spirit d2.1 report on application test cases, system ... ·...

45
SPIRIT – D2.1: Report on application test cases, system requirements and design for 400G/Terabit transmission Page 1 SEVENTH FRAMEWORK PROGRAMME THEME [ICT2013.3.2] [Photonics] SPIRIT Softwaredefined energyefficient Photonic transceivers IntRoducing Intelligence and dynamicity in Terabit superchannels for flexible optical networks Grant Agreement no. 619603 D2.1 Report on application test cases, system requirements and design for 400G/Terabit transmission Lead beneficiary for this deliverable: OTE Contact Person: Ioanna Papafili, George Agapiou Address: 2, Pelika & Spartis str., Maroussi, Greece Phone: +30 210 611 4706, 4663 Fax: +30 210 611 4650 email: {iopapafi, gagapiou}@oteresearch.gr Date due of deliverable: Actual submission date: Deliverable Authors: Participants: Workpackage: Security: Nature: Version: Total number of pages: M4 31/03/2014 George Agapiou, Dimitris Apostolopoulos, Stefanos Dris, Marco Camera, Roberto Magri, Santo Nani, Ioanna Papafili, Stamatis Perdikouris, Maria Spiropoulou, Hercules Avramopoulos OTE, TEI, ICCS/NTUA WP2 PU R 1.0 45

Upload: ngoxuyen

Post on 20-Aug-2018

214 views

Category:

Documents


0 download

TRANSCRIPT

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  1  

SEVENTH  FRAMEWORK  PROGRAMME  

THEME  [ICT-­‐2013.3.2]  [Photonics]  

 

SPIRIT  

Software-­‐defined  energy-­‐efficient  Photonic  transceivers  IntRoducing  Intelligence  and  dynamicity  in  Terabit  

superchannels  for  flexible  optical  networks  Grant  Agreement  no.  619603  

D2.1  Report  on  application  test  cases,  system  requirements  and  

design  for  400G/Terabit  transmission    

Lead  beneficiary  for  this  deliverable:   OTE  Contact  Person:   Ioanna  Papafili,  George  Agapiou  

Address:   2,  Pelika  &  Spartis  str.,  Maroussi,  Greece  Phone:   +30  210  611  4706,  4663  

Fax:   +30  210  611  4650  e-­‐mail:   {iopapafi,  gagapiou}@oteresearch.gr  

 Date  due  of  deliverable:  Actual  submission  date:    Deliverable  Authors:          Participants:    Workpackage:    Security:    Nature:    Version:    Total  number  of  pages:    

M4  31/03/2014  George   Agapiou,   Dimitris   Apostolopoulos,   Stefanos  Dris,   Marco   Camera,   Roberto   Magri,   Santo   Nani,  Ioanna   Papafili,   Stamatis   Perdikouris,   Maria  Spiropoulou,  Hercules  Avramopoulos  OTE,  TEI,  ICCS/NTUA  WP2  PU  R  1.0  45  

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  2  

 

 

 

 

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  3  

Executive  Summary  This   document   is   the   deliverable   “D2.1:   Report   on   application   test-­‐cases,   system  requirements   and   design   for   400G/Terabit   transmission”   of   Work   Package   2  “Application   scenarios   &   system   design,   component   specifications   and   software  development”  within   the   ICT   SPIRIT  Project  619603.   The  aim  of  Deliverable  D2.1   is   to  derive   the   system-­‐level   specifications   of   the   SPIRIT   flexible   transceiver   based   on  application-­‐driven  network  requirements  following  a  top-­‐to-­‐bottom  approach.  

In   the   first   part   of   this   document,   we   have   assessed   the   broadband   market   and  considered  studies  reporting  on   Internet  measurements  and  forecasting.  Based  on  the  findings   of   these   studies,   we   defined   the   major   application   scenarios   of   interest   to  SPIRIT.   Additionally,   we   have   identified   the   application-­‐driven   network   requirements  mainly  in  terms  of  bitrate  and  latency,  and  described  the  reference  network  architecture  of   OTE,   which   will   be   considered   in   the   evaluation   and   demonstration   of   the  functionality  of  the  SPIRIT  components.  

Furthermore,   the   requirements   from   a   system   perspective   are   presented,   and  appropriate   test   cases   for   performance   evaluation   of   SPIRIT’s   devices   are   identified.  Operation   in   a   fixed-­‐grid   WDM   environment   is   envisaged,   both   in   a   mixed  coherent/legacy   traffic   scenario,   as  well   in   an   all-­‐coherent   channel   setting.  Moreover,  the   requirement   for   programmable   and   flexible-­‐format   operation   will   ensure  compatibility  with   future   elastic   optical   networks,  with   an   emphasis   on   the   emerging  flex-­‐grid  standard.  

