CaON
Converged and Optical Networks Cluster
FP7 Future Networks
White paper
Date: 14/03/2016
Chairs:
Prof. Dimitra Simeonidou (dsimeo@essex.ac.uk)
Sergi Figuerola (sergi.figuerola@i2cat.net)
Co-chairs:
Juan Fernández Palacios (jpfpg@tid.es)
Andrea Di Giglio (andrea.digiglio@telecomitalia.it)
1
List of Contributors
Contributors
Company/institute
e.mail address
Dimitra Simeonidou
UEssex
dsimeo@essex.ac.uk
Sergi Figuerola
I2CAT
sergi.figuerola@i2cat.net
Juan F. Palacios
TID
jpfpg@tid.es
Andrea Di Giglio
Telecom Italy
andrea.digiglio@telecomitalia.it
Anna Tzanakaki
AIT
atza@ait.gr
Nicola Ciulli
Nextworks
n.ciulli@nextworks.it
Andrea Bianco
Polito
andrea.bianco@polito.it
Reza Nejabati
UEssex
rnejab@essex.ac.uk
Georegous Zerva
UEssex
gzerva@essex.ac.uk
Mikhail Popov
Acreo
Mikhail.Popov@acreo.se
Josep Prat
UPC
jprat@tsc.upc.edu
Xavier Masip
UPC
xmasip@ac.upc.edu
Marcelo Yannucci
UPC
yannuzzi@ac.upc.edu
Raul Muñoz
CTTC
raul.munoz@cttc.es
Ramon Caselles
CTTC
ramon.casellas@cttc.es
Marcos…..
Acreo
Marco.Forzati@acreo.se
Joan A. García-Espín
I2CAT
joan.antoni.garcia@i2cat.net
Tania Vivero Palmer
TID
2
List of Acronyms
API
Application Programming
Interface
OLT
Optical Line Termination
CaON
Converged and Optical Networks
ONU
Optical Network Unit
CapEx
Capital Expenditures
OOFDM
Optical Orthogonal Frequency
Division Multiplexing
CD
Chromatic Dispersion
OpEx
Operational Expenditures
CDN
Content Delivery Network
OPS
Optical Packet Switching
DC
Data Centre
OSS
Operation and Support System
E-NNI
External network-to-network
Interface
PCE
Patch Computation Element
EPON
PLI
Ethernet Passive Optical Network
Physical Layer Impairment
FSAN
Full Service Access Network
POF
(Step-Index) Plastic over Fibre
(SI-POF)
GMPLS
Generalized Multi-Protocol Label
Switching
RACS
Resource and Admission Control
Sub-System
GPON
Gigabit Passive Optical Network
RDF
Resource Description Framework
IaaS
Infrastructure as a Service
RN
Remote Node
ICT
Information and Communications
Technology
ROADM reconfigurable optical add-drop
multiplexer
IETF
Internet Engineering Task Force
RWTA
Routing Wavelength and Time
slot Assignment
IMF
Information Modelling Framework
SDN
Software Defined Networks
LTE
Long Term Evolution
SDK
Software Development Kit
MAN
Metropolitan Area Network
SLA
Service Level Agreement
MIMO
Multiple-Input and MultipleOutput
SOA
Service-Oriented Architectures
MTOSI
Multi-Technology Operations
System Interface
TSON
Time-Shared Optical Network
NDL
Network Description Language
udWDM
Ultra Dense Wavelength Division
Multiplexing
NIPS
UNI
Network + IT Provisioning
Service User-to-Network
Interface
UHD
Ultra High Definition
NMS
Network Management System
UPnPQoS
Universal Plug and Play Quality
of Service
OAM
Operation and Administration and
Management
VXDL
Virtual eXecution Description
Language
OBS
Optical Burst Switching –or–
Operational Business Support
WSON
Wavelength Switched Optical
Network
OFDMA
PON
Orthogonal Frequency Division
Multiple Access Passive Optical
Network
3
Index
1.
Introduction........................................................................................................... 4
2.
Justification/Rationale........................................................................................... 4
3.
Technologies enabling the CaON reference model .............................................. 6
3.1.
Optical network IT convergence ....................................................................... 7
3.2.
Optical network virtualization ........................................................................... 8
3.3.
Cross-layer considerations .............................................................................. 11
4.
CaON Physical technologies in support of FI services ......................................... 12
4.1.
Core ................................................................................................................. 12
4.2.
Metro............................................................................................................... 13
4.3.
Flexible and Elastic Core/Metro optical Networks .......................................... 13
4.4.
Access .............................................................................................................. 15
4.5.
Access/metro and in-building/home networks .............................................. 18
5. CaON Control and Management Plane Technologies in Support of Future
Internet Services .......................................................................................................... 19
5.1.
Control plane evolution................................................................................... 19
5.2.
Management plane evolution: From rigidness to programmable
management ................................................................................................................ 22
5.3.
Evolution in Optical Networks towards cognitive and self-managed
networks and its impact on control and management planes .................................... 23
6.
Energy efficiency and Green networking ............................................................ 24
7.
Standardisation ................................................................................................... 25
7.1.
Optical data plane technology ........................................................................ 26
7.2.
Optical control plane ....................................................................................... 26
7.3.
IT and network integration.............................................................................. 27
8.
References ........................................................................................................... 28
2
Executive Summary
TBD….
3
1. Introduction
This white paper exposes the key role that optical networks and its associated infrastructures
have towards the success of Future Internet. It takes into consideration technical inputs gathered
across different projects composing the FP7 CaON clusteri, and presents the main trends for
optical networks research. These research topics are positioned with relevance to the CaON
reference model. This is a reference architecture model agreed among the projects belonging to
the cluster and reflects the high level architecture that the CaON cluster foresees for the Future
Internet. This positioning paper aims at complementing the relevant Photonics 21 and
Net!works white papers.
This positioning paper is structured as follows: it presents the rationale and trends of Future
Internet with regards to optical networks, followed by an overview of enabling technologies for
the CaON reference model. After presenting the reference model, the physical technologies with
their control and management planes are presented. Moreover, some standardisation strategies
are identified, together with the impact of energy efficiency and green IT.
2. Justification/Rationale
Optical infrastructure is the physical substrate that historically has enabled the wide
deployment of the Internet and continues to be critical for Future Internet. Flexibility,
transparency, capacity, low cost per bit, isolation capabilities and advanced provisioning
services make optical infrastructure a key enabler for the evolution and convergence of Future
Networks..
The Internet has become one of the basic infrastructures that support the World economy
nowadays. In fact, networked computing devices are proliferating rapidly, supporting new types
of services, usages and applications: from wireless sensor networks and new optical network
technologies to cloud computing, high-end mobile devices supporting high definition media,
high performance computers, peer-to-peer networks and a never ending list of platforms and
applications. In the last years there has been a trend (and a requirement) for a convergence of
the different networked platforms towards a unifying architecture or reference model for
seamless end-to-end communication regardless of the device technology and access/metro/core
infrastructure domain segmentation. Particularly, some of these different areas, technologies and
innovations at the infrastructure level are going to generate a big impact on the evolution of our
society. We can establish an initial differentiation between mid-term and long-term approaches.
Being the former the convergence of IT & Telco towards cloud computing, with optimisation of
interactions between applications providers, resource, service consumers, network operators and
infrastructure providers (with SLA mapping); and the later the definition of new architectures as
key area of basic research for the coming years with new technologies at the core, metro and
access networks.
Emerging applications are entering the arena of telco services with an unprecedented enduser acceptance. Similarly than the Internet has settled into daily life, Cloud Computing is
making its way towards becoming the invisible stratus on which companies base their IT
processes and users get their content. From the network perspective, it means understanding
traffic demands to adopt the technology combination that best fits its support. Video and Cloud
4
Computing demands are stressing the network as never experienced during the past decade, and
will be the drivers of the network infrastructure evolution roadmap (fig. 1).
Figure 1: Global consumer Internet traffic
Figure 2: VM and physical server shipment evolution
Moreover, proclaims of the advantages of Virtualized resources over Physical ones are well
known and can be found wherever in the Internet, e.g. resource usage optimization [1], saves on
energy consumption [2]. The introduction of Cloud Services in a massive fashion entails new
constraints that may be convergent with the ones that come from the distribution of contents
among the network. Here is where the core network will adopt a key role in Cloud service
provisioning. It may provide:


Connectivity capabilities for residential and business customers towards the DCs and the
external Internet.
Highly reliable, low delay and high bandwidth demanding interconnections between the
cloud/CDN DCs themselves.
Due to the wide range of final services and high traffic demand between users and providers
Cloud and DCs infrastructures will have to adapt to unprecedented levels of elasticity and
contain unpredictability. However, current core and metro networks are not ready for these new
traffic demands and behaviour. Core transport is characterized by a variety of networks,
technologies and providers. Metro networks, in charge of aggregating traffic from access nodes
(e.g. DSLAM, OLTs, Nodes B, corporate, etc), are typically based on Ethernet Metropolitan
Area Network solutions from different providers. Within this scenario, core network may make
up a bottleneck. Strategically, core, metro and access networks operation and capacity should be
adapted to new services demand, in contrast to current core architectures where the adaption to
new services is mainly covered by over-dimensioning and over-provisioning (i.e. overdimensioning in LANs and over-provisioning in WAN). To successfully respond to the traffic
demands presented in the previous point, optical networks must support:








An extensive amount of requests from DCs while the rest of traffic remains unaffected.
Bandwidth and QoS assurance between end users and DCs (i.e. real time applications).
QoS enhancement (via better use of existing network and data center).
Flexible networking services enabling on demand fast data transfers.
High capacity and scalability
Costs optimization (DC and network).
Responsiveness to quickly changing demands and infrastructure customisation.
Enhanced service resilience (cooperative recovery techniques).
The inadequacy of the current core architecture to fulfil these requirements (Error! Reference
source not found.) evidences the need of the conception of a new architecture capable to enable
flexible connectivity services, specially adapted to new requirements with reasonable costs. A
key challenge for optical networks is the capability to perform automated and flexible
connectivity services between end users and DCs. This network model is conceived to:
5



Accelerate service provisioning and performance monitoring.
Enable on demand connectivity configurations (e.g. bandwidth) toend users.
Optimize both converged infrastructure costs and energy footprint (e.g. consumption,
carbon footprint) Guarantee the required QoS (e.g delay, jitter…) for real time and video
services.
Key requirements for a Cloud enabled network
Current Core
Arch.
Cloud Enabled
Network
Connectivity
Service
Internet (L3)
Static IP VPN
Static L2 VPN
Flexible Connectivity Services
LOW
HIGH
MED
Guaranteed BW
NO
YES
YES
Guaranteed QoS
NO
YES
YES
LOW
YES
YES
Cost/ bit
Global
Global
MAN
Flexible
BW
YES
NO
NO
Automated
Operation
YES
NO
NO
BW beyond
10Gbps
NO
NO
NO
Global
YES
YES
YES
Range
Table 1
3. Technologies enabling the CaON reference model
The CaON reference model (figure 3) presents a multi-dimensional, layered architecture for
the convergence of optical networks and future technologies and services. The main conclusion
from the CaON cluster is that the ICT convergence plays a key role at the infrastructure level.
This convergence is the basis to bring innovation at upper layers and enable a real and powerful
cloud networked infrastructure deployment where the optical network can dynamically react to
different and new applications behaviour.
This is a bottom-up reference model, where the infrastructure and provisioning layers,
together with cross-layer SLA and management, are the key focus for future research trends
within the CaON cluster community.
Cloud/Service Layer
(e.g. app middleware layer)
Network Control Plane Layer
(i.e. network provisioning layer)
SLA Layer
Application Layer
(i.e. final consumers)
Management Layer(s)
The physical infrastructure layer covers from the core to the access optical network. Within
the infrastructure layer we can identify the virtualisation capability. It provides a more flexible
way to deal with infrastructure resource utilization by overcoming the multilayer and current
network segmentation, and a whole new set of functionalities (flexibility and new dynamic
provisioning services) that enables the convergence of optical infrastructures to support cloud
services delivery. Moreover, it facilitates the emergence of new business models by enabling the
entrance of new players. However, with regards to virtualisation there are still many research
topics that need to be addressed and further discussed (i.e. how isolation is managed and the
impact that non-linear effects have on it).
Virtualisation Layer
Physical Infrastructure(s*)
* = (s) to reflect network & IT and multiplicity of infrastructures
Figure 3: CaON reference model
More particularly, the provisioning layer is focused on a control plane architecture that may
provide a new set of functionalities at the infrastructure level, enabling:
6







Scalable multi-domain and multi-technology scenarios with open control planes and
enhanced UNI’s interfaces.
Automated end-to-end service provisioning and monitoring between different network
segments and operators with coordinated management planes.
Network resources optimization by integrated control of different network technologies (e.g.
IP and optical).
Network/IT resources optimization by means of cross-stratum interworking mechanisms.
Operation over virtual instances of the network infrastructure.
Convergence of analogue and digital communications unifying heterogeneous technologies.
Unified OAM mechanisms able to operate in a complex behaviour (multi-technology, multidomain and multi-carrier).
On top of the provisioning layer there is the service layer. It establishes the link between the
network infrastructure and the applications (cloud service requirements). This is the layer where
the network exposes its services, resources and capabilities, enabling:






Application to network interface: this interface may enable the request of new and advanced
services from the cloud to the network control plane.
On demand services provisioning with advanced re-planning functionalities.
Co-advertisement, co-planning, co-composition and co-provisioning of any type of network
resource and IT services (i.e. connectivity + IT resources at the end-points coordinated in a
single, optimal procedure)
Enhanced Traffic Engineering framework for resource optimization, advance allocation and
energy consumption, in support of energy-efficiency.
Implementation of network prototypes comprising the innovative data and control plane
solutions designed along the projects, in particular, pre-commercial software (control plane,
network-service interworking…) and hardware prototypes (sub-wavelength switching,
multi-granular nodes, etc).
Industrial exploitation: Accelerated uptake of the future networks and service infrastructures
enabling increased access capacity and flexibility, as well as cost and power consumption
minimization for intensive bandwidth consuming applications and cloud services.
At the cross-layer level, the CaON reference model considers two vertical layers. These are
the SLA layer, another interesting topic within the convergence approach, and the Management
layer. The former takes into consideration the mapping of the SLA requirements from the
application layer down to the infrastructure (virtual) resources. The later is in charge of
extending management functions across the different sets of resources, including virtual ones,
and layers in coordination with the control plane and the provisioning layer.
3.1. Optical network IT convergence
The IT and Telco convergence mainly deals with dynamic flexible behaviour of network
infrastructures and the integration of their operation and management processes with the IT
infrastructures systems and services. However, the end challenge is on the capability to provide
application-aware infrastructure through a new and well-defined set of Network/Infrastructure
Service Interfaces. Actually, the dynamicity of those applications and collaborative group
environments require that such infrastructures are provisioned on demand and capable of being
dynamically (re-) configured. Dynamicity is also necessary to optimize the resource usage and
reduce the service provisioning time, which so far is still slow and manual compared to
application service needs. In fact, these applications will continue to evolve in features, size and
amount of customers, as the associated business requirements change. Thus, the availability,
7
performance, security and cost-effectiveness of application-aware infrastructure remain critical,
as they support business decisions and data in a fast-paced, economy-driven environment.
Current provisioning services over hybrid infrastructures (managed networks and IT),
composed of both IT resources (i.e. compute and storage) and high capacity optical networks,
need unified management and provisioning procedures. This means the usage of cognitive,
flexible, elastic and adaptive technologies for core and metro optical networks, with dynamic
control plane functionalities and programmability features, as those in Software Defined
Networks (SDN) ,for the whole integration with the DC network infrastructures is a must. SDN
gives owners and operators of networks better control over their networks, allowing them to
optimize network behaviour to best serve their and their users needs. However, current disjoint
evolution has ended up with totally decoupled solutions for each type of resource and
infrastructure, those under the network operator domain and those under the DC administrator
domain. Therefore, there is a key technical challenge towards this ICT convergence and hence,
be able to optimize the (i) infrastructure sharing for lowering OpEx/CapEx costs, and (ii) the
(dynamic) services and applications deployed on top of these hybrid infrastructures with energy
efficiency considerations. In this context, convergence also considers the trend toward
infrastructure resource virtualisation and federation, thus providing full flexibility at the
infrastructure level.
3.2.1. Management and control planes convergence
Management and control planes convergence is required as a must for future-proof, and
Internet-scale enterprise applications. Distributed applications, consuming resources spread all
over the world, require DCs and network core/metro convergence in order to optimize the
service workflow and overall performance for cloud computing. Dynamic provisioning of one
type of infrastructure resources only considers part of the problem, and typically leads to a
waste of resources due to over-provisioning, mostly in networks, and sharing limitations in all
kinds of resource usage. It must be noted that, as time goes by, hardware is increasing its power
(switching, computing, storage, etc.) and embedding degree, which means that a higher control
in granularity is needed too, both at the network and IT level. In the end, the challenge is on
providing a common and transparent infrastructure able to integrate different technologies and
services, where virtualisation is not the end solution but an adequate technique for overcoming
many limitations. Some future research considerations are:








Keep IT/Telco converged infrastructure provisioning service (IaaS) time at a minimum.
Unified and converged resource description languages and frameworks.
Multi-granular, cognitive, elastic, flexible and adaptive optical networks (e.g. hardware
configuration).
Isolation and flexibility of circuit-oriented networks (using resource virtualisation).
Definition of the impact of these new technologies on legacy business models.
Inter-administrative domain issues between networks and DCs.
Non-standard service provisioning (alien wavelength services).
Carrier grade cloud and DC integrated infrastructure services.
3.2. Optical network virtualization
8
As commented, current physical infrastructures are mainly constrained by the amount of
resources they can deal with, and this has to be solved. New infrastructures will be composed of
heterogeneous resources that allow the delivery of any type of services between different nodes.
Resources like network elements, connectivity, storage and computation are those that take part
as core elements of the physical substrate and enable the creation of cloud infrastructures. The
challenge, however, is on the level of flexibility, optimization and transparency to deliver a
service and the need to map the abstraction (virtual representation) of physical resources and
network topologies with the applications and service requirements. No matter what the
infrastructure is, it would be homogeneously controlled and managed to deliver any requested
service. Virtualisation will help on overcoming the multilayer and current network
segmentation. Thus, at this point is where network virtualisation will bring the envisaged
flexibility for the network infrastructures.
Although many virtualization technologies exist for storage and computational resources, a
virtualization framework for the network infrastructure is not yet available. This framework
should provide the capability to virtualise the physical network infrastructure, federate
administrative resource domains from different providers, and provide the needed open
interfaces, APIs and SDKs to allow that control and management planes deliver any type of
service; independently of whether the physical substrate is analogue (fix and radio) or digital
based. Virtualisation has to provide the full capabilities to partition the physical substrate into
virtual resources, or create a virtual resource from the aggregation of physical and virtual
resources too.. One of the outstanding features behind virtualisation is isolation. All the virtual
resources must be isolated from each other. It is because they will be concurrently managed and
operated, and will share the same physical substrate. In that sense, Virtual Infrastructures (VIs)
will consist of dynamic composition, interconnection and allocation of these virtual resources.
Additionally, these VIs will offer its infrastructure capabilities as a service to third entities or
control/management planes.
Actually, virtualisation will have a large impact in networking that is not restricted to the
physical substrate. Its flexibility will allow and facilitate the deployment of new services at the
control and management plane (higher layers), with new type of open interfaces, business
models and relationships between entities. Moreover, the systematic and dynamic deployment
of VIs will allow creating customized infrastructures for new cloud applications.
At the analogue domain, optical network virtualization it is expected to be a key technology
for addressing future global delivery of high-performance, network-based applications such as
Cloud Computing, DCs connectivity and UHD video media services, among others. An optical
virtual network infrastructure would be composed of a set of virtual optical nodes and virtual
optical links, over a shared physical substrate, interconnected and managed by a single
administrative entity. Isolation and coexistence are the two most important characteristics of
virtualized optical networks, while the existing layer 2 and layer 3 virtualization solutions, such
as VLAN and VPN, respectively, take advantage of the digital nature of network equipments
and transport formats. Unlike L2 and L3, optical network resources and transport formats are
characterized by their analogue nature. Optical layer constraints, such as wavelength continuity
and physical layer impairments (PLIs) differentiate optical and other network resources.
Therefore, future research should take into account the physical characteristics of optical
networks and its implication on optical network elements and transport technologies, and how
coexistence of analogue and digital systems have to be provided.
9
3.3.1. Resource description
The term resource should refer to all physical resources (network devices/physical links) used
to provide connectivity across different geographical locations and IT equipment providing
storage space and/or computational power. Therefore, a pure optical network resource, any
other network resource or any IT resource should be considered as resources.. There is the need
to have a Virtual Network/Infrastructure Description Language that allows a complete detailed
description of virtual resources (Infrastructure/network/IT) as well as integrating the notion of
timeline consumption.
Semantic resource description and information modelling framework are needed to define
and implement models that can be used for the definition of optical, layer 2 or layer 3 networks
and IT resources. GEYSERS, as a first step and in order to work with the resources and
compose them, defines abstracted models that represent the corresponding resources as a set of
uniform attributes, characteristics and functionalities while hides unnecessary characteristics
from the resource itself [ABOSI09], as a continuation of the work done in PHOSPHORUS
[WILLNER09]. By means of this abstraction process the resources coming from independent
physical domains can then be used by the Network Control Plane and the Service Middleware
Layer in order to provision their corresponding services on top of the resources. However, many
topics need still to be analysed and covered in resource description, like elasticity aspects of
virtual resources, QoS or complete isolation, among others.
3.3.2. Infrastructure description languages
An Information Modelling Framework (IMF) provides common information modelling tools
in order to create homogenised resource data models and specify interfaces to seamlessly
manage different kinds of resources. Thus, an IMF must cover the type of information that the
data model should be able to describe, the relationships between different kinds of resources and
the capabilities that can be exposed through interfaces. Some of the aspects that a data model
must consider are: resource attributes (IT and network), virtual infrastructure description,
energy and consumption, quality of service and security. Moreover, an IMF needs to support
and facilitate basic data operations for abstraction, composition and partitioning, that is, all the
virtualisation types. The result of the IMF consists then in a model supporting aspects related to
physical location, access interfaces, QoS/QoE attributes, multi-layer technology description,
time constrains and a description syntax (e.g. RDF/XML).
From the network point of view, most of the IMFs being used nowadays offer support for two
existing description languages, the Network Description Language and the Virtual eXecution
Description Language. Since an IMF may require flexibility and extendibility, semantic
approaches should be adopted in order to describe the resources and facilitate their logical
manipulation. The basic hierarchy of an information model should be built using the concept of
a Resource as the top element. This concept can be a Device, a DeviceComponent or a
NetworkElement. Basically this hierarchy enables to describe devices, their components and the
network elements connecting these devices. Different types of device components exist, each
one with different properties. Memory, processing and storage components can be used to
describe the platform of an IT resource. Switching components can be used to describe switches
or routers, while specific types of optical switching components need to be included to describe
the specific properties that are required for the virtualization process of these optical
components.
10
3.3.3. Use case example: Nomadic virtual PC over optical networks
This is a use case that reflects the need of convergence between optical networks and cloud
applications. Delivering a Virtual PC service to mobile commercial users is a task which
requires having certain network parameters such as minimum latency, jitter or transmission
(bandwidth) speeds at certain permissible threshold values. Note that when we refer to mobile
Virtual PC service users we mean users who will demand the service from different locations,
and not necessarily users who will be consuming the service while on the go (on a roaming
basis). A user consuming this virtual PC service will only tolerate certain (low) delay/latency,
which implies a need to have the VMs executing on the edge node closest to the user’s physical
location for the service to be commercially feasible and acceptable. These conditions impose the
need to physically move the VMs as quickly as possible from the previous edge node of
execution, to the current one using the optical data plane. The time invested in the transfer will
be perceived by the user as part of the “boot-up” time, and it has to be kept at a minimum. In
that sense, current research within the optical network metro architecture as described in the
MAINS project [add reference] may suit a variety of these novel services.
3.3. Cross-layer considerations
As a new step towards the convergence between cloud environments and optical transport
networks, there is the need of an innovative interface between the Service Layer and an
enhanced GMPLS-based Network Control Plane [GEYSERS]. This interface, called Network +
IT Provisioning Service User-to-Network Interface (NIPS UNI), within the GEYSERS scope,
enables active cross-layer cooperation for end-to-end service delivery.
The NIPS UNI aims at being a key enabler for the seamless and on-demand provisioning of
the heterogeneous set of networking and IT resources associated to cloud networking services.
The mechanisms offered to exchange cross-layer information about capabilities, availabilities,
route quotations and QoS requirements will allow more efficient orchestrations of the global set
of resources; in fact, network and IT resources can be jointly and automatically selected and
optimized, thus better satisfying the requirements of distributed applications for tailored
performances and reliability. The work done under the NIPS UNI specification, which defines
both the semantics and the procedures for Service Layer and Network Control Plane intercooperation along the entire service lifecycle from setup to tear-down, will allow further
research of cross-layer integration within the CaON reference model. In fact, it offers services in
support of scheduled connections, cooperative or automatic selection of IT end-points,
quotations for end-to-end connectivity, dynamic service modification, monitoring
functionalities, and cross-layer strategies for coordinated recovery of IT and network services.
Traditional UNIs act as a demarcation point between network service providers and
subscribers over which just connectivity services are requested, offered and monitored. On the
other hand, the NIPS UNI may evolve towards an interface that widens its services to the
aggregate composed of both networking and IT resources. In these terms, it becomes a logical
interface that allows network service providers to offer customized transport network services,
tailored according to the application requirements, to cloud providers.
This trend is also foreseen in other domains. In the MAINS project, focused on a new
multiservice metro network architecture that allows the application/service layer to access subwavelength optical layer resources on-demand and at the granularity of optical packets and/or
11
optical bursts, the control functionalities of the sub-wavelength data plane are implemented
through a supervising sub-wavelength capable GMPLS control plane, which also exposes a
unified Network Service Interface (MAINS Network-Service interface, MNSI). Through this
interface, aligned with the standard OIF UNI 2.0 network services, the service layer interacts
with the network control to reserve/configure connectivity services in the sub-wavelength
transport plane.
4. CaON Physical technologies in support of FI services
The key requirements of innovative ultra-high bandwidth networks refer to scalability,
flexibility, assurance of end-to-end quality of service and energy efficiency, beside reduction of
total cost of ownership. In the data plane, current equipment and network architectures still
provide limited scalability, are not cost-effective and do not properly guarantee end-to-end
quality of service. Thus, the control plane has to define an end-to-end control structure that
allows different technologies and domains to inter-work efficiently, incorporating virtualization
of network resources. Based on these rationales the main objective for a future transport
network is that it should be/offer:




Compatible with Gbit/s access rates.
Equipped with a multi-domain, multi-technology control plane and provide Optimal
integration of Optical and Packet nodes.
High scalability, flexibility and guaranteed end-to-end performance and survivability
Increased energy efficiency and reduced total cost of ownership
To face the scalability and flexibility problems for future transport network and, in the same
way, to guarantee a energy and cost savings, the approach of leveraging on architectures (in
parallel to be aware of the technology evolution) seems to have some advantages since it can be
shortly applicable and it can be also be compliant with legacy carriers networks.
4.1. Core
Main objectives of the Projects in the cluster dealing with core network evolution is the
definition of a transport network architecture, complying with requirements on scalability,
flexibility, end-to-end quality of service, energy consumption and cost, for both mid-term
(based on: elimination of IP transit routers; use of integrated wavelength switching and packet
transport) and long-term scenarios (further based on: multi-granular switching nodes; power
efficient ultra high capacity packet processing). In fact, it is demonstrated that an important
fractions of functions embedded in large-size routers are not actually used, represent the most
important share of energy consumption and it is one of the first item of expenditures. The key
areas of research for core network evolution involve the analysis of the feasibility of the
different architectures by means of performance and techno-economic impact studies, aiming at
network performance and cost. The assessment parameters are considering:
Reduction of energy consumption: Identify the best solutions to reduce the energy consumption
of the telco’s networks. Efficient combinations of O/E components needs to be investigated.
Combination of best of transport technologies. Research, develop, analyze and validate
optimum combination of L1(Optical) and L2(Packet Transport, OBS,…) transport technologies.
Control Plane for end-to-end service delivery. Pursue e2e services delivery across
heterogeneous domains in terms of technologies (circuit transport networks and connection12
oriented packet transport networks), control plane models (e.g. multi-layer/multi-region), OAM
mechanisms, vendors and operators. The identified control-layer and the related control-plane
architecture should be compatible with multi-domain and multi-technology scenarios, rely on a
hierarchical path computation element (PCE) approach, implemented in a wider resource and
admission control (RACS) framework.
Enabling the virtualization of resources. Enable the virtualisation of resources, and allowing the
cooperation among heterogeneous data-plane technologies to permit quick and low-cost
introduction of new services independent of underlying transport platform.
4.2. Metro
A broad range of emerging services and applications (wide-range of multi-media, distributed
applications such as Cloud, etc.) are driving the growing trend of network traffic with increasing
demand for high bandwidth and flexibility. In addition, such applications require guaranteed
multi-granular short-lived services i.e., from seconds to minutes with bandwidths from Mbps to
hundreds of Gbps. In order to provide these services, a new subwavelength switching network
architecture is required that can deliver dynamic access to transparent multi-granular flows as a
guaranteed (no contention) network service.
Optical packet switching (OPS) and optical burst switching (OBS) have been proposed to
support subwavelength services [1]. However, these techniques do not provide guaranteed
bandwidth services. It is also worth noting that current approaches consider ring solutions [1,5]
for metro. In that sense there is clear trend towards novel optical network solution – the Time
Shared Optical Network (TSON)[6] – to deliver both highly flexible statistically multiplexed
optical network infrastructure and on-demand guaranteed contention-free time-shared multigranular services. In that sense, TSON supports traffic flows from any source to any destination
in transparent optical networks for the metro region supporting the physical interconnection
requirements. It is based on user/application-driven bandwidth service requests, centralized
RWTA calculation, and one-way tree-based provisioning that allows for flexible
symmetric/asymmetric multi-granular bandwidth services with the use of either fixed or tunable
transceivers. It delivers contention-free optical switching and transport of contiguous and noncontiguous time-slices across one or multiple wavelengths per service. It also doesn’t require
global synchronization, optical buffering and wavelength conversion, thus, reducing
implementation complexity.
4.3. Flexible and Elastic Core/Metro optical Networks
Numerous studies have demonstrated and investigated the highly variable and complex nature
of internet traffic. Uncertainty in traffic demands, granularity, geographic and temporal
distribution may arise from the varied requirements of different applications, changes in
customer behaviour, uneven traffic growth or network failures. As such, networks need to be
able to cope with some level of uncertainty in order to provide an acceptable quality of service,
e.g. low blocking probability. One way to deal with uncertainty is to overprovision network
resources. However, this leads to inefficiency and higher costs. Another way is to equip
networks with flexibility according to the type of uncertainty that needs to be addressed. For
instance, dealing with uncertainty in the geographic distribution of traffic requires networks
with the flexibility to route channels to different destinations. Similarly, the requirement for
future optical transport and networks able to carry mixed bitrates, e.g. 10 Gb/s, 100 Gb/s, 400
Gb/s, 1 Tb/s and beyond, has triggered a great deal of interest in elastic optical networks. In
13
such networks spectrum allocation is performed in a flexible manner, depending on the
requirements of individual channels. For instance, it is possible to allocate contiguous 75 GHz
of spectrum for 400-Gb/s or 150 GHz for 1Tb/s. Moreover, transport of low traffic with
increased spectral efficiency is feasible by reducing channel spacing, e.g. 10 Gb/s with a 25GHz spacing. In addition, it is possible to support bandwidth variable transmission, whereby the
optimum bitrate for the required reach is used, which significantly increases network efficiency.
Therefore, this additional flexibility enables matching allocated spectral resources to channel
requirements, thereby providing efficient optical transport. However, technology limitations,
e.g. 12.5-GHz spectral slot size, restrict the use of elastic spectrum allocation to entire
wavelengths, i.e. 10 Gb/s granularity. For finer traffic granularities, several subwavelength
multiplexing techniques have been proposed, such as optical packet switching (OPS), optical
burst switching (OBS), orthogonal frequency division multiplexing (OFDM) and time-shared
optical networks (TSON). It has been recently shown that the combination of elastic spectrum
allocation and elastic time multiplexing may be used to provide extensive bandwidth
granularities in the optical domain.
Elastic optical networking presents a number of challenges. Most notably, the mix of channels
with high and low bandwidth requirements e.g. >1 Tb/s and 10 Gb/s, may give rise to spectrum
defragmentation. Spectrum gets fragmented as channels are added and removed leaving behind
non-contiguous empty slots. When a high bandwidth request arrives there may not be sufficient
contiguous bandwidth to accommodate it, which results in blocking. Techniques for spectrum
defragmentation involving relocation of existing wavelengths, e.g. by means of wavelength
conversion, have been proposed. However, before a high-bandwidth request arrives it is
uncertain which channels would need to be relocated. To support diverse traffic demands and
a broad range of granularities, future optical transport networks may need to support a
combination of transport functions such as elastic allocation, switching and resource
defragmentation in space, time or frequency. Furthermore, the demand for these and other
emerging functions operating on multiple dimensions, e.g. time, frequency, space, phase, etc.,
may be fluctuating or depend on the network region considered, e.g. metro, core.
The first demonstration of an elastic optical network based on OFDM transmission. Since then,
a number of studies have investigated elastic networking showing significant gains in network
mean traffic, required spectral resources , capacit], cost, etc. Other work has focused on the
development of bitrate-variable transceivers and spectrum defragmentation. Recently, there
have been important demonstrations on automated adaptive transmission and networking. In
spite of the increasing popularity of elastic optical networks, there has been very little work
focusing on elastic node and network architectures
In this context the concept of elastic optical transport is based on the ability to dynamically
partition the fibre bandwidth into variable-size spectrum slots. The size and shape of each slot
are usually tailored to the requirements of a specific channel or group of channels so that
efficient transport across the network is achieved. This fine slicing and shaping of passbands is
not accomplished by passive components, e.g. arrayed waveband grating (AWG). Instead,
active components typically based on liquid crystal on silicon (LCoS) or microelectromechanical systems (MEMS) are used. Elastic time multiplexing requires fast switching
devices, e.g. ns switching time, in order to achieve fine switching granularity and high
efficiency. Fast time switches are usually implemented with semiconductor optical amplifiers
(SOA) or electro-optic materials such as LiNbO3 or PLZT. The combination of flexible
spectrum switching and fast time switching technologies enables elastic time and frequency
allocation.
The main objectives of future research in this topic should focus on:

Scalable and flexible data plane technologies
14
Innovative transmission, switching and grooming technologies enabling transport beyond
1Tbps.

Control Plane for elastic optical networks
To design of new control plane solutions for scalable and adaptive flexible and elastic
optical networks in order to support end-to-end connection provisioning and recovery
services delivery crossing domains that are heterogeneous in terms of technologies and
control plane interworking models

Node and network architecture for elastic optical networks
Design elastic optical nodes and networks based on the relevant advanced data and
control plane technologies
4.4. Access
Next-generation of optical access networks are foreseen to provide multiple services
simultaneously over common network architectures for different types of customers. In recent
years, most studies are being focused on time division multiplexing passive optical network
(TDM-PON) and wavelength division multiplexing passive optical network (WDM-PON). The
TDM technology based on GPON and EPON and its future developments allow dynamic
bandwidth allocation but complex scheduling algorithms between several ONUs are needed,
therefore, through each time slot only one ONU can transmit or receive simultaneously
information. Consequently, the performance of this technology is highly sensitive to packet
latency and not transparent to other kind of traffic that shares the same link.
In the other hand, WDM-PON is able to deliver multiple services transparently to each ONU,
due to each ONU can use a dedicated wavelength. However, WDM-PON isn’t enough flexible
to dynamically allocate the bandwidth for several ONUs and transceivers, optical filters and
other devices are needed for this type of network, increasing the cost of the solution and making
it unfeasible for all type of customers. In contrast to previous technologies, the orthogonal
frequency division multiple access passive optical network (OFDMA-PON) can transparently
support various services, allows dynamic bandwidth allocation among services, in addition it
has resistance to some dispersions effects like chromatic dispersion (CD), consequently the
complete bandwidth can be divided into both orthogonal frequency-domain subcarriers and
time-domain slots, in this way each ONU can be assigned one or more subcarriers in a given
time slot. Mainly, the OLT in an OFDMA-PON system is able to support heterogeneous ONUs
using a single receiver per PON port. In that sense, the ACCORDANCE projects enables a
seamless OFDMA-based access network where all different Telco services are consolidated,
allowing a full coexistence of fixed and wireless applications.
The benefit of this technology is based on the OFDM modulation format that also offers
additional advantages, such as:


Allows using the spectrum with more efficiency due to the use of multilevel modulation
formats like QPSK or M-QAM.
Simple to scale to higher constellations sizes and higher bitrates.
15



Better chromatic dispersion tolerance: in OFDM, the speed at which each sub-carrier is
modulated is less than the aggregate rate and with the use of a cyclic prefix can mitigate the
effects of chromatic dispersion and achieve a greater reach.
High spectral efficiency, orthogonal sub-carriers without guard bands can be overlapped
making it more efficient than FDM schemes.
OFDM is currently used in access networks over copper pair as xDSL and wireless (WiFi,
WiMAX or LTE), so using OFDM in the optical network will simplify convergence
between different technologies in the access network and could facilitate more efficient
traffic management.
Thus, from telco point of view, although not yet mature, this technology is attractive
compared with other fixed access technologies. Mainly due to its flexible architecture, its cost
access system for the delivery of heterogeneous services, the high-speed that can be achieved,
its high spectral efficiency and its powerful bandwidth granularity. Although several
technologies for access networks are introduced below, the experience gained along the CaON
projects brings some key topics to be addressed for next generation of access networks:



End-to-end low round-trip delay for multimedia communications.
Access network scalability, in terms of connected users, BW and distances, sharing a
limited infrastructure, integrating radio-PON and providing an effective resiliency, as the
network extends to a higher dimension.
Next Generation Access models: Open neutral network versus operator vertical model
4.3.1. Optical Orthogonal Frequency Division Multiplexing OOFDM
The OOFDM technology roadmap can be divided into three major phases (fig 4).
Fig.4. Roadmap for OOFDM technologies (from ALPHA project, D4.5p).
4.3.2. Radio-over-fibre technology
Radio over fibre may be deployed in two application domains: Access networks for mobile
telephony networks (GSM, GPRS, UMTS, LTE), and In-building and home networks for
wireless broadband (WLAN, 60GHz, UWB, …) where steadily growing capacity demands are
put on the wireless connectivity for communication terminals. These growing capacity needs
per user can be solved in several ways: by decreasing the radio cell size, by increasing the
transmission capacity per radio frequency channel, and by multiple antenna techniques
(MIMO). Fig 5 presents an indicative timeline of RoF technologies networks.
16
Fig.5. Roadmap for Radio-over-Fibre technologies (from ALPHA project, D4.5p).
4.3.3. Large-core Plastic Optical Fibres for in-building and home
networks
The current conventional step-index POF (SI-POF) offers a solution for home networking
that can be immediately used due to the existing commercial products (mostly for Fast
Ethernet). POF is seen as a valid alternative to the electrical solutions, like Cat-5e/6a or coaxial
cables. To the best of our knowledge, ALPHA [ref] has been the first project where a Gigabit
POF transceiver prototype has been developed. For future developments, the power budget of
the system could be increased in order to make the solution more robust, in particular working
on the coupling condition between fibre and photodiode. Another area to investigate is the use
of blue or green laser diode.
Fig.6. Roadmap for POF technologies (from ALPHA project, D4.5p).
The future roadmap expected for POF technologies is illustrated in Fig. 6, and is conditioned
by the home networking market evolution. The success of POF technology in the in-building
network segment will also depend on some factors that are outside an EU research project.
4.3.4. Resilient hybrid WDM/WDM-PON
This first network solution, developed in the FP7 SARDANA [ref] project, gracefully
integrates the GPON optical TDM multiplexing, at a higher rate, with the optical WDM
17
multiplexing in hybrid architecture. With this integration, the fine granularity and scalability of
TDM combines with the huge bandwidth capacity and power efficiency of WDM. This
technology is implemented over an alternative architecture with respect to the conventional tree
WDM/TDM-PON, consisting on the organization of the optical distribution network as a WDM
bidirectional ring and TDM access trees, interconnected by means of cascadeable optical
passive Add&Drop remote nodes (RN). This type of technology aims at serving more than 1000
users spread along distances up to 100 km, at 10 Gbit/s, with 100 Mb/s to 1 Gb/s per user in a
flexible scalable way. The ring+tree topology can be considered as a natural evolution, from the
conventional situation where Metro and Access networks are connected by heterogeneous
O/E/O equipment at the interfaces between the FTTH OLTs and the Metro network nodes,
towards an optically integrated Metro-Access network.
4.3.5. OFDMA-PON
Another investigated promising PON technical solution is based on OFDMA (Orthogonal
Frequency Division Multiple Access); this technology/protocols can introduce ultra high
capacity, even reaching the 100Gbps regime, in extended reach optical access network
architecture, as proposed in the European ACCORDANCE project [2]. OFDM is implemented
through the proper mix of state-of-the-art photonics and electronics. Such architecture is not
only intended to offer improved performance compared to evolving TDMA-PON solutions but
also inherently provide the opportunity for convergence between optical, radio and copperbased access. Although OFDM has been used in radio and copper-based communications, it is
only recently that is making its way into optics and is expected to increase the system reach and
transmission rates without increasing the required cost/complexity of optoelectronic
components. In that sense, ACCORDANCE hence aims to realize the concept of introducing
OFDMA-based technology and protocols (Physical and Medium Access Control layer) to
provide a variety of desirable characteristics, such as increased aggregate bandwidth and
scalability, enhanced resource allocation flexibility, longer reach, lower equipment cost &
complexity and lower power consumption, while also supporting multi-wavelength operation.
4.3.6. Ultra-Dense-WDM-PON
For longer term development, an alternative to the exploitation of the electrical-over-optical
domains could consist of the direct intensive use of the optical spectrum, while minimizing the
electronics requirements in terms of bandwidth and power consumption. This can be achieved
by ultra-dense WDM multiplexing (udWDM-PON), with very narrow filtering techniques or by
coherent homodyne detection. New developments in photonics and signal processing can enable
this next-generation large-scale access networks based on ultra-dense wavelength division
multiplexing (U-DWDM) targeting more than 1000 users on a single architectural platform with
low-cost deployment. Pure optical OFDM is a step further in this direction.
4.5. Access/metro and in-building/home networks
The ALPHA project has developed solutions and respective roadmaps (fig. 10) for the cross
domain control and management of access/metro and in-building home networks. The solutions
and roadmaps have been based on the existing technologies with extensions and provide an
evolutionary path for the development of integrated control and management in the domains of
metro, access and home networks. These solutions address the formulated requirements for:
18




