INSPACE network and system definition and specification€¦ · The control plane functional requirements and the proposed architecture are summarized. A detailed description of the

FP7-ICT-GA 619732

SPATIAL-SPECTRAL FLEXIBLE OPTICAL NETWORKING ENABLING SOLUTIONS FOR A SIMPLIFIED AND EFFICIENT

SDM

SPECIFIC TARGETED RESEARCH PROJECT (STREP) INFORMATION & COMMUNICATION TECHNOLOGIES (ICT)

INSPACE network and system definition and specification

D2.1

Document Type : Deliverable

Dissemination Level : PU

Lead Beneficiary : WONESYS

Contact Person : Jordi Ferre Ferran

[email protected] Tel.:+ 34-935901149

Delivery Due Date : 31/01/2015

Submission date : 17/02/2015

Contributing institutes : FINISAR, AIT, TID, WONE, HUJI, ASTON, CNET

Authors : D. Klonidis, I. Tomkos (AIT), J. F. Ferran (WONE), F. Pederzolli, D.

Siracusa (CREATE-NET), F. Jimenez (TID), D. Marom (HUJI),

This deliverable provides the definitions of the INSPACE system and network approach as well as the

specifications for the enabling subsystems and modules that will be studied and developed in the project. The

goal of this deliverable is to define the characteristics of the reference spatial optical communication system

including : type of transmission solutions that will be developed in WP3; general switching functions and

operations that must be supported by the design of the spatial-spectral flexible switches in WP4 ;and network

planning and routing of final developed nodes in WP5

Ref. Ares(2015)731842 - 20/02/2015

mailto:[email protected]

INSPACE D2.1 INSPACE network and system definition and specification Version 7.0

FP7-ICT-GA 619732 Public Document 2 | 34

Revision History

No. Version Author(s) Date

1 1.0 Jordi Ferre Ferran (WONE) 18/12/14

Comments: Initial Release, ToC

2 2.0 Dimitrios Klonidis (AIT), Jordi Ferre Ferran ( WONE) 14/01/15

Comments: Structure of deliverable sections according to MS2 including information in bullet points. Add description of INSPACE system in first section.

3 3.0 Federico Pederzolli, Domenico Siracusa (CREATE-NET) 20/01/15

Comments: Revised section 4, improved requirements and main definitions, added definitions on control plane architectures and selection.

4 4.0 Felipe Jimenez ( TELEFONICA) 05/02/15

Comments: Revised section 2

5 5.0 Dimitrios Klonidis (AIT) 13/02/15

Comments: Converting deliverable to the INSPACE deliverable template

6 6.0 Dimitrios Klonidis, Ioannis Tomkos (AIT) 16/02/15

Comments: Edit of section 3 and 6. Edit of executive summary. Revision of all sections

7 7.0 Shalva Ben Ezra (Finisar), Christian Sanchez-Costa (ASTON), Dimitrios Klonidis 17/02/15

Comments: Final version after internal revie

Comments:

Participants

The INSPACE Project Consortium groups the following Organizations:

No Partner Name Short Name Country

1 Optronics Technologies S.A. OPT Greece

2 Telefonica Investigation y Desarrollo TDI Spain

3 The Hebrew University of Jerusalem HUJI Israel

4 Research and Education Laboratory in Information Technologies AIT Greece

5 Optoscribe Ltd. OPTOSCRIBE United Kingdom

6 Center for Research and Telecommunication Experimentation for

Networked Communities CN Italy

7 Aston University ASTON United Kingdom

8 Finisar Israel Ltd. FINISAR Israel

10 W-Onesys, S.L. WONE Spain



Executive summary

INSPACE proposes a novel networking approach by extending the established spectral flexibility concepts to

the SDM domain and significantly simplifying the super-channel allocation and control mechanisms, by

removing current limitations related with the wavelength continuity and fragmentation issues. The new

concept utilises the benefits of the high capacity, next generation, few-mode/multi-core fibre infrastructures,

providing also a practical short term solution, since it is directly applicable over the currently installed multi-

fibre cable links. The realisation of INSPACE approach is enabled by the development of novel multi-

dimensional spatial-spectral switching nodes, which are fabricated by extending the designs of the existing

flexible WSS nodes, incorporating advance mode/core adapting techniques. The concept is further supported

by novel processing techniques that minimise the mode/core interference as well as new network planning

algorithms and control plane extensions that are enhanced with the space dimension.

This deliverable summarizes the key system definitions and specifications which guide both the technology

development work in WP3, WP4 and WP5 and the system level studies on the feasibility and benefits of the

INSPACE approach in the remaining tasks of WP2.

The necessity and the expected benefits from the introduction of the space dimensions in optical networking

(in addition to the existing spectral dimension) is discussed first concluding on the possible five channel

allocation options that are examined in INSPACE. These channel allocation options define the networking

approaches and determine the control and network planning solutions. The options have been derived based

on the possible SDM type of fibre media available today as well as the envisioned switching node

functionalities.

Next the reference network scenarios are defined considering primarily a medium scale European national

network (covering the Spanish geography) as a multi-domain optical mesh based infrastructure. This model

represents a typical European “state of the art” optical transmission network. The model can be easily

modified to an extended coverage network with more demanding transmission requirements, in which the

specific transmission needs are assumed to be fulfilled by means of sophisticated signal amplification and

processing. Finally, an example of an inter data centre connectivity scenario is optionally considered as a

Metro deployment where potentially high data volumes need to be transported.

The technology requirements are discussed next focusing on the switching technology capabilities (i.e. the

demultiplexing and processing of SDM contents) and the type of SDM fibre (i.e. the effects that are imposed

by the transmission medium on the multiplexed channels). These two aspects determine the exact definition

of the allocated multi-dimensional channels. The different types of SDM fibres that will be considered in

INSPACE project are reviewed, highlighting mainly the properties of MMF and MCF constituting the two

cases of coupled and uncoupled transmission. A special mention on the newly developed MCF with multi-

mode cores is provided as a prominent medium that allows the implementation of the spatial group switching

design. Moreover, the four possible switching node designs for various SDM networking cases are presented

compared and commented. The first design proposed enables the independent switching of the spatial and

spectral contents in the expense of added complexity. Two simplified designs are extracted from this first

design implementing the spectral and spatial mode switching designs. The fourth design is a combination of a

joint switching design and a spectral-spatial independent switching design, and implements a spatial group

switching node.

The control plane functional requirements and the proposed architecture are summarized. A detailed

description of the control plane requirements, including the interfacing and configuration requirements, is

provided in deliverable D5.1.

Some additional specifications related with practical issues for the node implementation are defined. These

specifications identify the thermal stability, physical stability and heat anticipation requirements and will be

used for the definition of the system footprint and the power consumption, in the relevant studies performed

in task 2.3.

Finally, the deliverable lists the key system specifications and requirements for INSPACE. The targeted

requirements are identified for all cases and where applicable extended requirements are identified. These



requirements are the basis for the development efforts in WP3, WP4 and WP5 as well as the networking

studies that follow in WP2. It is noted that he testing requirements for WP6 activities will be defined in D6.1

according to the final outcomes of the development work and the available resources for the test-bed. The

listed requirements here refer to the system evaluation and definition efforts and act as a guideline for the

node, transmission and control modules under development.



Table of contents

Revision History ................................................................................................................................................ 2

Participants ......................................................................................................................................................... 2

Executive summary ........................................................................................................................................... 3

Table of contents ............................................................................................................................................... 5

1. Introduction - Definition of the Spatial-Spectral Flexible Optical Network .............................. 6

1.1. WDM network evolution: Moving from wavelength reconfigurable to Spatial-Spectral

flexible optical networking .................................................................................................................................. 6

1.2. Channel Allocation Options ................................................................................................................ 8

2. System and network level definitions and requirement ............................................................... 11

2.1. Reference networks ............................................................................................................................ 11

Medium scale National network model ............................................................................................. 11

Extended coverage optical network model ....................................................................................... 14

Metropolitan optical mesh model ....................................................................................................... 15

Conclusions and next step ................................................................................................................... 15

3. Technology requirements ................................................................................................................. 16

3.1. Types and characteristics of optical fibres in support of SDM networking ............................. 16

3.2. Multi-dimensional switching node designs..................................................................................... 17

Node design for space-wavelength granularity ................................................................................. 20

Node design for spatial mode switching across all spectral channels .......................................... 20

Node design for spectral mode switching across all spatial channels .......................................... 21

Node design for Wavelength switching across spatial mode subgroups ..................................... 22

Comparison of WDM-SDM switching alternatives ........................................................................ 22

4. Network Control Plane (WP5 related) ........................................................................................... 24

4.1. Control Plane Functional Requirements ........................................................................................ 24

4.2. Control Plane architecture ................................................................................................................ 25

Architectural Archetypes ...................................................................................................................... 25

5. Node implementation requirements ............................................................................................... 27

5.1. Expected node foot-print and housing requirements .................................................................. 27

5.2. Thermal stability and heat anticipation requirements ................................................................... 27

5.3. Physical stability and vibration issues .............................................................................................. 30

6. Conclusions on the targeted INSPACE system specifications and requirements ................... 32

6.1. Spectrum related requirement .......................................................................................................... 32

6.2. Space related requirements ............................................................................................................... 32

6.3. Super-channel generation requirements ......................................................................................... 33

6.4. Switching requirements ..................................................................................................................... 34



1. Introduction - Definition of the Spatial-Spectral Flexible Optical Network

1.1. WDM network evolution: Moving from wavelength reconfigurable to Spatial-Spectral flexible optical networking

Networks are ubiquitous today, and running them efficiently is a paramount priority. As information

consumption continues to grow unabated, network infrastructure supporting the content delivery, data

exchange, and messaging services has to continuously scale to wider span, longer reach and broader trunks.