Evaluation  of  SPIRIT’s  devices  will  take  place  both  in  the  laboratory,  as  well  as  in  a  field  trial  with  deployed  fiber  and  co-­‐propagating  traffic  from  commercial  transponders.  The  test  cases  identified  aim  to  evaluate  component  performance,   in  addition  to  validating  SPIRIT’s  targeted  system  concepts  in  realistic  network  scenarios.  

 

 

 

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  4  

Table  of  contents  1.   Introduction ........................................................................................................................ 6  

1.1   Purpose  of  the  document................................................................................................ 6  

1.2   Document  outline............................................................................................................ 6  

2.   Application  scenarios  and  requirements............................................................................. 8  

2.1   Broadband  market  analysis ............................................................................................. 8  

2.1.1   Fixed  broadband...................................................................................................... 8  

2.1.2   Mobile  broadband................................................................................................. 11  

2.2   Internet  traffic  analysis  and  forecast............................................................................. 13  

2.2.1   Technology  evolution ............................................................................................ 13  

2.2.2   Global  traffic .......................................................................................................... 14  

2.2.3   Mobile  traffic ......................................................................................................... 15  

2.2.4   Cloud  traffic ........................................................................................................... 16  

2.3   Future  traffic  trends ...................................................................................................... 18  

2.4   Application  scenarios .................................................................................................... 20  

2.4.1   Convergence  of  fixed  and  mobile  networks.......................................................... 20  

2.4.2   Support  of  bandwidth-­‐intensive  end-­‐user  applications ........................................ 21  

2.4.3   Support  of  service  and  network  virtualization ...................................................... 22  

2.5   Application-­‐driven  network  requirements.................................................................... 24  

2.6   Reference  network  architecture ................................................................................... 25  

3.   System  requirements  and  design ...................................................................................... 28  

3.1   System  requirements .................................................................................................... 28  

3.2   System  specification...................................................................................................... 29  

3.2.1   Modulation  Formats.............................................................................................. 29  

3.2.2   Multi-­‐Carrier/Superchannel  Generation ............................................................... 31  

3.2.3   Software-­‐Defined  Operation ................................................................................. 34  

3.3   Test  Cases ...................................................................................................................... 34  

3.3.1   Laboratory  Testing  at  Ericsson .............................................................................. 34  

3.3.2   Laboratory  Testing  at  ICCS/NTUA.......................................................................... 38  

3.3.3   Field  Trial  Testing................................................................................................... 39  

4.   Conclusions ....................................................................................................................... 40  

5.   List  of  Figures .................................................................................................................... 41  

6.   List  of  tables ...................................................................................................................... 42  

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  5  

7.   Abbreviations .................................................................................................................... 43  

8.   References......................................................................................................................... 44  

 

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  6  

1. Introduction  

New  end-­‐user  applications  such  video  streaming,  online  storage,  online  social  networks,  etc.  are  bandwidth-­‐intensive  and  thus,  are  stretching  the  capacity  of  the  physical   layer  to   the   limits.  Moreover,   a   general   trend   is   observed   over   the   last   few   years   towards  software-­‐defined  architectures  and  virtualized  network  functionalities  so  as  to  support  cloud-­‐based  applications  in  terms  of  mobility,  agility  and  flexibility.  

In  order   to  accommodate  bandwidth-­‐hungry  applications,  as  well   as  virtualization  and  software-­‐defined   networking   architectures,   fully   programmable   optical   components  that  will  support  easy  rate  and  format  adaptation  upon  the  requirements  of  the  upper  layers  are  needed.  SPIRIT  [1]  will  fabricate  low-­‐cost,  energy-­‐efficient  flexible  transceivers  that  are  capable  of   flex  grid  and  gridless  operation,  while  being  compatible  with  both  current   and   future   applications.   Additionally,   both   single-­‐   and   multi-­‐carrier   (OFDM)  QAM   formats  will   be   supported.   Interfacing   to   external   digital   control   logic  will   allow  dynamic  adjustment  of  the  symbol  rate  (up  to  32GBaud)  and  modulation  format.  

The  design  and   fabrication  of  SPIRIT’s  devices   is  driven  by   the   requirements   set  by   its  target   applications;   system-­‐level   specifications   need   to   be   set   in   order   to   achieve  compatible   operation   with   current   and   future   optical   networking   applications.   The  system   specifications,   in   turn,   need   to   be   appropriately   translated   to   the   component  level:  Devices  will  be  needed  that  will  meet  the  functionality  and  performance  levels  in  terms  of  bandwidth,  resolution,  flexibility  and  programmability.  

