Control Plane - Control Plane – It is
component, systems or platforms for security, trust, responsible for taking and executing
performances and interoperability. decisions on how packets should be
forwarded. It focuses mostly on the
Peering and Software Defined Internet Exchange. Current forwarding plane and less on the operational
inter-domain routing protocol, BGP, is lacking in simplicity of plane. It could use operational and
management, security and flexibility in defining relationships. management planes information for fine-
SDN could allow the evolvability of BGPs independently of tuning and modification of the forwarding
from the underlying nodes hardware. SDN programmability plane actions.
could mitigate the main problems in inter-domain routing: CPU Central Processing Unit
security and accountability; pricing and contracts; and traffic DHCP Dynamic Host Configuration Protocol
management [61].
DNS Domain Name Service
Edge Self-Organization. In the future, the edge of the DPI Deep Packet Inspection
network will see a sheer number of nodes and devices (i.e.
aggregation nodes, terminals, machines, sensors, actuator...): FP Forwarding Plane – It is responsible for
controlling actions on packets in the data-
each of these elements with sufficient processing, storage and path, including dropping, changing and
communication capacities will become like a network node, forwarding packets. It contains forwarding
capable of sharing its capabilities. Them these nodes will be resources such as classifiers and actions.
able to aggregate to create local self-organized networks, HV Hypervisor
which, selecting appropriate gateways, will connect to the big
I/O Input/Output
network. Nodes decisions will be made based on local
conditions, adopting autonomic and self-organization LAN Local Area Network
capabilities [38], [66]. LB Load Balancer
Socio-economic impacts. In general, SDN and MP Management Plane is responsible for
virtualization appear to be expressions of a softwarization trend monitoring, configuring, optimizing,
in Industry. In ICT, this is lowering the thresholds for new maintaining state of systems, equipment,
Players to enter the market. This will also create new business devices and the overall networks. It
maintains and executes the processes that
roles and new ways of cooperating and competing. modify the operation of the control plane and
Competition, in fact, will move to the software level, whilst the it configures the forwarding and control
hardware will no longer be a differentiator. New business planes.
models and value chains and networks should be defined. As M2M Machine-to-Machine communications
an example, in the long term, one could even imagine scenarios
MVNO Mobile Virtual Network Operator
with a galaxy of ecosystems, where "trusted" network services
and functions could be provided by different Developers, NFV Network Functions Virtualization
exchanged and traded like in stocks exchange. NGN Next Generation Network

GLOSSARY AND DEFINITIONS NIC Network Interface Controller

NMS Network Management System
AAA Authentication-Authorization-Accounting NO Network Operator
ADN Application-Defined Networking is an OAM Operations Administration & Maintenance
enterprise data network that uses virtual
network and security components to provide OGF Open Grid Forum
a dedicated logical network for each
ONF Open Networking Foundation
application, with customized security and
network policies to meet the requirements of OpenFlow Specifications developed by the Open
that specific application. Networking Foundation
AP Application Plane – It is plane where OpenNaaS Specifications developed by the OpenNaaS
applications and services reside and execute. community
API Application Programming Interface (a set of OpenStack Specifications developed by the OpenStack
routines, protocols, and tools for building Foundation

14
White Paper based on the IEEE Workshop SDN4FNS (Trento 11th – 13th November, 2013)
29th January 2014
Editor and contact: antonio.manzalini@telecomitalia.it