Network technology has effectively scaled over the last 40 years through the introduction of various

technologies, starting from Plesiochronous Digital Hierarchy (PDH) switching technologies to current

Internet Protocol/Multiprotocol Label Switching (IP/MPLS) routing over dynamic optical mesh

architectures based on Reconfigurable Optical Add-Drop Multiplexers (ROADMs). However, there are

serious concerns about future growth potential given the maximum transport and switch capacities of

DWDM based core networks.

To demonstrate the abovementioned issue, it is instructive to take note of Figure 1, which plots the end user

(access) delivered data rates as well as the core switchable and transported capacity since the late 1970’s.

These two metrics have historically tracked each other, both growing exponentially with core transport

carrying three orders of magnitude greater capacity over access. Over the next ten years, we can safely

anticipate end-user access to grow from 100Mb/s to 1Gb/s, driven by the generalised deployment of high

capacity fibre based technologies for the access segment. This suggests that the network core will then have

to switch and route 1Tb/s channels. These rates cannot be supported over a single optical signal, due to

performance limitations of optoelectronic components (electronic multiplexers and framers, optical

modulators, ADC/DACs etc.), hence various optical multi-carrier approaches have been proposed, giving

rise to optical super-channels. These spectral super-channels pose several challenges in the context of

networking that are being addressed across the globe, pertaining to the choice of modulation formats

(OFDM, Nyquist, etc.), spectral efficiency and bandwidth utilization, processing complexity, cost and support

for switching with ROADMs. But even the spectral super-channel approach offers a limited growth potential

due to the “capacity crunch” [1], on account of the finite transport capacity of a given single mode fibre core

and the limited optical gain bandwidth [2]. These issues will finally lead to blocking situations for spectral

super-channels [3].

From an operator perspective, it is especially relevant to ensure an extended lifespan for green-field

deployments or network upgrades. The use of advanced modulation formats and new signal structures (like

super-channels) over a flexi-grid enabled optical infrastructure would allow to overcome the resource

exhaustion without a radical change on the network elements, but introducing extra complexity on the

planning and operational tasks. This approach, however, will alleviate the problem for a limited time.

Some additional mechanisms could be used in order to increase the network capacity, but all of them bring

relevant drawbacks.

An optimisation on the current amplification schemes by site reallocation and reduced span lengths would

provide some extra Optical Signal to Noise Ratio (OSNR) margin and allow the use of efficient modulation

formats on a wider scope. This, however, would not be generally feasible and would introduce additional

capital expenditures, operational burdens and reliability issues. As another option, additional L-band

amplification could be considered, but new amplifiers and WSS elements would be needed and operational

complexity will be also increased, all for a factor of two increase. Raman amplification would also be

beneficial for increasing the network OSNR levels but brings security concerns and, again, extra cost and

operational complexity.

1 T.Wu, Time Magazine, March 11th, 2010

2 D.Ellis, J. Zhao, D. Cotter, “Approaching the Non-linear Shannon Limit”, Journal of Lightwave Technology, Vol. 28, No. 4, pp 423 – 433 (2010)

3 A.D.Ellis, F.C.G.Gunning, “Implementation of Tbit/s Networks”, IEEE Photonics Conference 2011, Special Symposium on Terabit Optical Ethernet, paper MW3,

(2011)



Figure 1 - Access capacity (blue) and core switching and transport rate (red) trends over the last 40 years. The two metrics go hand in hand, suggesting the next switching load will be at the 1 Tb/s scale. Solid lines represent long term trends, whilst dashed lines represent likely future evolutions, with access traffic requiring network access at increasingly higher burst rates to support TDM protocols, and core capacities constrained by the nonlinear Shannon limit.

An increase of the network regeneration and grooming levels would respectively increase the available

capacity and its actual usage. The drawbacks are clear in this case, a generalized explicit regeneration will

increment the network costs to a very high extent, poses operational and reliability issues and has a relevant

impact on power consumption. Very high grooming levels would make the need for additional resources at

the ROADMs digital (Optical Transport Network-OTN) matrixes and, in the extreme, for extra resources at

the IP/MPLS layer (with the finest grooming granularity) that would be unnecessary if there were enough

capacity at the optical layer. As a consequence, all the above-mentioned issues for the regeneration case are

also applicable.

The use of low loss/large area fibres would also allow, to some extent, an increased use of transmission

technologies based on higher complexity modulation format as they would potentially increase the OSNR

level at the transponder receivers and diminish the impact of non-linearities. However, they would postpone

the capacity shortage for a limited time and would hardly compensate the deployment investments.

In the cases where transmission cables have spare fibres (not used or planned for transmission to other

destinations or reserved for supervisory tasks) then, they could be used in a parallel fashion to provide higher

capacities between two ROADM nodes. The operator could leverage unused WSS ports or installed higher

port count models to deal with the extra line signals terminating at the nodes. However, this would bring an

extra cost on control and amplification resources and, potentially, on line and add/drop WSS elements and

increases the complexity of the nodes. In the end, as this approach consist of a partial replication of the

transport infrastructure it lacks of the necessary scalability and is not efficient in terms of cost, space and

power consumption. Moreover, it is not generally feasible and, as the other approaches, it cannot be

considered a solution for the long term needs.

From all the above, it is clear that the telecom operators should closely follow the advances on emergent

technologies that, covering all key transport “pieces” (transmission, amplification, switching …) provide a

future proof solution and ensure that long term capacity and flexibility needs will be attended. In this sense,

Spatial Division Multiplexing Techniques are being widely researched and have strong support from the

major system vendors.

The nascent SDM technology offers much wider conduits of information by offering additional means for

transporting optical channels in one of the three following forms: parallel array of single mode fibres (SMF),

multicore fibres (MCF), and few mode fibres (FMF). Each of these solutions has its advantages and

1970 1980 1990 2000 2010 2020

100

1k

10k

100k

1M

10M

100M

1G

10G

100G

1T

10T

Core

Access

Sw

itchab

le c

ap

acity in

core

(b

/s)

Custo

me

r a

ccess r

ate

(b

/s)

Year



disadvantages in transport, amplification technology, and signal transmission impact. However, the

integration of SDM fibre links in networking scenarios has yet to be explored.

The proposed SDM alternatives now coming into play offer additional advantages over the pure increment of

available core capacity due to the many alternative solutions for the switching hardware, granularity, and

control. At one extreme, the transported data can be decomposed to its finest spatial and spectral

constituents at every network core, and switched with an extremely large electrical cross connect, while at the

other extreme the entire data carried on the SDM-WDM link is optically switched in bulk to the next link.

The optimum solution certainly resides somewhere in between. INSPACE shall explore the networking alternatives with SDM links for core networks, where SDM technology will first be introduced,

innovating at all levels from new switching hardware adapted to SDM, through network provisioning and control methodology.

1.2. Channel Allocation Options

The use of the space dimension in addition to the spectrum dimension could potentially lead to different

flexible optical networking options. The five main options that have been identified in INSPACE project are

summarized in Figure 2 and explained in the following paragraphs.

The pure SDM approach (as defined in the physical layer) denotes that at the end of each transmission link a

spatial-demultiplexer is placed prior to a spectral-demultiplexer to process and extract the spatially

multiplexed data. The spectrally flexible networking concept that is studied for common SMF-based systems

can be equally applied to SDM systems, assuming that a spatial-demultiplexing process applies at the end of

the links (Figure 2 – case A). In this case, the space dimension relates with the multiplexing stage of the

physical layer and has no role in networking layers. Therefore, in principle there is no difference with the

spectrally flexible networking concept that spans over multiple parallel links. Evidently, this scheme

represents the natural evolution of today’s transmission systems towards higher capacity systems, considering

the use of links with bundles of SMF. However, it is also applicable to MCF systems with uncoupled cores in

order to avoid cross-talk between cores carrying independently assigned end-to-end spectral SChs.

An alternative could be to use elasticity in space dimension instead of the spectrum dimension (Figure 2 –

case B). Channels can be allocated over a fixed spectral grid (as in the case of WDM systems) and may have

the ability to flexibly expand over some or all of the modes/cores in the fibre. For example a 100Gb/s DP-

QPSK demand can be allocated inside a 50GHz grid and occupy one core in a MCF, while a 400Gb/s

demand is allocated over 4 cores inside the same grid. Networking can then rely on common WDM

allocation and routing schemes. Also the transmitter and receiver designs are simplified due the use of a

common laser source at the transmitter site and the local oscillator site. In this case, the spatial elasticity

makes the spectral elasticity requirements obsolete. Moreover, a pure spatially flexible optical network

maintains the network planning and operation simplicity of fixed grid WDM networks, avoiding issues like

spectral fragmentation and complex resource optimization and control, as in the case of spectrally flexible

networks. However, this networking solution is not capable to make real use of the capacity increase benefits

of SDM. Limited resource utilization is expected due to the required spectral gaps between adjacent fixed-

grid channel and mainly due to the unutilized spatial resources from low capacity demand. The scheme is

applicable to any type of MMF or MCF with coupled or uncoupled cores.