1.1 Purpose  of  the  document  

The  purpose  of  this  document  is  manifold:    

! To   identify   and   describe   the   potential   application   scenarios   of   the   SPIRIT  transceiver   based   on   current   and   forecasted   traffic   and   market   trends,   the  objectives   for   telecommunications   in   Europe   until   2020,   and   upcoming   new  networking  paradigms  such  as  cloud  computing,  network  function  virtualization  and  software-­‐defined  networking.    

! Based  on  the  identified  application  scenarios  of  the  SPIRIT  transceiver,  to  derive  the  application-­‐driven  network  requirements  (e.g.  in  terms  of  rate  and  latency).  

! To  map  the  network  requirements  to  system-­‐level  requirements.  

! To  provide  the  initial  system  specification  of  the  SPIRIT  component  based  on  the  above  system  requirements.  

! To  define  specific  test-­‐cases  for  the  SPIRIT  prototype.  

1.2 Document  outline  

This  document  is  organized  as  follows:    

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  7  

Section   2   provides   an   overview   of   the   factors   that   drive   the   need   for   flexible,   fully-­‐programmable,  optical   components,  describes   the  application   scenarios   for   the  SPIRIT  transceiver,  and  defines  the  application-­‐driven  network  requirements  for  the  respective  scenarios.    

Section   3   outlines   the   system   requirements,   as   well   as   test   cases   to   be   used   in   the  evaluation  of  the  devices  of  SPIRIT.  

Finally,  Section  4  summarizes  this  document  and  presents  the  main  conclusions.  

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  8  

2. Application  scenarios  and  requirements  

In   this   chapter,   the   increasing   capacity   demand   of   end-­‐user   applications,   which   is  expected  to  drive  the  growth  of  aggregate  traffic  at  the  metro/core  network  of  ISPs,  is  discussed.   In  particular,  we  consider  current  and   forecasted   traffic  and  market   trends,  the  objectives  for  telecommunications   in  Europe  for  the  next  six  years,  as  well  as  new  networking   paradigms   that   drive   the   development   of   new   applications,   and   thus,  capacity   demand.   Furthermore,   specific   application   scenarios   for   the   SPIRIT   optical  component   are   discussed.   Finally,   the   bandwidth   and   network   requirements   of   the  SPIRIT   flexible   transceiver   in   the   aforementioned   scenarios   are   analyzed,   taking   into  account  the  prospective  capacity  demand  of  end-­‐user  applications.  

2.1 Broadband  market  analysis  

A   qualitative   and   quantitative   analysis   on   the   use   and   implementation   of   broadband  infrastructures   and   services,   both   for   fixed   and   mobile   networks   is   discussed   in   this  section.   Specifically,  we  deal  with   Internet  demographics   such  as  broadband  coverage  and   penetration,   number   of   fixed   and   mobile   users,   and   future   targets   for   the  telecommunications  market  in  Europe.  

Europe  20201 is  a  10-­‐year  strategy  proposed  by  the  European  Commission  on  March  3rd,  2010,   for   advancement   of   the   economy  of   the   European  Union   (EU).   The   initiative   of  Europe   2020   related   to   the   roll-­‐out   of   high-­‐speed   Internet,   constitutes   the   so-­‐called  Digital   Agenda2  (DA)   for   Europe.   The   DA   identifies   the   need   for  much   faster   internet  access  than  is  generally  available  in  Europe,  especially  for  providing  new  services  such  as  high  definition  television  or  videoconferencing.  To  match  world  leaders  like  South  Korea  and  Japan,  the  DA  identifies  the  need  for  Europe  to  achieve  download  rates  of  30  Mbps  for  all  of  its  citizens,  and  100  Mbps  for  at  least  50%  of  European  households  subscribing  to  internet  connections  by  2020.    

Two   important  metrics   for   the  quantification  of   the  broadband  market  are  broadband  coverage,   and   broadband   penetration.   Broadband   coverage   is   the   percentage   of  population   where   a   broadband   contract   can   be   activated   by   the   telecom   operator  without  the  need  for  new  deployments,  while  broadband  penetration  is  the  percentage  of  population  with  an  active  broadband  contract.  