OSS Operations Support System [12] H. Ali-Ahmad et al., “An SDN-based Network Architecture for
Extremely Dense Wireless Networks”, Proc. of IEEE SDN4FNS’13,
OTT Over The Top Trento, Italy, Nov. 2013;
PCE Path Computation Element [13] FP7 CROWD project - http://www.ict-crowd.eu;
[14] A. Basta et al., “A Virtual SDN-enabled LTE EPC Architecture: a case
PoP Point of Presence
study for S-/P-Gateways functions”, Proc. of IEEE SDN4FNS’13,
RAN Radio Access Network Trento, Italy, Nov. 2013;
[15] S. Namal et al., “SDN as an enabler for inter-technology load balancing
QoE Quality of Experience
and admission control”, Proc. of IEEE SDN4FNS’13, Trento, Italy,
SLA Service Level Agreement Nov. 2013;
[16] S. Namal et al., “Enabling Secure Mobility with OpenFlow”, Proc. of
SP Service Provider IEEE SDN4FNS’13, Trento, Italy, Nov. 2013;
SDN Software Defined Network [17] M. Mechtri et al., “SDN for Inter Cloud Networking”, Proc. of IEEE
SDN4FNS’13, Trento, Italy, Nov. 2013;
VN A virtual network (VN) is a network that
consists of virtual resources (e.g. processing) [18] F. Callegati and W. Cerroni, “Live Migration of Virtualized Edge
and virtual network links. The two most Networks: Analytical Modeling and Performance Evaluation”, Proc. of
IEEE SDN4FNS’13, Trento, Italy, Nov. 2013;
common forms of network virtualization are
protocol-based virtual networks, (such as [19] R. Trivisonno et al., “Virtual Links Mapping in Future SDN-enabled
VLANs, VPNs, and VPLSs) and virtual Networks”, Proc. of IEEE SDN4FNS’13, Trento, Italy, Nov. 2013;
networks that are based on virtual devices [20] J. Mueller et al., “Scalable On-Demand Network Management Module
(such as the networks connecting virtual for Software Defined Telecommunication Networks”, Proc. of IEEE
machines inside a hypervisor). SDN4FNS’13, Trento, Italy, Nov. 2013;
VM Virtual Machine [21] J. Zhu et al., “Software Service Defined Network: Centralized Network
Information Service”, Proc. of IEEE SDN4FNS’13, Trento, Italy, Nov.
VPN Virtual Private Network 2013;
vswitch Any Ethernet switch implemented in [22] F. Salvestrini et al., “Towards a distributed SDN control: Inter-platform
software alongside or inside a hypervisor. signalling among flow processing platforms”, Proc. of IEEE
There are proprietary and open SDN4FNS’13, Trento, Italy, Nov. 2013;
implementations of vswitch. [23] I. Bueno-Rodríguez, “OpenNaaS based SDN framework for dynamic
WAN Wide-Area Network QoS control”, Proc. of IEEE SDN4FNS’13, Trento, Italy, Nov. 2013;
[24] M. F. Bari, “PolicyCop: An Autonomic QoS Policy Enforcement
Xen Open Source Hypervisor Framework for Software Defined Networks”, Proc. of IEEE
SDN4FNS’13, Trento, Italy, Nov. 2013;
REFERENCES [25] R. Bifulco and F. Schneider, “OpenFlow rules interactions: definition
and detection”, Proc. of IEEE SDN4FNS’13, Trento, Italy, Nov. 2013;
[1] White paper on “Software-Defined Networking: The New Norm for