A more efficient scheme compared to the pure spatially flexible networking concept can be proposed by

allowing spectrally flexible contents to be allocated over the space dimension (Figure 2 – case C). This

approach leads to the flex-grid SDM scheme and is primarily enabled by the novel joint spatial switching

elements proposed in INSPACE project. From the networking point of view the network planning and

operation complexity follows that of spectrally flexible networks with the main difference being in the

allocation options for the SChs. Similar to a spectrally flexible approach, the scheme also allows the

adaptation of the signal format (i.e. bandwidth) to the transmission reach target, which can be an important

issue for SDM transmission due to cross-channel interference in MMFs and MCF with coupled cores. The

flex-grid SDM solution is expected to have better resource utilization than the pure spatially flexible (i.e.

fixed-grid SDM) solution for large capacity demands due to the efficient use of the spectral resources,

however the optimum use of the spatial resources still remains an issue.



Figure 2 - The different flexible optical networking options offered with the use of the space dimension in addition to the spectrum dimension

Evidently, the most flexible networking approach is to combine elasticity in both spectrum and space

dimensions (Figure 2 – case D). In this case, the available end-to-end resources could be identified over a

pool of spatial and spectral slots in the links that form the end-to-end path, assuming that innovative multi-

dimensional nodes are in place able to perform switching in both dimensions independently. The additional

use of the space dimension is expected to increase significantly the complexity of the network planning and

resource optimization algorithms compared to the spectrally flexible networking approach, but can potentially

lead to significant benefits. Spectral fragmentation issues (see next section) can be resolved via switching of

the fragmented spectral slots in the space dimension, while also new approaches for network resource

virtualization and content aware IP mapping over the optical layer can be adopted with the spatial allocation

of virtualized network segments. The scheme requires the use of MCFs with uncoupled cores to achieve

independent switching of cores in the space dimension.

Finally, an alternative scheme for combined spectral and spatial networking could consider the allocation of

SChs on independent groups of cores or modes rather than individual ones (Figure 2 – case E). This case can

be seen as a combination of cases C and D presented above. The spatial groups can be treated as a single

switching entity either if they are fully utilized or not (as in case D), while spectral elasticity is maintained

within them (as in case C). This approach could potentially relax the technology related design complexity of



the node (i.e. by reducing the number of required spatial ports), in the expense though of some resource

underutilization compared to the previous case. The scheme is applicable to MCF with uncoupled cores and

possibly to MCF with coupled cores assuming that the spatial groups are properly allocated over distant cores

with minimum interference between them. Moreover, this scheme could be attractive in particular for the

case of the newly developed MCF with few-mode cores (FM-MCF) [4], where each one of the few mode

cores is a spatial group that can be handled independently to the rest of the fibre cores.

It is noted that the combined spectral and spatial flexible networking is a new research topic and significant

research effort is required in order to identify and quantify its full potentials and complexity limitations.

Moreover, a detailed evaluation of the different flexibility schemes presented above is currently missing from

literature and is one of the key goals of INSPACE project.

4 Y. Sasaki, et al., "Dynamic multidimensional optical networking based on spatial and spectral processing," Optics Express, vol. 20, no. 26, pp. B77-B84, Dec. 2012.



2. System and network level definitions and requirement

This section describes the reference networks and system level characteristics that will be used for the

network level studies, the techno-economic and energy consumption analysis and the feasibility studies in

WP2.

2.1. Reference networks

Several exemplary networks, resembling typical operator deployments are defined for SDM evaluation. For

the sake of model classification, the network scope/segment covered by each model will be used as a

parameter. Only terrestrial networks will be considered at this point, in order to put the focus on a limited set

of network scenarios.

- First, a medium scale European national network (covering the Spanish geography) is defined as a

multi-domain optical mesh based infrastructure. This model is defined with great detail and it can be

considered a typical European “state of the art” optical transmission network.

- Secondly, an example of extended coverage network with more demanding transmission

requirements is presented. The specific needs for such type of deployments are currently fulfilled by

means of sophisticated signal amplification procedures.

- Finally, an example of an inter data centre connectivity scenario will be included as a Metro

deployment where potentially high data volumes need to be transported.

A Pan-European network can be considered, in practice, a collection of multiple interconnected networks,

with implicit regeneration points and, as such, it is already covered by the first (national, single domain)

defined model.

Medium scale National network model

It is assumed in this case that the infrastructure is owned by the incumbent operator, with abundance of

transmission media (optical, radio, etc.) resources as well as appropriate premises for Telco equipment.

Moderate capacity structures based on CWDM or low speed DWDM (usually rings, but optical meshes are

becoming common) are normally supporting the Metro aggregation network, implementing both grooming

and switching at higher layers such as IP or Ethernet.

The application of INSPACE technologies is mainly focused on the support of high capacity pipes at the

regional or national levels, being both covered by the model defined in this document.

The network model is based on a hierarchical structure with 5 meshed regional domains and one “express”

national domain. Domain separation allows deploying multiple vendor technologies and avoids potential

scalability limitations when distributed protocols and dynamic procedures are involved in the optical path

setup. In that sense, network partition isolates regions from relevant failures occurring in the other domains

and helps maintaining the network stability.

Intra-regional communications make use of resources within the specific region, while inter-regional

communications may be established using some inter-regional links or through a dedicated (express) national

domain, whose ROADMs are collocated with some selected ROADMs at each regional domain. The use of

an express domain reduces the maximum number of concatenated filters of the any potential optical path,

thus increasing the number of feasible routes.

The network topology is shown in Figure 3 for regional domains and Figure 4 for national domain.



Figure 3 - Telefonica reference network. Regional domain.

Figure 4 - Telefonica reference network. National domain.



Directionless capabilities will be assumed at both regional and national domains, meaning that for all

Add/Drop nodes any added signal can be directed to any possible direction.

Colourless capabilities will be generally available at the national, regional or national + regional domains.

Coloured architectures are preferred cost wise, but they offer much lower flexibility and, in consequence,

increase the network operational expenses. Moreover, coherent technologies allow minimizing the optical

filtering at the Add/Drop nodes and, thus, make the cost of colourless architectures much more attractive

than in previous traditional (direct detection) networks.

It is considered that the network model is based on a “Broadcast and Select” architecture, with LCOS based

ROADMs with up to 9 degrees and using cascaded Nx1 WSS blocks for the colourless tributary side at Rx.

However, “Route and Select” architectures could be needed in the short term to cope with either high loss if

maximum ROADM degree increases or extended isolation ratios become necessary. Similarly, new

implementations would be needed in the future for the Add/Drop, such as MxN WSS, or Optical Multicast

switches. Also, the optical slots/channels will follow a fixed grid approach, having in mind that Flexi-grid

alternatives could bring advantages in terms of spectral efficiency that will make them an interesting option

for the future.

In the network model, it is envisaged that uncompensated schemes, i.e. DCM free, are always used in the

national domain, that means that at only coherent technologies will be considered in that domain.

Compensated schemes and “legacy” direct detection technologies are allowed in the regional domains,

although there is a trend to deploy coherent detection at every network segment.

Regarding amplification strategies, only C-band EDFA amplifiers are considered all along the Core optical

network. Hybrid EDFA and Raman amplifiers could be selectively deployed at some segment in the network

where better OSNR values are needed, but this will happen very seldom.

SMF will be used on all the network segments and links. It is assumed that oldest fibre layouts have been

renewed and the fibre links are in a relatively good state.

The fibre characteristics are:

Type: SMF G.652-D

Chromatic dispersion at 1550nm: 16.5 ps/nm*Km

Chromatic dispersion slope at 1550 nm: 0.057ps/nm2*Km

Attenuation coefficient: 0.22 dB/km.

PMD coefficient: the maximum value for a fibre link is 0.3995 ps/sqr(Km), while the minimum is

0.0213 ps/sqr(km).

The inter-amplifier links will be made of multiple segments of 2Km length. We consider an average value for

a good splice of 0.05 dB and an extra loss of 2dB coming from the different connectors between amplifiers

(at least 4 needed to connect amplifier and line cards at both ends).

The power management is based on a constant power per channel scheme, leveraging on WSS or standalone

VOAs for channel equalization.

For the amplifier deployment within the reference network, we will define an initial scenario and include

other potential locations (exchanges or cabinets) where additional amplifiers could be installed if needed. As

indicated at the beginning of the section, the model assumes the operator has plenty of transmission

resources and a great number of potential locations for active equipment.

The defined scenario tries to improve the OSNR levels across the network links, while following some

practical guidelines:

When the total link attenuation (fiber+splices+amplifier connectors...) is below or equal to 25dB no

inline amplifiers will be deployed.

When the total link attenuation fits in the range of 25dB to 40 dB, one inline amplifier will be

installed in the link at some location close to the middle of the link.



When the total link attenuation fits in the range of 40dB to 60 dB, twoinline amplifiers will be

installed in the link, at approximately regular distance intervals.

When total link attenuation is above 60dB, the procedure described above is followed for every 20dB

attenuation interval.

There will be some additional locations, which can be potentially used for in-line amplifiers.

Those locations are defined following these criteria:

There will be no additional facilities (exchanges, cabinets, etc, where in-line amplifiers could be

placed) for link lengths with attenuation values lower than 15dB.