2.1.1   Fixed  broadband  According  to  [2],  at  the  end  of  2012,  over  99.9%  of  European  homes  could  have  access  to  at  least  a  basic  broadband  network  considering  all  technologies,  i.e.,  including  fixed,  fixed-­‐wireless,  mobile  and  satellite  access.  Satellite  broadband  has  the   largest  physical  

                                                                                                               1  http://ec.europa.eu/europe2020/index_en.htm  2  http://ec.europa.eu/digital-­‐agenda/  

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  9  

coverage:  It   is  available  to  100%  of  the  population  in  24  of  27  Member  States.  Despite  the  high  coverage,  satellite  take-­‐up  is  still  marginal,  as   it  represents   less  than  1%  of  all  EU   broadband   lines.   Without   satellite,   99.4%   of   homes   are   covered   by   broadband  (standard  broadband).  Considering  only  fixed  and  fixed-­‐wireless  technologies  (standard  fixed  broadband),   coverage  goes  down   to  95.5%   leaving  a  gap  of  more   than  9  million  homes.  

Moreover,  Next  Generation  Access  (NGA)  technologies  capable  of  at  least  30  Mbps  are  available  to  53.8%  of  homes  as  of  the  end  of  2012,  while  Docsis  3.0  cable  has  by  far  the  highest  NGA  footprint  (39.4%),  followed  by  VDSL  (24.9%)  and  FTTP  (12.2%).  Expectedly,  broadband   coverage   is   significantly   lower   in   rural   areas.   Standard   broadband   covers  96.1%   while   standard   fixed   broadband   reached   only   83.2%   of   rural   homes.   Wireless  technologies  (satellite  and  mobile  HSPA)  exceed  the  rural  coverage  of  fixed  technologies  in   general.   Moreover,   NGA   remains   very   low   in   rural   areas,   with   12.4%   availability.  Figure  2-­‐1  and  Figure  2-­‐2  illustrate  the  standard  fixed  and  NGA  coverage  in  EU  countries,  respectively.  

 

Figure  2-­‐1:  Fixed  broadband  coverage;  end  of  2012.  (Source:  Point  Topic)  

As   of   January   2013   there   were   144.8  million   fixed   broadband   lines   in   the   EU,   which  corresponds   to   28.8   lines   per   100   inhabitants.   Although   the   annual   growth   has   been  continuously  slowing  down  since  2007,  the  fixed  broadband  market  grew  by  5.5  million  lines   in   2012.   There   is   still   potential   for   further   growth   in   the  market,   as   24%   of   EU  homes   do   not   have   an   internet   subscription.   Figure   2-­‐3   depicts   the   fixed   broadband  penetration  at  the  EU  level  between  years  2004-­‐2013.    

Additionally,   fixed   broadband   penetration   in   the   EU  was   slightly   higher   than   in   Japan  and   just   below   that   of   the   US   as   of   July   2012.   Figure   2-­‐4   shows   the   fixed   broadband  penetration  both  in  EU  and  OECD  countries  as  of  July  2012.  

 

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  10  

 Figure  2-­‐2:  NGA  coverage;  end  of  2012.  (Source:  Point  Topic)  

 Figure  2-­‐3:  Fixed  broadband  penetration;  2004-­‐2013.  (Source:  Communications  Committee)  

Concerning   the   fixed   broadband   market,   xDSL   continues   to   be   the   predominant  technology  in  the  EU  despite  the  decrease  of  its  share  from  80.9%  of  all  fixed  broadband  lines  in  January  2006,  to  73.8%  in  January  2013.  Nevertheless,  the  number  of  xDSL  lines  increased   by   1.7  million   in   2012.   All   of   this   increase   can   be   attributed   to   VDSL   lines,  which  currently  represent  a  mere  3.9%  of  xDSL  lines.    

Cable,   being   the   second  most   widespread   fixed   technology,   has   slightly   increased   its  market  share  from  15.4%  to  17.4%  since  2010.  The  number  of  cable  lines  increased  by  2  million,   slightly   surpassing   xDSL   in   growth   in   2012.   NGA   cable   based   on   DOCSIS   3.0  doubled  in  2012,  as  it  expanded  by  8.4  million  lines  making  cable  the  most  widely  used  NGA  technology  in  the  EU.  By  now,  the  vast  majority  of  European  cable  networks  have  been  upgraded  to  DOCSIS  3.0,  and  two  thirds  of  cable  subscriptions  have  already  been  

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  11  

migrated   to   this   standard.   As   for   the   other   technologies,   fibre   lines   (FTTH   and   FTTB)  went   up   by   31%   in   the   last   12   months,   but   still   represent   only   5.1%   of   all   fixed  broadband  lines.  