Networks” https://www.opennetworking.org/; [26] E. Escalona et al., “Using SDN for cloud services provisioning: the XIFI
use-case”, Proc. of IEEE SDN4FNS’13, Trento, Italy, Nov. 2013;
[2] White paper on “Network Functions Virtualisation”
http://portal.etsi.org/NFV/NFV_White_Paper.pdf; [27] M. Coudron et al., “Boosting Cloud Communications Through A
Crosslayer Multipath Protocol Architecture”, Proc. of IEEE
[3] SDN4FNS website: http://sites.ieee.org/sdn4fns; SDN4FNS’13, Trento, Italy, Nov. 2013;
[4] ONF, “SDN Architecture Overview” Version 1.0, December 12, 2013 [28] M. Vahlenkamp et al., “Enabling Information Centric Networking in IP
https://www.opennetworking.org/images/stories/downloads/sdn- Networks Using SDN”, Proc. of IEEE SDN4FNS’13, Trento, Italy, Nov.
resources/technical-reports/SDN-architecture-overview-1.0.pdf ; 2013;
[5] N. Feamster, J. Rexford, and E. Zegura, "The road to SDN: An [29] E. Patouni et al., “Network Virtualisation Trends: virtually anything is
intellectual history of programmable networks," ACM Queue, December possible by connecting the unconnected”, Proc. of IEEE SDN4FNS’13,
30, 2013 Trento, Italy, Nov. 2013;
http://www.cs.princeton.edu/courses/archive/fall13/cos597E/papers/sdn
history.pdf; [30] D. Siracusa et al., “Edge-to-Edge Virtualization and Orchestration in
Heterogeneous Transport Networks”, Proc. of IEEE SDN4FNS’13,
[6] A. Galis, S. Denazis, C. Brou,C. Klein (ed) –”Programmable Networks Trento, Italy, Nov. 2013;
for IP Service Deployment” ISBN 1-58053-745-6, pp450, June 2004,
Artech House Books, [31] R. Riggio et al., “EmPOWER: A Testbed for Network Function
http://www.artechhouse.com/International/Books/Programmable- Virtualization Research and Experimentation”, Proc. of IEEE
Networks-for-IP-Service-Deployment-1017.aspx; SDN4FNS’13, Trento, Italy, Nov. 2013;
[7] Path Computation Element (PCE) - IETF RFCs 4655 and RFC 5; [32] J. Batalle et al., “On the implementation of NFV over an OpenFlow
infrastructure: Routing Function Virtualization”, Proc. of IEEE
[8] IETF FORCES Forwarding and Control Element Separation; SDN4FNS’13, Trento, Italy, Nov. 2013;
http://datatracker.ietf.org/wg/forces/ ;
[33] CONTENT FP7 project http://content-fp7.eu;
[9] IETF i2rs - Interface to the Routing System
https://datatracker.ietf.org/wg/i2rs/; [34] K. Katsalis et al., “CONTENT Project: Considerations towards a Cloud-
based Internetworking Paradigm”, Proc. of IEEE SDN4FNS’13, Trento,
[10] S. Scott-Hayward et al., “SDN Security: A Survey”, Proc. of IEEE Italy, Nov. 2013;
SDN4FNS’13, Trento, Italy, Nov. 2013;
[35] Y. Kanada, “A Node Plug-in Architecture for Evolving Network
[11] Y. Bar Geva et al., “Tearing down the Protocol Wall with SDN”, Proc. Virtualization”, Proc. of IEEE SDN4FNS’13, Trento, Italy, Nov. 2013;
of IEEE SDN4FNS’13, Trento, Italy, Nov. 2013;
[36] W. John et al., “Research Directions in Network Service Chaining”,
Proc. of IEEE SDN4FNS’13, Trento, Italy, Nov. 2013;