Whenever the link attenuation is higher than 15dB, there is a possible amplifier location at an

equivalent distance of “Y” dB. The value of “Y” is taken randomly from a normal distribution with

average 10dB and standard deviation of 0.4 dB.

Extended coverage optical network model

This network model would apply to networks covering very extensive regions, where fiber resources are

usually scarce and deployment conditions are not always as good as in the previous case, representative of a

medium scale European country. Also, Telecom operators do not usually own their entire transmission

infrastructure, but rely on swapping agreements in terms of either individual fiber (within a cable) or capacity

(optical channels, subwavelength structures, etc.). By means of those agreements they can extend their

network coverage to the locations where they do not have a broad base of customers, but with some

potential limitations in terms of bandwidth or performance.

In this scenario, a limited number of relevant locations (where Internet exchange points reside) would usually

collect traffic from lower population areas. Traffic demands would then be pretty high on some few Metro

networks (main cities) and, up to some level, in some of the connections from small/medium size cities to

the Internet PoPs. Those connections could be supported over 100Gbps coherent technologies, able to reach

more than 1500Km without a need for regeneration.

Direct detection technologies, being cheaper, would be in principle more appropriate for lower capacity

connections, but in many situations coherent and DSP assisted technologies are needed just to overcome the

high physical impairments (such as dispersion) that are expected in some of the fiber links. Also, fiber layout

can be subject to frequent movements coming from road works or natural disasters that affect transmission

conditions in a relevant manner. In some other situations, sophisticated amplification mechanisms e.g. based

on Raman plus remote optical pumping must be needed due to the absence of any power supply along a very

high span link. As an example, fiber can be laid over power wires, with spans reaching even 350 Km up to the

sites at transformation stations were optical signals can be amplified.

The following simplified network diagram shown in illustrates the concept, where optical nodes are

interconnected through a variety of links, based on simple EDFA amplification when distances are moderate,

to co-propagating or counter-propagating RAMAN (with optionally EDFA) for longer distances to RAMAN

with remote optically pumped schemes for the most demanding links.

Figure 5 - View of a network diagram wher optical nodes are interconnected through a variety of links.



Metropolitan optical mesh model

This network model would apply to Metropolitan environments in populated areas with high aggregate data

loads coming from advanced residential services, corporate communications, inter datacenter flows, etc.

The dynamicity and flexibility levels of these scenarios are not as high as in the regional or national optical

meshes, so that they can be based on directional and colored low degree ROADMs (typically up to 4 degrees)

with outer rings based on 2 degrees FOADMs. The network will be carrying a mix of traditional 10Gbps

signals and 40Gbps/100Gbps with direct detection.

The number of optical nodes will be around 10/15, with link distances between 1 and 50 Km and low gain

EDFA or even no signal amplification. Tributaries will be of multiple kinds, like GbE, Fiberchannel or SDH

circuits, further groomed into higher speed optical signals.

For some scenarios, like those related to inter datacenter communications, latency can be a critical issue in

this environment. This can have an impact on the nature of the optical signals (like the FEC) or the

dispersion compensation mechanisms (FBG preferred over DCF).

The network model initially considers, only, directional and colored architectures. However, the deployment

of colorless architectures will be driven by the generalization of optimized Add/Drop blocks based on power

splitting instead of selective filtering, derived from coherent reception with implicit optical filtering.

Conclusions and next step

The applicability of INSPACE technologies seems to be more straightforward in the first (medium scale

regional/national networks) and third network scenario (Metro environments in highly populated areas)

where there is a need to support very high traffic loads and the deployment conditions are not extremely

requiring.

The telecom operators are currently investing high amounts of money on new fiber layouts, for the sake of

renewal of old fiber infrastructure or to avoid the capacity limitations of current radio links. Those

investments have to be paid by the services revenues along an extended period of time, meaning that the fiber

capacity has to absorb the future traffic demands for many years to make the investments profitable. This can

be considered a critical advantage of INSPACE implementations, as the potential capacity of the transmission

media gets multiplied. The same advantage, although not so critical, will be found with the active equipment

comprising the end-to-end transmission link, which will allow reducing equipment updates and the related

migration costs.

The deployment of optical mesh structures is becoming widespread in most scenarios, even in Metropolitan

environments traditionally based on rings. The transmission and switching capacity of the optical nodes is

growing accordingly, with higher numbers of channels per span and higher ROADM degrees (for line and

Add/Drop side). The INSPACE developments will help maintain this trend and ensure the support of the

ever increasing service requirements without replicating the network infrastructure and increasing the capital

and operational expenses.



3. Technology requirements

The exact definition of the allocated multi-dimensional channels depends on the switching technology

capabilities (i.e. the demultiplexing and processing of SDM contents) and the type of SDM fibre (i.e. the

effects that are imposed by the transmission medium on the multiplexed channels).

This section reviews the types of SDM fibres that will be considered in INSPACE project and provides the

possible switching node designs for various SDM networking cases as these defined in subsection 1.2.

3.1. Types and characteristics of optical fibres in support of SDM networking

Optical fibres in support of SDM transmission may come in many forms (see Figure 6). Single-mode fibre

(SMF) is designed to allow light guiding of a single spatial mode in the core region by tailoring the refractive

index profile and core dimensions. Multiple cores can be placed within a single fibre cladding, forming a

multi-core fibre (MCF), with each core now supporting a single spatial mode. Hence MCF offers a capacity

multiplier equal to the core count. Alternatively, the core dimensions or the refractive index contrast can be

modified to support additional optically-guided spatial modes, which closely resemble Laguerre-Gaussian

modes. These few-mode fibres (FMF) offer a capacity multiplier equal to the mode count. A somewhat more

exotic fibre design is that of the annular core fibre (ACF), which supports multiple spatial modes confined to

the annular core region. The annular structure is designed to support a single radial mode and multiple

azimuthal modes.

Figure 6 - Different types of fibres in support of SDM, showing their geometrical form and propagating spatial mode distributions. (A) Single-mode fibre, (B) Multi-core fibre, (C) Few-mode fibre, (D) Annular-core fibre, (E) Multi-core supporting few modes, and (F) Multi-core arranged in coupled subgroups.

One of the key differentiating metrics between SDM supporting fibres is whether the modes remain

uncoupled in transmission or potentially may be coupled due to manufacturing imperfections, as well as

environmental effects such as bends, stress and temperature gradients. Coupled transmission implies that the

modes intermix, yet the information is maintained within the set of modes. However, coupled transmission

does require that these mixed modes must remain together in order to unravel the mixing at the receiver

using digital signal processing. FMF and ACF are inherently prone to mixing as the modes spatially overlap

and hence are categorized as a coupled SDM transmission medium. Even though in a simple MCF the cores

are distinct, they may still couple if the cores are closely packed; conversely, the MCF can be specifically

designed to remain uncoupled. One flavour of SDM purports to use an array of existing SMF; obviously this

SDM solution can be categorized as uncoupled as the fibres are separated. The SDM fibre can also be

designed to support multiple coupled spatial mode subgroups, yet have no coupling between the groups. Two

such examples are a FMF-MCF hybrid, where there are uncoupled multiple higher index cores and each core



supports several spatial modes, and a MCF design with an uneven spacing allowing only the closely packed

cores to couple.

The adoption of new SDM-supporting fibres in the optical network potentially increases the capacity per

fibre by factor M, the number of guided spatial modes. Considering each fibre mode still spectrally spans the

optical communication band, then WDM can be applied, carrying N wavelength channels per spatial mode.

Hence the WDM-SDM fibre capacity can be defined by a two-dimensional array, with wavelengths (λ1, …,

λN) and spatial modes (σ1, …, σM) defining its columns and rows (see Figure 7).

While there are many SDM fibre alternatives that can be considered for implementing future WDM-SDM

optical networks, for mesh node switching purposes, they can be categorized into three general archetypes:

1. Uncoupled spatial modes: Spatial channels remain distinct in fibre propagation and all ancillary network equipment, as would be experienced in uncoupled MCF or an SMF bundle. Therefore, individual spatial modes can be switched from one SDM fibre link onto another or add/drop operations applied for any spatial mode/wavelength channel combination.

2. Coupled spatial modes: Spatial channels mix throughout fibre transmission, as occurring in FMF and coupled MCF. As a result of the channel mixing, multiple-input, multiple-output (MIMO) processing is required in order to unravel the mixed information, which occurs after coherent detection of all the spatial modes at the SDM receiver. Since the information is mixed across all spatial modes, modes cannot be separated for switching to other destinations or information loss will be experienced. A complete MIMO receiver has to be employed in order to separate the SDM data, an operation performed at the channel destination, and not desirable at every mesh network node as network transparency will be lost.

3. Coupled spatial subgroups: Spatial modes may mix only within subgroups of the total spatial mode count. The subgroups are defined by the SDM fibre design, and the spatial modes belonging to a subgroup must not be separated in switching operation. Subdividing the spatial modes into subgroups of smaller size eases the switching limitations with respect to full mode coupling, while not reaching the full flexibility of uncoupled modes.

Figure 7 - Parsing the fibre’s SDM and WDM channels for switching, where space modes (σ1, …, σM) and wavelength channels (λ1, …, λN) are fully utilized. (A) Each mode/wavelength channel can be independently switched, (B) Switching performed on mode basis across all wavelengths, (C) Switching performed on wavelength basis across all modes, and (D) Switching performed on wavelength basis and spatial mode sub-groups.