 Figure  2-­‐4:  Fixed  broadband  subscription  per  100  inhabitants;  EU  &  OECD;  July  2012.  

(Source:  Communications  Committee  and  OECD  [3])  

In   total,  NGA  technologies,   including  FTTx,  VDSL  and  cable,  are  available  to  54%  of  EU  homes,  but  take-­‐up  is  only  around  12%.    

2.1.2   Mobile  broadband  Concerning   mobile   broadband,   we   focus   on   third   generation   (3G)   HSPA   and   fourth  generation   (4G)   LTE   networks.   On   average,   there   was   96.3%   population   coverage   of  third  generation  HSPA  networks  in  the  EU  in  December  2012,  while  rural  coverage  varies  greatly  among  countries,  but  on  average   it   is  higher  than  any   fixed  technology  [2].  On  the   other   hand,   the   European   coverage   of   fourth  generation   LTE   networks   tripled   in  2012;   currently,   LTE   is   available   to   26.2%   of   population,  while   it  mainly   covers   urban  areas.  Figure  2-­‐5  and  Figure  2-­‐6  illustrate  the  coverage  of  HSPA  and  LTE  in  the  EU  as  of  the  end  of  2012.  

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  12  

 

Figure  2-­‐5:  HSPA  coverage;  end  of  2012.  (Source:  Point  Topic)  

 

Figure  2-­‐6:  LTE  coverage;  end  of  2012.  (Source:  Point  Topic)  

Mobile   broadband   penetration   reached   54.5%   (use   of   handheld   devices   and  computers),   although   the   growth   slowed   down   last   year.  Moreover,   83.4%   of  mobile  broadband  subscriptions  were  used  in  handheld  devices.  Figure  2-­‐7  depicts  the  mobile  broadband   penetration   for   both   technologies   at   EU   level,   per   type   of   device   for   the  years  from  2009  to  2013.  

 

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  13  

 

Figure  2-­‐7:  Mobile  broadband  penetration  at  EU  level;  2009-­‐2013.  (Source:  Communications  Committee)  

2.2 Internet  traffic  analysis  and  forecast  

2.2.1   Technology  evolution    Over  the  last  few  years,  the  rise  and  adoption  of  a  new  computing  paradigm  has  been  observed:   That   of   cloud   computing.   This   includes   new   service   models   such   as  Infrastructure-­‐as-­‐a-­‐Service   (IaaS),   Platform-­‐as-­‐a-­‐Service   (PaaS),   Software-­‐as-­‐a-­‐Service  (SaaS),  or  Network-­‐as-­‐a-­‐Service  (NaaS).    

The   first   three   involve   data   center   virtualization   and   imply   the   provision   of  infrastructure  (e.g.  storage,  computation),  platform  (e.g.  virtual  machines),  or  software  (e.g.  online  storage  applications  such  as  Dropbox3  or  Google  Drive4)  as  services  to  third  parties.  On  the  other  hand,  the  fourth  one,  i.e.  NaaS,  involves  a  network  infrastructure  provider   employing   network   virtualization   by  means   of   software-­‐defined   networking  and  network   function  virtualization  so  as   to  provide  network   resources   (e.g.  access   to  nodes,   links  with  specific  QoS  requirements)   to  his  customers  dynamically   (i.e.  on-­‐the-­‐fly),  in  a  flexible  and  scalable  manner.    

Software-­‐defined  networking  (SDN)  is  a  recently  proposed  approach  (cf.  OpenFlow  [5])  that  enables  network  management   through  an  abstraction  of   lower   level   functionality.  This   is   practically   done   by   decoupling   the   control   plane   (i.e.   the   modules   that   make  decisions  about  traffic  routing),  from  the  data  plane  (i.e.  the  underlying  systems  such  as  routers  and  switches  that  forward  traffic  to  the  selected  destination).  Within  the  context  of   software-­‐defined   networking,   network   functions   virtualization   (NFV)   [6]   was  

                                                                                                               3  https://www.dropbox.com/  4  https://drive.google.com/  

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  14  

introduced  by  ETSI  Industry  Specification  Group  (ISG)  that  proposes  the  virtualization  of  network  functions,  previously  carried  out  by  proprietary,  dedicated  hardware.  

An  example  of  NaaS   is  network   slicing  or  Bandwidth-­‐on-­‐Demand   (BoD).   In  particular,  network   slicing   involves   splitting   of   available   network   capacity   (e.g.   bandwidth)   into  channels,  each  of  which  is  independent  from  the  others,  and  can  be  assigned  to  specific  server  of  device  based  on  their  QoS  requirements  for  a  specific  amount  of  time.  Under  this  model   link   rates   can  be  dynamically  adapted   to   the   traffic  demands  of   the  nodes  connected  to  the  link  in  real-­‐time.  