Unified network management of heterogeneous networks (networks of networks)
Context-aware networking
Flexibility, scalability, efficiency and robustness (Intelligent and Controllable)
Green networking
Integrated control
Unif ied control plane f or
home and access,
Full f ixed/mobile convergence
Cross domain end-to-end provisioning
GMPLS controlled NG AON access
Integration of NG-PON with wireless AP (WiMAX/LTE)
Integrated UPnP-QoS (or similar) and GMPLS
Per domain QoS
Access: Management based Std. Ethernet, XG-PON/10G-EPON, MPLS, and IP QoS
Home: UPnP-QoS (or other CP), parameterised f low management
Prioritising flows
Access: Reconf igurable management plane based on e.g. AON/Carrier Ethernet
Home: Prioritised access f or in-home f lows through the gateway
Best effort, separate or no control plane
Access: f ixed access bandwidth, IP connectivity to access
Home: no central QoS controller in home
0
now
+1 year
short
term
+5
medium
term
+10
long
term
very long
term
Figure 10 High-level roadmap for cross domain issues and end-to-end QoS provisioningin access and home networks’
In the access, the GMPLS control plane will be more advanced and the users can request for a
specified bandwidth. UPnP-QoS or a similar control plane in the home will automatically
request the necessary resources in the access network through the gateway. In the very long
term, a unified and common control plane like GMPLS will act as glue and mediator between
the user and the access network, and full fixed/mobile convergence will be supported.
5. CaON Control and Management Plane Technologies in Support
of Future Internet Services
5.1. Control plane evolution
The Future Internet grow is enabled not only by the bare optical transport technologies with
enhanced capabilities, but also by the control plane tools and procedures that can guarantee the
related provisioning, monitoring and survivability of the involved resources and services.
The research in network control planes is currently focused on consolidating the control
procedures adopted for the underlying optical infrastructure, and on extending a generalized
(single-instance) control approach to include more and more technologies. Both objectives
involve different architecture aspects with different degrees of maturity: they can range from
more evolutive extensions to the control plane protocols when there is the need to incorporate
new advances in optical data plane technologies (such as new Optical Transport Network
multiplexes or grid-less networking), and can scale up to more extreme and demanding
interactions between the control plane and the network service layer (e.g. the cloud) for
controlling new types of enhanced connectivity services (i.e. beyond the point-to-point).
Additionally, since optical networks represent the core substrate responsible for inter-carrier
data transport, other key research topics addressed in this area include possibly standardized
multicarrier and multivendor control solutions to make more effective and open (i.e. vendor-
19
independent) the current implementations. Some of the mainstreams in the current control plane
evolution, in decreasing order of importance and compelling requirements, are:



Opening the control plane domains towards true multi-vendor and multi-carrier scenarios
Decoupling of the optical transport from the control plane(s)
More flexible and powerful User to Network Interfaces (UNI); i.e. equipping the control
plane with more advanced interfaces to external end-user “systems” (e.g. clouds) for any
type of bandwidth-on-demand provisioning service, and above all seamlessly integrated
with the service layer workflows
5.1.1. “Opening the CP domains/systems” for true multi -vendor
and multi-carrier interactions
After many attempts for inter-vendor interoperability through standardized protocol
extensions (IETF), also supported by industry-driven Implementation Agreements (OIF E-NNIs
and UNIs), the issue of multi-vendor equipments within one operator’s network is still
unresolved, above all in case of multiple switching technologies. Several reasons contributed to
this limitation: primarily the possibility offered by the current standards for different
interpretations of complex procedures, which led to a diversity of deployment options by
vendors and different degrees of compliance for implemented features; then, the different pace
of market availability of specific technological solutions by a single vendor (e.g. ROADMs and
WSON equipments under GMPLS control) with respect to the slower consolidation of the
related reference standard modelling and control procedures. Both these causes led to the
proliferation of many proprietary (vendor-specific) extensions and different equipment
behaviours, above all in the optical domain; subsequently, a sort of “protected market-niches”
for vendors has been created, as they can deliver their systems as highly integrated “all-in-one
black boxes” (i.e. bundles of control and transport plane components, possibly extended to the
management plane). The limitation on control plane openness is further complicated at the intercarrier interfaces, where many other issues needs still to be solved; for example, the definition
of reference mechanisms to dynamically establish trust relationships among carriers is still
undefined, as well as technology-agnostic signalling procedures and service semantics (e.g. for
QoS) that can ease the cooperation among carriers. Similarly, there is no agreement on possible
reference model(s) for sharing more detailed Traffic Engineering topology information, that can
provide data beyond the rough endpoint reachability but still preserving the confidentiality of
the carrier’s internal infrastructure.
A sibling challenge in this context is the increasing interest by carriers to operate multiregion/multi-layer equipments (i.e. supporting different switching technologies), either by one
single vendor or by multiple ones, under a single control plane instance. This challenge is
relevant for both homogeneous technology networks applying proprietary control plane
extensions (e.g. for WSON GMPLS), and for heterogeneous technology networks (e.g. MPLS
and GMPLS). Nowadays, network operators are often forced to design their control domains
that directly map one specific vendor technology, thus interfacing to “black-box” GMPLS
systems at the management plane and with limited functionalities. The evolution of the control
plane architecture should allow a more in-depth control of the control plane processes, in
particular for what concerns the route computation and the resource allocation policies.
A potential approach to this problem area is in researching modes for an effective “splitting
of the control plane architecture”, i.e. moving some of the intelligence out of the GMPLS
systems towards the Operation and Support System (OSS). Key rationale for the split approach
20
is the possibility to more strictly correlate the routing decisions taken by the control plane with
the systems where the operators set and manage their Traffic Engineering policies (e.g. for
provisioning and planning). The bridge between the enlarged OSS “brain” and the GMPLS
“arm” is provided by the Path Computation Element (PCE) architecture, both in terms of
established and developing standards (e.g. hierarchical PCE).
The GMPLS and PCE architectures are powerful and flexible enough to allow throttling the
boundary between the “arm” and the “brain” according to a variety of splitting points. For
example, an interesting solution for network operators who need self-defined procedures to
route circuits across their network is to maintain a centralized paradigm for the actual service
provisioning, by means of a Network Management System augmented via a stateful PCE
(usually referred to as the “nominal wavelength service provisioning”), while using the
distributed GMPLS signalling and routing for any subsequent fast recovery mechanisms.
5.1.2. Decoupling of the optical transport from the control
plane(s)
Another possible application of the “split architecture” concept can be the integration of
devices by different vendors in a single control framework. In this case, the splitting point can
be much lower in the architecture, i.e. right above the node hardware and the related node agent.
The key goal is to decouple the control plane implementation and procedures from equipments,
with the main rationale of “moving intelligence out of the box”, and making it vendorindependent. In this perspective, this “external” control plane is the unifying glue for
provisioning, recovery and traffic engineering procedures across different vendors within the
same operated network. This approach relies on the assumption that vertical interoperability
between the vendor-independent intelligent entity and the node devices is more streamlined
with simpler interfaces (low level operations and application programming interfaces) and based
on having, for example, unique or centralized points of deployment. One of the potential
enablers in this research area is the popular OpenFlow protocol and its Software Defined
Networking (SDN) framework. The major applicability areas for OpenFlow are currently the
connectionless IP or MPLS-controlled networks, i.e. it is confined at the edge/aggregation (from
campus up to metropolitan networks); however, there is an emerging interest towards
developing its adaptation for circuit switched networks, and in particular for wavelength
switched optical networks.
Despite of the technical impacts of the aforementioned “opening” and “splitting” trends, the
separation of intelligence layers (“brain”) and convenience layers (“arm”) can also generate a
business impact. In fact, they allow traditional third party players (e.g. software houses, stack
vendors) and network customers (network administrators, but also Over-The-Top with large
DCs and network operators) to participate actively in network operations and control, with the
possibility to introduce new business actors and market dynamics.
5.1.3. Enhanced User to Network Interfaces (UNI)
Network operators have often been traditionally “hostile” towards dynamic UNIs, motivating
this approach with the increased management complexity that results from the injection of
customer-driven states (i.e. circuits) within their network and not under their direct control.
Nevertheless, many emerging end-user systems require a better integration with the network
provisioning procedures for on-demand and tailored connection services. Examples of these
21
systems are cloud computing and Service Oriented Architectures (SOA) at large, which all rely
on the network as a vital commodity and could highly benefit in treating it as an integrated
resource within their orchestration processes. An ever-increasing number of distributed (super-)
computing applications have highly-demanding requirements for dynamicity and flexibility in
network and Information Technology (IT) resource control (e.g. automated scaling up/down),
but their network service(s) is still treated as “always-on” and much more static in nature. Their
application layer is unable to exploit the automatic control potentialities of the current optical
(and not-optical) network technologies, thus resulting in inefficient resource utilization in the
network, above all in case of fault recovery. This all points towards a network interface beyond
the traditional UNI, and specifically towards Cloud/Service-to-Network interfaces with
generalized semantics to integrate the characteristics of both IT sites/resources and network
nodes (i.e. resource types, capabilities and availabilities, sites, attached services, capabilities and
capacities of network, computing and storage elements, etc.). These more powerful interfaces
should go closer to the cloud “way of thinking” about the network resources (of which the
circuit is just the ultimate service instantiation), and support a number of advanced components,
such as workflow descriptions, interaction properties, Service Level Agreements /
Specifications (SLAs/SLSes), AA credentials, security contexts and accounting models.
5.2. Management plane evolution: From rigidness to
programmable management
One of the main roles of Network Management is to ensure that the services provided by the
network are offered to the clients with the desired level of performance, quality, and
availability, usually based on a Service Level Agreement (SLA). Typical functions of network
management are network provisioning, fault management, and performance monitoring, which
are handled by Network Management Systems (NMS’s) that embed the capability to be
customized to the different network equipment. Unfortunately, the overall set of management
functions have been developed mainly on a per layer (network technology) basis. Thus, IP
networks are typically managed through customized systems and individual tools, such as HP
OpenView, IBM Tivoli, OpenNMS or Nagios. However none of these tools provides support
for all potential providers’ needs and requirements, mainly because of the lack of programmable
features, the lack of well set standard protocols for network device configuration, the lack of
consensus around the preferred protocols and especially in terms of defining uniform data
models, the high cost of commercial tools and the limited capabilities of the open source tools,
what all in one also conducts to multiple interoperability issues between the IP network
management systems with other management systems. On the other hand, the transport network
is dominated by NMS’s particular to each vendor, where MTOSI appears as the interface to
communicate in a standard way to different NMS’s. The main limitation on a Transport
Network Management System is certainly the level of integration of management capabilities
for devices that operate at different layers than the Transport System. Thus, the management of
a typical operator’s network with many layers is based on separate ecosystems composed of
different NMS’s and isolated tools, without easy interaction between them. In fact, the interrelation between different layers is kept by in-house systems and databases that are hard to
develop and maintain. Moreover, operations involving several layers are full of manual steps
and end up in long and costly processes. This isolation leads to high operational costs, lack of
interoperability and a continued need of upgrades in different systems, what definitely drives to
a non-desired management scenario, hence requiring innovative solutions to optimize the
overall network management process.
22
In order to reduce the complexity of managing a network with multiple layers, two basic
directions can be followed, namely simplification and/or coordination. The former aims at
reducing as much as possible the complexity of the network by a flattering of the layers, while
the latter refers to the development of tools that can interact with the existing network
management systems at the different layers so as to make them work in concert. Any of these
approaches requires extensive research to face the following challenges:






Interoperability between different NMS’s.
Lack of coordination between layers (L1, L2, L3).
Lack of standards and lack of consensus on interfaces and protocols, especially in terms of
defining uniform data models (for NETCONF, MTOSI, etc).
The need to reduce the complexity and duplication of network devices and roles.
The need to reduce manual and error prone intervention as much as possible.
Lack of programmable features allowing providers to compose and orchestrate a set of
operations as a result of an event or a pre-defined policy in the network.
5.3. Evolution in Optical Networks towards cognitive and
self-managed networks and its impact on control and
management planes
Next generation optical networks will progressively deploy cognitive technologies, becoming
cognitive optical networks. In short this term refers to networks that are able to learn, optimize
and adapt themselves in reaction to state changes with little to no (operator) intervention.
Clearly, the adoption of such technologies will have strong implications and impact on the data,
control and management planes. It is noteworthy that Cognitive Optical Networks are becoming
feasible thanks to the adaptive capabilities of both hardware and software components. Specific
examples of dynamic adaptation involve optical transmission (with software-defined /cognitive
transceivers with learning-capabilities) as well as optical transport (with cognitive framing and
encapsulation) and optical switching (with self-flexible and adaptive on demand switching,
leveraging the new grid-less spectrum management paradigms and approaches). Current and in
development technology capabilities, such as format transparent wavelength or signal format
conversion, regeneration or network-wide optical frequency/time/phase determination, will
support the realisation of such cognitive functions. Moreover, hardware programmable elements
could be also deployed to turn state-of-the art optical modules into cognitive-enabled optical
system.
Further research and development should be focused on developing an open platform to
dynamically re-purpose, evolve, self-adapt and self-optimize functions/devices/systems of the
optical network infrastructure. An open platform for these optical/opto-electronic technologies
would allow for environment-aware, self-x systems that can change any parameter based on
interaction with the environment with or without user assistance. This platform would need to
interact with both the control and management planes, potentially requiring either
adaptations/extensions of the current framework or even radically different new approaches.
New control and management plane architectures, protocols and algorithms should support
highly flexible cognitive future optical infrastructure in a heterogeneous optical environment
(i.e., an environment where the cognitive capabilities of its components is heterogeneous). In
particular, research on cognitive control and management plane should be carried out to enable
23
network-wide infrastructure dynamic self-adaptation, self-handling across heterogeneous
systems, and should target both a) a framework that considers optimized multi-dimension
(frequency, time, space) resource allocation, control and provisioning with self-healing and
evolvable operations; and b) evolvable and open control plane platforms based on modular
structures with environmental-awareness utilizing a mix of self-x or user-x controls (i.e. driven
by the user or driven by the network in the normal course of its own cognitive capabilities). This
research should provide a balance between minimal control and management overheads and yet
deliver a trustworthy environment of multi-operator, multi-domain contexts is of critical
importance.
6. Energy efficiency and Green networking
The steadily rising energy cost and the need to reduce the global greenhouse gas emissions
have turned energy into one of the primary technological challenges. Information and
Communications Technology (ICT) in general and optical technologies in particular, are
expected to play a major active role in the reduction of the world-wide energy requirements.
Indeed, recent studies show that ICT is today responsible for a fraction of the world energy
consumption of about 4%, a percentage expected to double in the next decade. This evolution is
illustrated in Error! Reference source not found. [1], [2], where the reported data are based on a
‘business-as-today’ scenario, i.e. assuming that energy-efficiency efforts from industry,
regulation and consumers will remain similar to these of the past years. Error! Reference source
not found. also shows that there is no specific sector dominating the ICT power consumption,
indicating that the need for energy-efficient solutions is relevant to all ICT sectors, spanning
from DCs to network devices and to users appliances.
Figure 3 – Estimation of ICT energy consumption evolution
Currently, access networks are responsible for a major part of the network power
consumption, as access related devices, although consume less power than those in the core
network, are deployed in much higher quantities. To date, fixed access networks are mainly
implemented, by copper-based technologies such as ADSL and VDSL. However, to address the
rapidly increasing broadband access penetration and the new and emerging services, access
technologies such as fibre to the X (FTTX) or even fibre to the home (FTTH) are becoming
available to the end users. This adoption of energy efficient fibre based technologies is expected
to limit the energy consumption in the access network segment despite the heavily increased
capacity. Recent studies (e.g. [3]) also predict that the power share of the metro and core
network segments will grow rapidly. This is due to the dramatic increase in the traffic expected
to be supported by these network segments and to the fact that although they deploy energy
24
efficient optical transmission technologies, they rely heavily on traditional electronic devices for
switching and routing functions. Electronic devices consume high power and their consumption
increases in a nonlinear fashion with the bit rate [4]. It is therefore critical that energy efficiency
considerations are applied in the design, implementation and operation of these networks.
Optical networking can play a key role towards the support of energy efficient and hence
sustainable future ICT solutions. The level of energy efficiency that can be achieved is very
much dependent on the specific architectural approaches that will be followed, the technology
choices that will be made, as well as the use of suitable planning/routing algorithms and service
provisioning schemes. In this context, it is also important to design and operate optical networks
taking into consideration the details of the services and applications that they support as well as
the end devices they interconnect, as considering the relevant specificities and constraints can
have a direct impact on the overall energy efficiency of the infrastructure.
More precisely, at the equipment level it is important to assess if, when and where optical
technologies can be more energy-friendly than electronics, not only considering new or
enhanced low-power devices but also fully re-designing network node architectures to exploit at
best optical component features. Hybrid opto-electronic design can be an important asset
especially in the medium term to also ensure graceful upgradeability. In terms of network
architectures two strategies should be considered: to gracefully upgrade current infrastructures
on one hand and to design new clean slate network paradigms on the other hand. This includes
re-discussing the bandwidth efficient but energy hungry packet switching paradigm vs the
circuit or burst switching techniques, which seem better suited for optical technologies, and
understanding the trade-off between lightpath provisioning compared to hop-by-hop electronic
switching. Improving the engineering practice at the design, planning and operational level
includes redefinition of energy aware management paradigms, introduction of new simpler
protocols, definition of energy friendly resilience schemes including the possibility of quickly
switching on and off devices upon failures, as well as support of planning and routing
algorithms which reduce the network overprovisioning to enhance energy features. Finally,
attempts to match the characteristics of currently popular applications such as P2P, grid or cloud
services to the underlying optical-based network infrastructure can further enhance energy
savings in future network infrastructures, for operators, service providers and users.
7. Standardisation
Research projects bring relevant solutions to its application fields. However, translation from
research to industry is a slow and difficult path that unfortunately remains uncompleted most of
the times. Standardization is however key for the industrialization of research solutions. It takes
time and money for particular solutions to reach a global market that increasingly tends to
replace silos by open solutions. Standardization carries this demanded openness by
interoperability among the different vendors/providers solution with the cost reduction of mass
production (operators can purchase equipments from any vendors/providers; vendors/providers
can sell equipments to any operators).
Nevertheless, in the current model of standardization followed by EC Research Projects it is
hardly manageable to standardize results within the meantime of the project. The limited
timeframe of a project related to the timeframe of a standardization process, tied to the fact that
standardization efforts are commonly launched at advanced laps of the project, makes difficult
25
the success of the standardization of a particular solution. An inter-project common
standardization strategy will pave the way to overcome these difficulties by increasing the
influence of European projects and consortiums on standardization bodies while providing a
long term presence during the whole standardization process. The current standardization
strategy covers three fronts, which from now and on will be coordinated by the CaON cluster.
7.1. Optical data plane technology
The driving standardization body in this case is the ITU-T. There are several EC Research
Projects dedicating efforts to standardization addressing the encompassed technologies.