3.2. Multi-dimensional switching node designs

A multi-dimensional switching node may use all three granularity levels (space, wavelength, and time) or a

subset of two of them depending on the technology capabilities. Utilizing the time domain requires active

components and hence is reserved to IP routers that can utilize burst type communication or TDM leading to

significant energy inefficiency. The EU project INSPACE will focus on the introduction of SDM at the



optical level, where information can be routed transparently and most efficiently. However, transitioning to

an optical network based on high capacity SDM fibre links is still far off and entails resolving many issues

related to its physical-layer implementation, such as identifying the best SDM fibre option, the optical

amplification means, the efficient space multiplexing and demultiplexing methods, and in particular the all-

optical switching approaches that should occur at network nodes where information-bearing channels have to

be routed towards their destinations within an optical mesh topology.

At the optical network nodes, the WDM-SDM traffic on each inbound fibre link has to be either redirected

to outbound fibre links as part of the network information flow, or dropped to receivers for local

consumption at the node’s geographical location. Additional data is typically reintroduced, or added, in its

place originating at clients associated with the node. The SDM fibre categories dictate the permissible

switching operations that can transpire at the optical network nodes. We identify four alternative switching

scenarios which strongly relate to the SDM fibre categories, forming multi-dimensional switching nodes of

different granularities (i.e. as discussed before):

A. Independent spatial mode/wavelength channel switching (space-wavelength granularity): The WDM-

SDM fibre capacity can be switched independently for every spatial mode and wavelength

combination. This forms the finest switching capacity granularity, leading to the greatest flexibility at

the cost of increased realization complexity. Employing independent mode/wavelength switching

requires uncoupled SDM fibre.

B. Spatial mode switching across all wavelength channels (space granularity): The WDM-SDM fibre

capacity is switched at the spatial mode level, independent of wavelengths. Hence the entire

communication band per mode is jointly switched. At low spatial mode counts, space switching

granularity is coarse (all WDM channels) but simple to realize. Employing independent mode

switching also requires uncoupled SDM fibre.

C. Wavelength switching across all spatial modes (wavelength granularity): The WDM-SDM fibre

capacity is switched at the wavelength level across all spatial modes, forming spatial superchannels

that are routed through the network as one entity. Since the spatial modes are not separately routed,

the network topology is similar to today’s SMF networks, yet benefitting from the SDM capacity

multiplier. Employing wavelength switching across all modes is obligatory for coupled SDM fibre,

but can also be applied to uncoupled SDM fibres.

D. Wavelength switching across spatial mode subgroups (fractional space-full wavelength granularity):

The WDM-SDM fibre capacity is switched at the wavelength level across smaller spatial mode

subgroups. The switching operation is still in support of spatial superchannels within each subgroup,

but applied independently to the subgroup elements. Employing the fractional space and full

wavelength granularity capacity switching is in support of coupled subgroups SDM fibre, but can also

be applied to uncoupled SDM fibres.

Routing in the WDM-SDM optical network is constrained by the employed switched capacity granularity, as

the network provisioning algorithms must assign each information flow request onto a route that can be

supported by the switching nodes and is contention-free. An additional degree of freedom implicit in the

switching capacity granularity involving the wavelength space (options A, C, and D), is the ability to flexibly

define the switched spectrum.

The independent space-wavelength granularity (option A) is the smallest capacity block size and offers the

greatest routing flexibility as even single wavelength and spatial mode requests can be accommodated. The

alternative solutions (options B-D) utilize larger switching capacity granularities, by addressing all wavelengths

(B), spatial modes (C), or spatial mode subgroups (D) as one entity. Such switching solutions may become

inefficient when addressing small capacity requests, but the reduced hardware required to realize these

degenerate switching solutions may be favourable implementation-wise. If using coupled SDM fibre, then

jointly switching all spatial modes (option C) is mandatory. The capacity granularity can be reduced by

provisioning narrower spectral bands by the switching hardware, which may require some customization. For

example, if the switching hardware supports minimal bandwidth provisioning of 35GHz, then a six-wide

SDM fibre can offer enough capacity to support one terabit per second as the minimal switched capacity,

which is a reasonable starting point for future WDM-SDM optical networks.



Figure 8 - Implementations for independent spatial mode/wavelength channel switching. (A) Optical cross-connect for full connectivity (only two fibre links shown), and (B) Route-and-select per each spatial mode.



After having defined the switched capacity granularity at the WDM-SDM optical network nodes, we turn our

attention to their implementation details. The switching node must complete two functions: routing traffic

from an input SDM fibre to an output SDM fibre and performing channel add/drop to be terminated at

optical transceivers. Each solution entails its unique switching hardware, and various levels of complexity are

associated with each granularity level. However, some elements are recurring and we briefly explain their

operation.

The WSS has been introduced earlier, and its flexible-grid implementation is assumed here. The WSS utilizes

SMF at its input/output ports, and must be properly interfaced to the SDM fibre solutions. For MCF, a

breakout device separates the M cores to M individual SMF. For FMF, a mode demultiplexer converts the M

modes to M individual SMF. This operation does not necessarily require the modes to be mapped to

individual output fibres; a unitary mode-mixing operation may be associated with the demultiplexer that can

be subsequently undone in MIMO processing.

Node design for space-wavelength granularity

Two possible implementations of independent spatial mode/wavelength channel switching (space-wavelength

granularity) offering different levels of flexibility are shown in Figure 8. The first implementation makes use

of a large optical cross-connect. To interface between the uncoupled SDM fibre solution and the OXC, each

independent spatial mode is processed with a 1×K WSS. The WSS subdivides the WDM channels on each

spatial mode according to destination, whether to an output SDM fibre (on a particular spatial mode) or to a

drop port. On the output side a K×1 WSS multiplexes the channels onto a spatial mode of the SDM fibre.

Single drop channels can be terminated directly at conventional receivers, and multiple drop channels can be

separated further with another WSS. This architecture provides full routing flexibility thanks to the OXC,

especially wavelength contention by enabling SDM ‘lane changes’, or routing of a wavelength channel from

one spatial channel to another instead of the much harder wavelength conversion solutions proposed for

SMF networks. Additionally, the transceiver elements are accessible to all fibre directions (and are further

colourless and contentionless), resolving the directional limitation in route-and-select solutions. Having WSS

pre- and post-process WDM channels enables flexible bandwidth allocation, as well as conserve OXC fibre

ports as WDM channels destined to the same direction can be routed jointly.

The second independent spatial mode/wavelength switching solution eliminates the OXC, which is a costly

element and single point of failure threat. Here, route-and-select is performed for each spatial mode

independently. Each independent spatial mode is subdivided by a 1×K WSS and routed to output fibre

destinations (mapped to the same mode on the output fibre, eliminating SDM lane change operation), or to a

set of receivers associated with the spatial mode (i.e., directional in both fibre and mode sense). Hence,

eliminating the OXC results in routing constraints.

The biggest disadvantage of the independent space-wavelength granular switching node designs is the amount

of hardware required to implement them. Essentially, this entails scaling the WSS count M-fold, same as the

capacity increase. Hence the mesh node cost scales linearly with the capacity gain, which is counter to the

value proposition of SDM. We’re seeking a sub-linear cost increase with capacity to maintain network

economics, by way of better device and sub-system integration. With independent space-wavelength granular

switching this requirement is not met.

Node design for spatial mode switching across all spectral channels

Eliminating the WDM switching elements and realizing a space-granular capacity switching design

significantly reduces the node hardware and cost (see Figure 9). An OXC receives all spatial modes along its

input ports and switches each mode to an output SDM fibre, thereby completing the routing of the entire

optical communication band (all WDM channels) to the output destinations. For the add/drop operation, the

OXC can switch the dropped spatial mode to a port where a WSS separates the channels to be detected and

the remaining channels are fed through to a second WSS which combines these channels with additional add

channels back into the OXC for output fibre assignment. The WSS count in the node depends on the

number of spatial modes allowed to be dropped, with a minimum of one spatial mode per fibre. But as this

number increases, which is required to offer reasonable routing flexibility, the WSS count can again become

prohibitively high.



Figure 9 - Switching node designs for implementing space granularity, routing the entire communication band per spatial mode,

Node design for spectral mode switching across all spatial channels

Alternatively, the wavelength-granular capacity switching design (see Figure 10) utilizes a recently-introduced

WSS modification, specifically designed for routing of spatial super-channels. The WSS is based on a

conventional (SMF based), high-port-count WSS, and is made to operate with all the spatial modes of the

input SDM fibre feeding a first subset of the WSS ports. The internal wavelength switching mechanism of the

WSS (based on beam steering), steers the set of input ports onto a second subset of the WSS ports. When the

ports are arranged in a linear, equi-spaced array, then the fibre ports are imaged from the first subset onto the

second, switching in parallel all the fibres and hence the entire spatial mode set. This joint-switching WSS can

be used to construct the conventional route-and-select architecture of the SMF networks, with the M-fold

parallelism applied across all modes with a single switching module. The routed spatial super-channels

traverse a first M×(1×K) for destination selection and a second M×(K×1) for combining the wavelength

channels to the output SDM fibre. The dropped spatial super-channels are interfaced to SDM transceiver

elements where MIMO processing is performed for information extraction in the case where the modes are

mixed due to mode coupling from the SDM fibre. The scalability of this wavelength-granular solution to high

mode counts (tens of modes) is presently undetermined, as joint switching WSS have limited fibre port

counts. One of the main targets for the INSPACE project is to examine alternative designs and technologies

for the development of joint switching WSS with support for high port counts.