Moreover,  Virtual  Private  Networks  (VPNs)  constitute  an  older  example  of  the  recent-­‐rising   NaaS,   however   by   means   of   SDN   their   operation   will   be   much   more   flexible.  Finally,  another  case  falling  under  the  scope  of  NaaS  is  mobile  network  virtualization.  In  this  case,  a  Mobile  Network  Operator  (MNO)  builds  and  operates  a  network  and  sells  its  communication   access   capabilities   to   third   parties,   e.g.   Mobile   Virtual   Network  Operators  (MVNOs),  charging  them  by  capacity  utilization.  An  MVNO  is,   thus,  a  mobile  communications   service   provider   that   does   not   own   the   radio   spectrum   or   wireless  network   infrastructure   over   which   it   provides   services;   rather,   it   offers   its  communication   services   using   the   network   infrastructure   of   an   established   Mobile  Network  Operator.  

The  evolution  of  technology  as  described  in  previous  paragraphs  has  led  to  a  significant  increase  of  the  aggregate  IP  traffic  in  core,  metro  and  access  networks.  In  the  following  sections,   we   report   findings   of   Cisco   studies   on   the   volume   of   IP   traffic   that   crosses  today’s  networks,  as  well  as  forecasts  for  the  next  4  years  up  to  2017.    

2.2.2   Global  traffic  Global  IP  traffic  is  forecast  to  surpass  the  zettabyte5  (ZB)  threshold  by  2017  [7],  i.e.,  1.4  ZBs   per   year,   or   120   exabytes   (EBs)   per   month.   In   particular,   global   IP   traffic   has  increased  more  than  fourfold  within  2008-­‐2012,  and  will   increase  threefold  until  2017.  Overall,  IP  traffic  will  grow  at  a  compound  annual  growth  rate  (CAGR)  of  23%  from  2012  to  2017  as  presented  in  Table  2-­‐1.    

The  high  increase  of  data  center  IP  traffic  is  mainly  due  to  the  faster  delivery  of  services  and  data,   increased  performance  of  applications,  and   improved  operational  efficiency.  Indeed  data  center  traffic  dominated  IP  traffic  since  2008  [9],  however,  it  is  undergoing  a  fundamental  transformation  due  to  the  rise  of  cloud  computing  (see  Section  2.2.1).  

Metro-­‐only   traffic   (traffic   that   traverses  only   the  metro  network  of   ISPs   and  bypasses  long-­‐haul  traffic  links)  will  surpass  long-­‐haul  traffic  in  2014  as  it  will  grow  nearly  twice  as  fast  as  long-­‐haul  traffic,  and  it  will  account  for  58%  of  total  IP  traffic  by  2017  (Table  2-­‐2).  The  higher  traffic  growth  in  metro  networks  is  due  in  part  to  the  increasingly  significant  role   of   Content   Delivery   Networks   (CDNs),   which   bypass   long-­‐haul   links   and   deliver  

                                                                                                               5  1  zettabyte  (ZB)  =  10007  bytes  (Bs)  

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  15  

traffic  to  metro  and  regional  backbones.  Specifically,  CDNs  will  carry  over  half,  i.e.,  51%,  of  Internet  traffic  in  2017  from  34%  in  2012.    

Table  2-­‐1:  Global  IP  traffic;  2012-­‐2017.  (Source:  Cisco  VNI,  2013)  

 

Additionally,  Internet  video  traffic  will  be  69%  of  all  Internet  traffic  in  2017,  up  from  57%  in  2012;  note  that  this  percentage  does  not   include  video  exchanged  through  peer-­‐to-­‐peer  (P2P)  file  sharing.  In  particular,  the  sum  of  all  forms  of  video  (TV,  VoD,  Internet,  and  P2P)  will  be  in  the  range  of  80%-­‐90%  of  global  consumer  traffic  by  2017.  Specifically,  it  is  estimated   that   it  will   take   an   individual   over   5  million   years   to  watch   the   amount   of  video  that  will  cross  global  IP  networks  each  month  in  2017,  while  every  second,  nearly  a  million  minutes  of  video  content  will  cross  the  network.  