15
White Paper based on the IEEE Workshop SDN4FNS (Trento 11th – 13th November, 2013)
29th January 2014
Editor and contact: antonio.manzalini@telecomitalia.it

[37] A. Galis et al., “Softwarization of Future Networks and Services – Next [53] D. Matsubara, T. Egawa, N. Nishinaga, M.-Ki, Shin, V. P. Kafle, A.
Generation SDNs”, Proc. of IEEE SDN4FNS’13, Trento, Italy, Nov. Galis, - “Open the Way to Future Networks – a viewpoint framework
2013; from ITU-T” – invited paper “The Future Internet- Future Internet
[38] A. Manzalini and R. Saracco, “Software Networks at the Edge: a shift of Assembly 2013: Validated Results and New Horizons” Lecture Notes in
paradigm”, Proc. of IEEE SDN4FNS’13, Trento, Italy, Nov. 2013; Computer Science 7858, Springer, pp370, May 2013, ISBN 978-3-642-
38081-5; http://www.springerlink.com/content/978-3-642-38081-5/;
[39] FP7 IP MobileCloud Networks: Mcn consortium, project number:
318109. http://mobile-cloud-networking.eu; [54] A. Galis, J. Rubio-Loyola, S. Clayman, L. Mamatas, S. Kukliński, J.
Serrat, T. Zahariadis, “Software Enabled Future Internet” - 5th
[40] Interworking and JOINt Design of an Open Access and Backhaul International Conference on Mobile Networks and Management
Network Architecture for Small Cells based on Cloud Networks: iJOIN. (MONAMI 2013), 23-25 Sept 2013, Cork, Republic of Ireland,
https://www.ict-ijoin.eu/; http://mon-ami.org/2013/show/home;
[41] D. Kreutz, F.M.V. Ramos, P. Verissimo, “Towards secure and [55] http://irtf.org/sdnrg;
dependable software-defined networks”- the second ACM SIGCOMM
workshop on Hot topics in software defined networking, August 2013, [56] D. Kreutz, F.M.V. Ramos, P. Verissimo, “Towards secure and
http://conferences.sigcomm.org/sigcomm/2013/; dependable software-defined networks”- the second ACM SIGCOMM
workshop on Hot topics in software defined networking, August 2013,
[42] S. Shin et al. “FRESCO: Modular Composable Secu- rity Services for http://conferences.sigcomm.org/sigcomm/2013/;
Software-Defined Networks”. In: In- ternet Society NDSS 2013;
[57] S. Shin et al. “FRESCO: Modular Composable Secu- rity Services for
[43] Y.3001 ITU-T recommendation – “Future networks: Objectives and Software-Defined Networks”. In: In- ternet Society NDSS 2013;
design goals” – 2011 http://www.itu.int/rec/T-REC-Y.3001-201105-I;
[58] Open Source SDN projects http://www.sdncentral.com/comprehensive-
[44] D. Matsubara, T. Egawa, N. Nishinaga, M.-Ki, Shin, V. P. Kafle, A. list-of-open-source-sdn-projects/;
Galis, -“Toward Future Networks: A Viewpoint from ITU-T” - IEEE
Communications Magazine, March 2013, Vol. 51, No. 3, pp: 112 – 118; [59] Open Daylight Foundation - http://www.opendaylight.org;
[45] E. Haleplidis, S. Denazis, K. Pentikousis, J. Hadi Salim, D. Meyer, O. [60] A.Manzalini, R. Minerva, F. Callegati, W. Cerroni (2013). Clouds of
Koufopavlou, SDN Layers and Architecture Terminology draft- Virtual Machines at the Edge, IEEE Com. Mag. “Future Carriers
haleplidis-sdnrg-layer-terminology-03, December 5, 2013, Networks”, July 2013;
http://tools.ietf.org/html/draft-haleplidis-sdnrg-layer-terminology-03; [61] Scott Whyte. Project CARDIGAN: An SDN-Controlled Exchange
[46] J. Rubio-Loyola, A. Galis, A. Astorga, J. Serrat, L. Lefevre, A. Fischer, Fabric. http://www.nanog.org/meetings/nanog57/
A. Paler, H. de Meer, “Scalable Service Deployment on Software presentations/Wednesday/wed.lightning3.whyte.sdn.controlled.exchange
Defined Networks”–IEEE Communications Magazine/ Network and .fabric.pdf, 2012;
Service Management Series, ISSN: 0163-6804; December 2011 [62] IETF NETCONF Configuration protocol
http://dl.comsoc.org/ci1/; http://tools.ietf.org/html/rfc4741;
[47] OpenStack - Open source software for building private and public [63] IETF SNMP - Simple Network Management Protocol
clouds- http://www.openstack.org; http://tools.ietf.org/html/rfc5343:
[48] ETSI ‘Software-aware and Management-aware SDN” - 3rd ETSI Future [64] iJOIN FP7 Project http://www.ict-ijoin.eu.
Networks Workshop 9-11 April 2013 - [65] J. Mueller, T. Magedanz, 'Towards a Generic Application Aware
http://docbox.etsi.org/Workshop/2013/201304_FNTWORKSHOP/eproc Network Resource Control Function for Next-Generation-Networks and
eedings_FNT_2013.pdf; Beyond', International Symposium on Communications and Information
[49] M. Banikazemi, D. Olshefski, A. Shaikh, J.Tracey, and G. Wang, Technologies (ISCIT), DOI:10.1109/ISCIT.2012.6381026, ISBN:978-1-
“Meridian: An SDN Platform for Cloud Network Services”, IEEE 4673-1156-4, Page(s): 877 - 882, Gold Coast, Australia, October 2–5,
Communications Magazine • February 2013; 2012, www.iscit2012.org;
[50] L. Erran Li, Z. Morley Mao J. Rexford, CellSDN: Software-Defined [66] A. Manzalini, P.H. Deussen, S. Nechifor et alii "Self-optimized
Cellular Networks, Open Network Summit (Research Track). Santa Cognitive Network of Networks", in Oxford Journals "The Computer
Clara, CA, USA (April 2013); Journal"; 2010, Volume 54, Issue 2, pp 189-196.
[51] A. Gudipati_, D. Perry_, L. Erran Li, S.Katti, SoftRAN: Software
Defined Radio Access Network, HotSDN 2013;
[52] Open Systems Interconnection (OSI) model (1994)- www.ecma-
international.org/activities/Communications/TG11/s020269e.pdf;

16