Figure 10 - Switching node designs for implementing wavelength granularity, routing all modes per wavelength.



Node design for Wavelength switching across spatial mode subgroups

The final variant is a space and wavelength hybrid design, which offers fractional space and full wavelength

granular capacity switching (see Figure 11). This switching scenario is matched to an SDM fiber solution

offering M spatial modes, where the modes can be divided to M/P independent subgroups, where modes are

coupled within the individual subgroups having P modes. Each mode group must be switched jointly due to

the inherent coupling, yet the groups can be switched independently of one another. The switching solution is

the basic route-and-select topology applied to each subgroup using modified WSS that support the joint-

switching concept. The solution is replicated M/P times, matching the subgroup count. This hybrid solution

offers finer granularity than switching all modes (wavelength granularity), at a price of increased switching

hardware, yet the price is a fraction (1/P) of the capacity gain. While the fractional space-full wavelength

solution requires specific forms of SDM fibre, it can also be applied to uncoupled SDM fibre. This switching

solution can more effectively address SDM fibres with very high spatial mode counts.

Figure 11 - Switching node design for implementing hybrid fractional space-full wavelength switching granularity, routing spatial super-channels spanning spatial sub groups.

Comparison of WDM-SDM switching alternatives

SDM transmission is a promising solution to the capacity limitation of SMF, but addressing the physical SDM

elements of novel fibre types, supporting optical amplifiers, and mode multiplexers, without careful attention

to optical networking implications misses an important element of the entire value proposition. In this paper

we highlighted some of the implications of designing a WDM-SDM optical mesh network, concentrating on

the switching node designs by which information flows need to be provisioned. We identified four categories

of capacity granularities to be provisioned, applied across the space and wavelength domains. Each category

can be realized with different optical switching gear at the network node, affecting the realization complexity

and cost, flexibility, and scalability. These findings are summarized in Table 1.

It is premature at this early stage to deduce if there is an optimal solution to the WDM-SDM optical network.

Different networking applications may likely have divergent conclusions. Assessment of the complete optical

network must to take into account the physical layer attributes, the expected information flow scales and how

efficiently they can be met given the minimum capacity granularity that is routed by the network, blocking

probabilities due to contention for the provisioning of information flows, and cost of implementation,

amongst other. As such, a complete analysis involves the contributions from different skillsets and it will

likely require the concerted effort of many researchers in the field to analyze the performance level and

benefits offered by WDM-SDM optical networks.

The first step for this analysis is to define the different flexible and multi-dimensional optical networking

options that expand the allocation of traffic over both spectral and spatial dimensions. Such options are

related with the different fibre types and switch designs. These options are presented in the following section.



Table 1 - Comparison of WDM-SDM switching alternatives

Space-wavelength

granularity

Switching design 1

Space granularity

Switching design 2

Wavelength

granularity

Switching design 3

Fractional space-full

wavelength granularity

Switching design 4

Realization With OXC: High-port count OXC

and at least 2M conventional

WSS per I/O fibre link.

Without OXC: 2M conventional

WSS per I/O fibre link. 4M if

WSS placed on add/drop.

Moderate port count OXC,

and 2 WSS per mode

selected for WDM channel

add/drop.

4 joint switching WSS per

I/O fibre link in route-and-

select topology applied to

all spatial modes in parallel.

4×M/P joint switching WSS

modules per I/O fibre link.

Flexibility With OXC: Each mode/WDM

channel independent provisioned

and routed. Supports SDM lane

change. Single point of failure.

Without OXC: Each mode/WDM

channel independent provisioned

and routed. Spatial mode

maintained. Prone to wavelength

contention.

The complete optical

communication band is

routed across network.

Coarse granularity. If

WDM channels need to be

extracted from many

modes then WSS count

quickly escalates.

Each spatial super-channel

provisioned across all

modes. Susceptible to

wavelength contention.

Add/drop bound to

direction.

Compromise solution using

small SDM groups. More

efficient when provisioning

low capacity demands.

Scaling With OXC: Can quickly escalate

to very large port counts.

Switching node cost linearly

scales with capacity, no price

benefit to SDM.

Conventional OXC can

support foreseeable mode

and fibre counts. OXC is

single point of failure.

Pricing favourable but with

greater add/drop require

more WSS modules.

Cost roughly independent of

SDM count. Inefficient for

low capacity connections

due to minimum BW

provisioned across SDM.

Large SDM Rx/Tx are

integration and DSP

challenge.

Cost scales as group count.

Groups can be turned on as

capacity grows, offering pay-

as-you-go alternative.

Maintaining small group sizes

facilitates MIMO processing

at Rx.



4. Network Control Plane (WP5 related)

The main function of the Control Plane (CP) of an optical network is to provide an interface through which

to perform connection setup, teardown and network monitoring. In order to do so, the CP must be able to

interface with the various devices that make up the optical network and to configure them in a manner

consistent with the provided high level goals (such as instantiating a connection). Additionally, modern

control planes include functionality to compute the most appropriate resource assignment and further

(possibly optional) functionalities, as it is described in the following sub-section.

4.1. Control Plane Functional Requirements

The INSPACE control plane will have to fulfil the following requirements; please note that their definition is

part of D5.1, and the following is only a summary:

[REQ. 1] Network Programmability: the CP will expose an Application Programming Interface (API)

through which to control the network. This is similar to what is already possible with GMPLS networks

(more details about GMPLS are provided below and in deliverable D5.1), but permits simpler extension

and integration with third-party management software.

[REQ. 2] SDM service handling: the INSPACE control plane shall be able to setup, tear down, modify

and refuse to instantiate SDM connections. This requirement can be further subdivided into:

[REQ. 2.1] Service setup: the CP shall be able to setup SDM services on the basis of knowing the ingress

and egress nodes, and at least the required capacity (bandwidth).

[REQ. 2.2] Service teardown: the CP shall enable the removal of an established SDM service, thus freeing

the resources it occupied for future use.

[REQ. 2.3] Service modification: the CP shall be able to move, enlarge or shrink the set of resources

associated with a SDM service, either directly or via a Make-before-Break approach.

[REQ. 2.4] Admission Control: the CP interface to connection requestors shall support the possibility of

denying a request, due to the finite amount of optical resources available.

[REQ. 2.5] Resource and Capability discovery: the CP must be able to gather knowledge about the

characteristics and capabilities of all relevant objects and devices in the network at every node that

performs resource assignment.

[REQ. 3] Extended SDM service features: the CP shall implement a number of “advanced” features that

are present in non SDM control planes, such as survivability, status querying, unsolicited status updates

and alarms, re-optimization, network monitoring and virtualization. More in detail:

[REQ. 3.1] Status Notification: the CP shall be able to asynchronously produce notifications of events

involving active lightpaths (such as unexpected service terminations due to failures) and relay them to the

controlling logic.

[REQ. 3.2] Network re-optimization: the CP shall be able to re-arrange the set of active connections to

minimize a target function, such as spectral occupation or a fragmentation metric.

[REQ. 3.3] Survivability: the CP shall enable the provisioning of protected services, and/or automatically

attempt to restore connectivity of failed services while minimizing traffic losses.

[REQ. 3.4] Network monitoring: the CP shall collect and collate statistics of the network’s behavior. Within

the INSPACE project, this aspect will focus on delivering alarm conditions to the controlling logic, rather

than cover the wide range of information that can be collected from counters and physical sensors.

[REQ. 3.5] Virtualization: two different flavors of virtualization can be expected from the control plane:

[REQ. 3.5.1] User-requested virtual topologies, where the user defines the virtual nodes and links

he is interested in, which have to then be mapped (designed and activated) on the physical infrastructure

by the network operator (i.e., multiple concurrent service requests are supported).

[REQ. 3.5.2] Infrastructure as a Service (IaaS), where users can define virtual infrastructure on

top of the physical one. The CP shall reserve resources for virtual topologies and map connections over

them to connections in the physical topology. Within the INSPACE project, the implementation of this

virtualization flavor is not foreseen.



4.2. Control Plane architecture

From the architectural point of view, the INSPACE control plane needs two main interfaces to the outside

world: a north-bound interface, which connect the control plane (CP) with its client layers (for example, a

Network Management System from which connection requests originate), and a south-bound interface,

which connect the CP with the data plane and, in distributed settings, is responsible for the synchronization

between all decision points. Inside the CP, we can generally distinguish between some form of traffic

engineering database (TED), and the actual controller implementing the control logic. The relations

among the above mentioned components are depicted in Figure 12.

More in detail, the north-bound interface, is the client-facing interface, and may take the form of a protocol

or an API (likely a simple REST API for the purposes of this project). It must export functionalities from

requirements, including service setup, teardown and modification, admission control, solicited and unsolicited

information retrieval and more.

The task of the controller is to receive requests from the clients via the north-bound interface and implement

the necessary configuration to serve them (after the needed computations) via the south-bound interface.

Both the controller and the TED may be centralized in a single node or node cluster, or distributed among all

nodes. In the former case, each node maintains at least a local knowledge base and runs an agent handling

communications with the control node(s).