Photonic Access: The FSAN standardization ITU-T Task Force is focused in the
standardization of NG-PON2 technologies (as a "Disruptive" NG–PON technology with
no strict requirement in terms of coexistence with GPON on the same ODN), where the
SARDANA, ACCORDANCE and FIVER European projects are participating. The
ALPHA European project is working on the standardization of WDM-10G TDM PON.
Metro and Core: The efforts are centered High Speed Transmission (100G+), Flexi-grid
technologies and MPLS/photonic integration. The STRONGEST European project has
presence in these standardization works.
Power efficiency: The STRONGEST European project is working on it for Metro and
Core networks while the TREND European project is active on Access networks.
With regards to the access, the standardization process is an ongoing work, the groups are
entitled to consolidate their efforts and remain very active in proposing their solution and
specifications to the standardization bodies (especially for NG-PON2). The time frame of NGPON2 in standardization is shown in Fig. 11
Fig. 11
7.2. Optical control plane
The driving standardization body in this case is the IETF, with two major initiatives to
provide vendor/provider interoperability and automatic end to end connectivity:
26


GMPLS: The practical totality of the projects concerning Metro and Core Networks
adopt GMPLS as the preferred control plane platform due to its wide technology
umbrella it permits: the STRONGEST and GEYSERS projects for wavelength
switching technologies, the MAIN project for sub-wavelength technologies, the ETICS
project for both (covering inter-carrier issues).
PCE: As a key element (enabler of inter-operability) of MPLS and GMPLS networks, it
is also subject to standardization efforts of STRONGEST, MAINS, GEYSERS, ETICS.
7.3. IT and network integration
It is an incipient field with standardization efforts open by different bodies towards the
convergence of the two words.


The OGF is the driving standardization body in this case, organized in several working
and research groups: the MAINS project has precense in the OGF-NSI WG (Network
Service Interface Working Group) and the GEYSERS project in the ISOD-RG
(Infrastructure Services On-Demand Provisioning Research Group).
Meanwhile there is an intend to create a IETF working group on Cross-Stratum
Optimization, where the GEYSERS project may have an important presence.
27
8. References
Energy efficiency
[1]
[2]
[3]
[4]
M. Pickavet, W. Vereecken, S. Demeyer, P. Audenaert, B. Vermeulen, C. Develder, D.
Colle, B. Dhoedt, and P. Demeester, “Worldwide Energy Needs for ICT: the Rise of
Power-Aware Networking,” in IEEE ANTS Conference, Bombay, India, Dec. 2008.
M. Pickavet, R. Van Caenegem, S. Demeyer, P. Audenaert, D. Colle, P. Demeester, R.
Leppla, M. Jaeger, A. Gladisch, H.-M. Foisel, “Energy footprint of ICT,” in Broadband
Europe 2007, Dec. 2007
J. Baliga, K. Hinton, R. S. Tucker, “Energy consumption of the Internet”, proceedings
COIN-ACOFT 2007, Melbourne (Australia), pp 1-3, June 2007
R. Tucker et al., “Energy consumption in IP networks”, in European Conference on
Optical Communication ECOC’2008, Brussels, Sept. 2008.
Section 4
[5]
[6]
[7]
[8]
[9]
[1] D. Chiaroni, et.al., “Demonstration of the Interconnection of Two Optical Packet
Rings with a Hybrid …“, PD3.5, ECOC 2010
[2] F. Vismara, et.al. , “A Comparative Blocking Analysis for Time-Driven-Switched
Optical Networks”, ONDM 2011
[3] M. A. Gonzalez-Ortega, et.al, “LOBS-H: An Enhanced OBS with Wavelength Sharable
Home Circuits”, ICC 2010
[4] B. Wen, et.al, “Routing, wavelength and time-slot-assignment algorithms for
wavelength-routed optical WDM/TDM networks”, ICTON 2010
[5] Dunne, J., "Optical Packet Switch and Transport: A New Metro Platform to Reduce
Costs and Power by 50% to 75% …”, WOBS 2009
[10] [6] G.Zervas et al “Time Shared Optical Network (TSON): A Novel Metro Architecture for
Flexible Multi-Granular Services”. ECOC 2011 Geneve.
[11] [ABOSI09] C.E. Abosi, R. Nejabati, and D. Simeonidou: Design and Development of a
semantic information modelling framework for a service oriented optical Internet.
International Conference on Transparent Optical Networks, 2009. ICTON 09. Azores.
Portugal.
[12] [WILLNER09] A. Willner, C. Barz, J.A. García Espín, J. Ferrer Riera, S. Figuerola and P.
Martini: “Harmony - Advance Reservations in Heterogeneous Multi-domain
Environments”. IFIP TC-6, Networking 2009, Lecture Notes in Computer Science, 2009,
Volume 5550/2009, 871-882, DOI: 10.1007/978-3-642-01399-7_68
28
i
The porjects involved on the CaON cluster are : GEYSERS, ONE,......
29