Table  2-­‐2:  Metro  and  Long-­‐Haul  traffic;  2012-­‐2017.  (Source:  Cisco  VNI,  2013)  

 

2.2.3   Mobile  traffic  Global  mobile  data   traffic  grew  81%   in  2013,  while   reaching  1.5  EBs  per  month  at   the  end   of   the   same   year,   up   from   820   petabytes   (PBs)   per   month   at   the   end   of   2012.  Moreover,  over  half  a  billion  (526  million)  mobile  devices  and  connections  were  added  

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  16  

in   2013,  while  mobile   devices   and   connections   in   2013   grew   to  7  billion,   up   from  6.5  billion  in  2012.  

According  to  [8],  traffic  from  wireless  and  mobile  devices  exceeded  50%  of  IP  traffic   in  2012,   in  contradiction   to   former  projections   that   it  would  surpass   fixed   traffic   in  2016  [7].  Moreover,   in  2013,  a  4G  connection  generated  14.5  times  more  traffic  on  average  than   a   non-­‐4G   connection.   Although   4G   connections   represent   only   2.9   percent   of  mobile  connections  today,  they  already  account  for  30  percent  of  mobile  data  traffic.  

Globally,  mobile  data  traffic  is  expected  to  increase  11-­‐fold,  growing  at  a  CAGR  of  61%  between  2013  and  2018,  reaching  15.9  EBs  per  month  by  2018  [8]  while  by  the  end  of  2014,   the   number   of  mobile-­‐connected   devices  will   exceed   the   number   of   people   on  earth,  and  by  2018  there  will  be  nearly  1.4  mobile  devices  per  capita.  There  will  be  over  10   billion  mobile-­‐connected   devices,   including  machine-­‐to-­‐machine   (M2M)  modules—exceeding  the  world’s  population  at  that  time  (7.6  billion).  

Finally,  over  two-­‐thirds  of  the  world’s  mobile  data  traffic  will  be  video  by  2018.  Mobile  video  will  increase  14-­‐fold  between  2013  and  2018,  accounting  for  69%  of  total  mobile  data  traffic  by  the  end  of  the  forecast  period  (Figure  2-­‐9).  

   

Figure  2-­‐8:  Mobile  IP  traffic;  2013-­‐2018.  (Source:  Cisco  VNI  Mobile,  2014)  

Figure  2-­‐9:  Mobile  video  traffic;  2013-­‐2018.  (Source:  Cisco  VNI  Mobile,  2014)  

2.2.4   Cloud  traffic  Two  major   technology   trends   arose   within   the   last   years:   Cloud   computing   and   data  center   /  network  virtualization.   In   this   section,  we  capture  some  characteristics  of   the  impact  of  these  trends  in  terms  of  IP  traffic  increase.  

Specifically,  in  [9],  it  is  forecasted  that  global  data  center  IP  traffic  will  reach  7.7  ZBs  per  year  or  644  EBs  per  month,  up   from  214  EBs  per  month   in  2012,  by   the  end  of  2017.  Moreover,   about   two   thirds  of   this   traffic  will   be   comprised  of  workload  processed   in  the  cloud;  thus,  global  cloud  IP  traffic  will  reach  in  2017  5.3  ZBs  per  year  or  443  EBs  per  month  up  from  98  EBs  per  month  in  2012.  

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  17  

Note  that  while  the  amount  of  global  traffic  crossing  the  Internet  and  IP  WAN  networks  is  projected  to  reach  1.4  ZBs  per  year   in  2017  [7],   the  amount  of  data  center   traffic   is  already  2.6  ZBs  per  year—and  by  2017,   it  will  triple,  reaching  7.7  ZBs  per  year  [9].  The  higher  volume  of  data  center  traffic  is  due  to  the  inclusion  of  traffic  that  is  not  only  from  the  data  center  to  the  user  and  vice  versa,  but  also  traffic  between  data  centers,  as  well  as  traffic  inside  the  data  center.    

Table  2-­‐3  provides  details  of  the  growth  of  global  data  center  traffic,  overall  and  by  type,  from  2012  to  2017.  It  can  be  observed  that  most  of  the  traffic,  i.e.  up  to  75%,  remains  within  the  data  center.  However,  traffic  that  traverses  IP  WAN  links,  i.e.  traffic  from  the  data  center  to  the  user  and  vice  versa,  and  inter-­‐data  center  traffic  (e.g.  for  replication  purposes),  reaches  25%  of  global  data  center  traffic,  i.e.  1.8  ZBs  per  year  in  2017.  

Significant  promoters  of  cloud  traffic  growth  are  the  rapid  adoption  of  and  migration  to  cloud  architectures,  along  with   the  ability  of  cloud  data  centers   to  handle  significantly  higher  traffic  loads.  Cloud  data  centers  support  increased  virtualization,  standardization,  and  automation.  These  factors  lead  to  increased  performance,  as  well  as  higher  capacity  and  throughput.  