Finally, the south-bound interface is responsible for both database synchronization (device, topology,

resource and capabilities discovery) and service signalling. In a centralized setting, this translates to both

solicited and unsolicited configuration/state/alarm messages from the nodes to the centralized controller, as

well as configuration messages pushed by the latter to the formers to instantiate services. Conversely, in a

distributed setting, the synchronization function would be responsible for flooding all the necessary data to all

participating nodes, and the signalling function would travel from node to node along each service’s intended

path.

Figure 12 - Main architectural elements of the INSPACE Control Plane.

Architectural Archetypes The architectural archetypes to be considered for INSPACE are detailed in the following.

North-bound Interface

Controller TED

South-bound Interface

South-bound Interface South-bound Interface South-bound Interface

Client…

Client Client

Optical Node Optical Node Optical Node



Fully Distributed GMPLS: the Generalized Multi-Protocol Label Switching suite is a staple of modern

optical transport networks. The fully distributed nature of GMPLS means that path computation happens on

the optical nodes, which may or may not have much spare processing capacity to run complex routing

heuristics, or even lack necessary information, which may not flooded via OSPF-TE (Open Shortest Path

First with Traffic Engineering extensions) to keep the protocol lightweight. Another issue is that in GMPLS

device configuration is carried out by the RSVP-TE protocol (Resource Reservation Protocol with Traffic

Engineering extensions) starting from the head node of a path, which makes the programmability approach

of allowing almost direct access to each device in the network much more convoluted than it needs to be. In

fact, to modify a set of N connections in just a middle node, N RSVP-TE requests would have to be

generated, each at the source node of the relative service, carrying the same configuration of the original

services but for the desired changes in the node being modified. Furthermore, the intrinsically distributed

nature of path computations for services originating at different nodes means that near-concurrent requests

risk reserving resources desired by both, thus triggering the failure of one or both service setups and the

necessity of a re-trial over another, likely longer path and set of spectral resources. This lack of coordination

also renders the implementation of some form of coordinated re-optimization much more difficult than if

operations were serializable.

Hybrid PCE/GMPLS: a Path Computation Element (PCE) is a logically centralized element that is

responsible for path computations in a network. By centralizing the path computation function (and the

associated knowledge) on a dedicated node with spare processing capacity, it may allow the employment of

more sophisticated Routing and Spectrum Assignment algorithms than are possible with simple GMPLS.

Stateful PCEs, which keep state between requests, can be used to overcome the concurrency limitation of

pure GMPLS; active stateful variants with instantiation capabilities are also able to autonomously setup or

release connections independently from clients, thus behaving as a full network controller. In such an

architecture the PCE communication Protocol (PCEP) acts as both the north-bound interface towards clients

(unless hidden behind a User to Network Interface, with the nodes themselves acting as PCEP clients, as is

usually the case) and south-bound interface (in conjunction with the GMPLS protocols). In this sense, it can

be considered a first, albeit very limited, form of Software Defined Networking.

Fully Centralized SDN: at a high level, Software Defined Networking (SDN) is concerned with the

decoupling of the control and data planes, which, while a relatively new concept in the context of packed

switched LANs where this paradigm first emerged, is already the norm in optical transport networks. An

ancillary characteristic of SDN solutions is their inherent programmability, with a logically centralized

controller element responsible for the configuration of a domain, and able to offer in-depth access to device

state and configuration through an easy-to-use software API. Therefore, devices no longer need to run smart

logic on-board, and complex control logic can be outsourced to external, user-developed “network”

applications.

The full SDN solution, possibly developed on top of existing open-source projects, would appear to allow a

simpler implementation of the SDM network model, requiring at worst an amount of work comparable to

that of a GMPLS or PCE/GMPLS solution. In addition, it would allow to avoid vendor lock-in (being open-

source), as well as being freely modifiable by third parties to, for example, support additional devices or

functionalities. Furthermore, given the explicit project goal of providing network programmability, a full

open-source SDN control plane is the most adherent to the stated objectives. For these reasons, a full SDN

architecture is proposed in WP5, rather than a simple PCE/GMPLS hybrid with added north-bound

capabilities. Please note that this does not preclude the use of a GMPLS stack as one of the south-bound

protocols, nor the use of a PCE-like component as part of the logically centralized controller.

More details about the INSPACE control plane framework are available in D5.1. More specifically, D5.1 also

contains, at a relatively high level of abstraction, the INSPACE topology and connection models, as well as

the selected overall architecture of the INSPACE control plane, its internal components and how they

interact, and some protocol-related choices for the interfaces towards the clients/client applications (i.e., the

north-bound interfaces) and the network devices (i.e., the south-bound interface).



5. Node implementation requirements

This section provides the specifications that will be followed by the INSPACE project for the node

implementation. These specifications define the footprint, thermal and physical stability and heat anticipation

requirements. They will be used for the definition of the system footprint and the power consumption, in the

relevant studies performed in task 2.3.

5.1. Expected node foot-print and housing requirements

Advanced Telecommunications Computing Architecture (ATCA or Advanced-TCA) is the largest

specification effort in the history of the more than 100 companies participating. Advanced-TCA is targeted to

requirements for the next generation communications equipment. This series of specifications incorporates

the latest trends in high speed interconnect technologies, next-generation processors, and improved

Reliability, Availability and Serviceability (RAS). We will follow ATCA specifications for INSPACE principles

based cross connect mechanical specification.

Figure 13 - Example of a shelf with multiple slots according to ATCA specifications

An ATCA board is 280 mm deep and 322 mm high. The boards have a metal front panel and a metal cover

on the bottom of the printed circuit board to limit electromagnetic interference. The locking injector-ejector

handle (lever) actuates a micro switch to let the Intelligent Platform Management Controller (IPMC) know

that an operator wants to remove a board, or that the board has just been installed, thus activating the hot-

swap procedure. ATCA boards support the use of PCI Mezzanine Card (PMC) or Advanced Mezzanine Card

(AMC) expansion mezzanines.

The shelf supports RTMs (Rear Transition Modules). RTMs plug into the back of the shelf in slot locations

that match the front boards. The RTM and the front board are interconnected through a Zone-3 connector.

The Zone-3 connector is not defined by the ATCA specification.

Each shelf slot is 30.48 mm wide. This allows for 14-board chassis to be installed in a 19-inch rack-mountable

system and 16 boards in an ETSI rack-mountable system. A typical 14-slot system is 12U or 13U rack units

high. The large ATCA shelves are targeted to the telecommunication market so the airflow goes in the front

of the shelf, across the boards from bottom to top, and out the rear of the shelf. Smaller shelves that are used

in enterprise applications typically have horizontal air flow. We will try to adopt INSPACE solution to the

ATCA medium shelf design supported by ATCA defined backplane.

5.2. Thermal stability and heat anticipation requirements

This section provides criteria for temperature and humidity robustness of network equipment. The criteria

cover the following:



Operation temperature and humidity environments

Temperature margin

Fan-cooled equipment

Heat dissipation and energy efficiency

Surface temperature

Equipment airflow

These are discussed in detail in the following paragraphs:

Operation temperature and humidity environments

The table below provides the normal operating temperature/humidity levels and short-term operating

temperature/humidity levels in which network equipment shall operate. The equipment shall not sustain any

damage or deterioration of functional performance during its operating life when operated within the

conditions described in next table. Note that equipment aisle refers to conditions at a location of 1524 mm

(60 in) above the floor and 381 mm (15.0 in) in front of the rack.

Table 2 - Equipment Aisle Air Temperature and Humidity Limits

2Short-term refers to operating conditions that include the temperature margin

Temperature margin

The temperature margin evaluation is intended to determine the system response to temperatures above the

short-term extreme. This is not intended to change design criteria or operating temperature range. It is

intended only to provide additional information. Equipment response to temperatures up to 100C above the

short-term high temperature extreme of above table shall be determined.

Fan-cooled equipment

The criteria in this section are intended to ensure that fan-cooled equipment operate as intended and can be

maintained efficiently. It is important that equipment will continue to function normally with a single fan or

blower failure over the entire long term operating temperature range. Equipment cooled by forced

convection shall not sustain damage or deterioration of functional performance when operated with any

single fan failure at a 40oC equipment-aisle air temperature for a short-term of up to 96 hours. Equipment-

Aisle Air Temperature and Humidity Limits are depicted on next figure.



Figure 14 - Equipment-Aisle Air Temperature and Humidity Limits

Heat dissipation and energy efficiency

Management of energy consumed and heat dissipated by telecommunications equipment is a major challenge

for service providers. Crucial to this management is accurate reporting of expected equipment power and

heat loads. The heat dissipation criteria of this section are based on the cooling capacities of traditional

network facilities. For additional information on equipment and room cooling methods, refer to GR-3028-

CORE, Thermal Management in Telecommunication Central Offices: Thermal GR-3028.

Equipment rate of heat dissipation should not exceed the values shown in the following table.

Table 3 - Equipment rate of heat dissipation

Surface temperature

The criteria and methodology of this document are based on ATIS-0600004.2006, Equipment Surface

Temperature. It is a requirement that equipment surfaces that face aisles or surfaces where normal

maintenance functions are anticipated shall conform to the temperature limits established in Table 2, when

the when the equipment is operating in a room with an ambient air temperature of 230C.



Table 4 - Temperature Limits of Touchable Surfaces

Equipment airflow

The most common way to cool modern telecommunications equipment is with air convection often using

internal equipment fans to create forced-convection cooling. The location of air inlets and exhausts in

equipment can affect the efficiency of heat removal in both equipment and the room. For this reason, a

standard airflow pattern is desirable. Service providers usually rely on equipment environments with

ventilation cooling air provided from overhead ducts and diffusers. An estimated 95% of the equipment

buildings operated by service providers utilize this room cooling scheme. In addition, most environments

maintain cooler front (maintenance) aisles and warmer rear (wiring) aisles.

Thermal stability and heat anticipation requirements for the INSPACE node

GR-3028-CORE classifies the location of equipment air inlet and exhaust on network equipment, and

provides the basis for the definition of the following and general specifications for the INSPACE node:

Table 5 - Absolute Maximum ratings

(exceeding these ranges may result in permanent module damage)

Description Unit Min Max Notes Storage Case Temperature

0C -40 85

Humidity % 5 95 Non-condensing

Supply Voltage V 48 Class N per HBM, meets IEC 61000-4-2 level 4.

ESD kV 8

Shipping Temperature 0C -40 85

Table 6 - Normal operating conditions

Description Unit Min Max Notes Case Temperature 0C -5 70

Humidity % 90 Non-condensing

Supply Voltage V 48 Class N per HBM, meets IEC 61000-4-2 level 4.

Cooling Air Flow Rate m/s 3 5 Air flow @ 25C

0 3 Air flow @ -5C Ambient temperatures Mfg. (Final Test)

0C -5 65 Forced air flow

Power Consumption W TBD

5.3. Physical stability and vibration issues

This section provides the generic criteria for earthquake, office vibration, and transportation vibration for

network equipment.

Earthquake Environment

During an earthquake, telecommunications equipment is subjected to motions that can over stress equipment

framework, circuit boards, and connectors. The amount of motion and resulting stress depends on the

structural characteristics of the building and framework in which the equipment is contained, and the severity



of the earthquake. Next table shows the map of earthquake risk zones. Zone 4 corresponds to the highest

risk areas, Zone 3 the next highest, and so on.

Table 7 - Map of earthquake risk zones

Office Vibration Environment and Criteria

Telecommunications equipment may be subjected to low-level vibration in service that is typically caused by

nearby rotating equipment, outside rail or truck traffic, or construction work in adjacent buildings or spaces.

This vibration can cause circuit board “walkout,” malfunctions, or other service interruptions or failures.

Description Unit Min Max Notes

Mechanical Shock 0-100 Hz sine & 11ms Half sine 5 shocks each axis

dB/octave -12 Frequency range 5–10Hz

12 Frequency range 5–10Hz

< 10G max



6. Conclusions on the targeted INSPACE system specifications and

requirements

This section lists the key system specifications and requirements for INSPACE. The targeted requirements

are identified for all cases and where applicable extended requirements are identified. These requirements are

the basis for the development efforts in WP3, WP4 and WP5 as well as the networking studies that follow in

WP2. It is noted that the testing requirements for WP6 activities will be defined in D6.1 according to the final

outcomes of the development work and the available resources for the test-bed. The listed requirements here

refer to the system evaluation and definition efforts and act as a guideline for the node, transmission and

control modules under development.

6.1. Spectrum related requirement

Operating spectrum

The main target for the operating spectrum is C-Band. This affects the targeted bandwidth of the node and

the transmission elements.

Studies will also examine the possibility for extended operation over the C- and L-Band

Channel allocation in spectrum

For maximum flexibility a flex-grid spectrum allocation approach is specified. The switching elements of the

INSPACE have to support the flex-grid operation, i.e. support the switching of superchannels with variable

bandwidth allocation options.

No maximum spectral width is defined. However, for practical reasons, related with the generation and

allocation of super-channels in the spectrum dimension, typical maximum spectral width values of up to

250GHz will be considered.

According to the channel allocation options defined in section 1.2 some benchmarking cases will also

consider fix-grid spectrum allocation.

Frequency slot (minimum grid spectral width)

The main specification for the minimum spectral width is 12.5GHz. This number is defined in compliance

with the expected standards.

However, from the technology point of view, a 6.25GHz slot is also feasible. The added benefit from the use

of a narrower frequency slot in combination with the added design complexity will be examined in the related

networking studies.

6.2. Space related requirements

Types of fibres

The targeted specifications include the following types:

- MMF supporting 3 modes (LP01 and the generated set of LP11)

- MCF with isolated cores supporting 7 cores

These types will be the main ones to be explored according to fibre availability. It is noted that for the case of

MCF a 4-core fibre is possible to be used instead of a 7-core fibre.

For the extended networking studies the additional types of fibres to be considered are:

- MMF supporting 5 modes (LP01 and the generated sets of L11, LP21)

- MCF with isolated cores supporting 12 cores

- MM-MCF with 6 cores and 3 modes per core

The case of an SMF bundle will be considered as benchmarking case in all studies.



Channel allocation in space (according to fibre type)

For the case of MMF the space dimension is allocated partially or fully to one channel and is not reused, by

another channel in the same spectral slots.

For the case of MCF with isolated cores the space dimension can be flexibly allocated to any number of

channels up to a number equal to the number of cores. The full flexibility in space dimension will be

examined for the independent spatial mode/wavelength channel switching node implementation design

(switching design 1). In order to reduce the node complexity, the case of spatial mode group switching will

also be examined (switching design 4). Comparisons with respect to switching designs 2 and 3 will be

performed for this type of fibre.

The case of MM-MCF is of particular interest for the subgroup switching node design solution and will be

explored in both the networking studies and with respect to the node design complexity.

6.3. Super-channel generation requirements

Minimum super-channel capacity

At minimum a channel can be defined over one frequency slot (12.5GHz) in one mode/core in space using

the simplest modulation format of BPSK. It results in a minimum switching granularity entity of around

12Gb/s (considering spectral guard bands) achieved in the case of independent spectral-spatial switching.

However, a small super-channel capacity reduces significantly the spectral efficiency of the system due to the

additional spectral guard bands required to be used to allow the super-channel processing at the node.

The minimum super-channel capacity is defined for 100Gb/s demands. This number is chosen as a common

and practical reference value for the networking studies and according to the next generation core network

demands and traffic models. It is not limited by the switching or processing capabilities.

Maximum super-channel capacity

It is not defined. It can take extremely high numbers. Practically it depends on the transmitter capabilities.

Data formats and symbol rate

Studies will mainly target coherent BPSK and QPSK schemes with polarization multiplexing. The studies can

be extended to cover also the case of 16QAM.

The two schemes to be explored are the typical single carrier coherent WDM scheme and the electronic

OFDM scheme.

The symbol rate is determined by the format and the frequency slot in combination with the required spectral

guard band. For the defined frequency slot of 12.5GHz the symbol rate can range between 10Gbaud and

12Gbaud. The exact value will be determined for the final testing activities in WP6 at the end of year 2 of the

project

Spectral guard bands

This parameter depends of the resolution and addressability of the WSS element and it will be determined in

detail in WP4 studies.

The main target is to use spectral guard bands of 1 frequency slot considering typical WSS solutions available

today. (For small addressability values it can be defined in terms of WSS slots). The use of technologies

achieving finer resolution and narrower spectral guard bands will be explored in WP4 for a targeted spectral

guard band of 6.25GHz.

Spatial guard bands (optional)

The idea of using spatial guard bands is proposed for the case of spatial group switching (switching design 4)

when coupled MCF are used. According to this proposal certain cores can be left unassigned in order to

prevent strong cross talk with the neighbouring cores allowing the independent allocation of channels on the

rest of the cores (or group of cores).

This option is out of the initial objectives of INSPACE and its examination is optional.



Super-channel allocation options

Refer to subsection 1.2 for the different options.

Main target is to fully evaluate (theoretically and experimentally) options C and E and examine theoretically

option D. Options A and B are considered as benchmarking cases and will be examined theoretically

Optical grooming

The use of optical grooming is optional, since it is not defined in the initial objectives of INSPACE. However

it is identified that optical grooming may lead to savings in spectral efficiency at the cost though of added

control complexity.

The main target is to avoid optical grooming and consider the typical allocation of super-channels on an end-

to-end basis

Optical grooming can be explored optionally for the space dimension only (i.e. spatial and no spectral

grooming, by combining super-channels that are allocated over the same spectral slot into different core

groups that are jointly switched in core nodes. (Relevant networking studies are required first to be conducted

to examine any possible benefits).

6.4. Switching requirements

Types of spatial-spectral flexible switches

As defined in subsection 3.2 there are 4 designs proposed by INSPACE project:

- Independent spatial and spectral switching (switch design 1)

- Spatial mode switching across all wavelength channels (switch design 2)

- Spectral mode switching across all space channels (switch design 3)

- Independent spatial and spectral switching of spatially defined groups (switch design 4)

All these designed will be examined and compared according to the channel allocation options

Implementation

INSPACE targets the implementation of switch designs 3 and 4 for MMF and MCF types respectively.

Design 1 will be studied theoretically due to its increased design complexity and cost. The same is also true

for design 2 which can be implemented with existing switching solutions and will be examined as a

benchmarking case to compare all switching options.

Switching element specifications

The switching element must support 12.5GHz grid with resolution of 10GHz or less. It must be also able to

scale to all cores or modes or combination of them and operate over at least the C-band.

The exact design specifications will be provided by WP4 (complying though to the system requirements

defined in this deliverable)

.

Documents

INSPACE network and system definition and specification€¦ · The control plane functional requirements and the proposed architecture are summarized. A detailed description of the