Table  2-­‐3:  Global  data  center  traffic;  2012-­‐2017.  (Source:  Cisco  Global  Cloud  Index,  2013)  

 

Real-­‐time  and  time-­‐sensitive  applications  are  contributing   to   increased  cloud  adoption  in  both  the  business  and  consumer  segments.  For  business,  the  necessity  to  provide  fast  and  flexible  access  to   large  data  archives   is  an   important  objective  for   IT  organizations  considering  cloud-­‐based  solutions.   In  addition,  enabling  advanced  analytics   to   tap   into  the  wealth  of   information  contained  in   largely  unstructured  data  archives  can  create  a  valuable   competitive   business   advantage.   Moreover,   enhanced   collaboration   services  delivered   through   the   cloud   can   increase   employee   productivity   and   customer  satisfaction.    

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  18  

 Figure  2-­‐10:  Personal  content  locker  traffic  growth;  2012-­‐2017.  (Source:  Cisco  Global  Cloud  

Index,  2013)  

On   the   other   hand,   in   the   consumer   space,   applications   such   as   video   and   audio  streaming  are  strong  contributors  to  cloud  traffic  growth,  while  newer  services  such  as  personal   content   lockers  are  also  gaining   in  popularity.   In  particular,  personal   content  lockers  such  as  Dropbox  and  Google  Drive  are  applications  by  means  of  which  users  can  store  and   share  music,  photos,   and  videos  online,   through  an  easy-­‐to-­‐use   interface  at  relatively   low   or   no   cost.   Furthermore,   the   proliferation   of   tablets,   smartphones,   and  other   mobile   devices,   which   allow   access   to   personal   content   lockers   in   a   manner  convenient  to  the  user,  drives  the  increasing  popularity  of  such  applications.  Specifically,  personal  cloud  traffic  is  expected  to  increase  from  1.7  EBs  annually  in  2012  to  20  EBs  in  2017,  at  a  CAGR  of  63%  [9]  (Figure  2-­‐10).  

2.3 Future  traffic  trends  

Broadband   speed   improvement   results   in   increased   consumption   and   use   of   high-­‐bandwidth   content   and   applications.   According   to   [10],   the   global   average   fixed  broadband  speed  will   continue   to  grow  and  will  nearly  quadruple   from  2012   to  2017,  from  11.3  Mbps  to  39  Mbps.  Several   factors  that   influence  the  fixed  broadband  speed  forecast   include   the   deployment   and   adoption   of   FTTx,   high-­‐speed   xDSL   and   cable  broadband  adoption,  as  well  as  overall  broadband  penetration.  On  the  other  hand,  the  average  mobile  broadband   speed   in  2012  was  526  kbps,  while  the  average  speed  will  grow   at   a   CAGR   of   4%,   and   will   exceed   3.9  Mbps   in   2017.   Additionally,   smartphone  speeds,  generally  3G  and  higher,  are  currently  almost  four  times  higher  than  the  overall  average,  while  they  are  expected  to  triple  by  2017,  reaching  6.5  Mbps.  

As   discussed   in   Section   2.1.2   and   Section   2.2.3,   mobile   data   traffic   is   driving   an  exponential   increase   in   data   transmission   through   the   Internet.   This   is   due   to   the  increasing  adoption  of  mobile  devices   like  smartphones  and  tablets,  which  are  moving  away   from  being   purely   "utility"   devices,  with   entertainment   increasingly   occupying   a  central  role  in  the  usage  of  internet  on-­‐the-­‐go.  In  particular,  smartphones  accounted  for  more   than   half   of   all   handset   shipments   in   2013,   and   the   percentage   is   expected   to  

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  19  

continue  to  grow,  mainly  due  to  decreasing  prices  and  the  perceived  value,  and  greater  integration  of  mobile  apps  into  everyday  life  [4].    

Moreover,  the  emergence  of  faster  and  widely  accessible  internet  connection  is  having  an   impact   on   the   television   and   movie   sectors.   In   particular,   concepts   like   SmartTV,  PayTV   or   VoD   (video-­‐on-­‐demand)   have   become   equally   familiar   as   traditional   direct  television  or  movie  viewing.  The  increase  of  the  IP  video  traffic  will  drive  and  accelerate  the  global  IP  traffic  growth  until  2017  [10].  Globally,  IP  video  traffic  will  account  for  73%  of  traffic  in  2017.    

Additionally,  Live  TV  is  currently  distributed  by  means  of  a  broadcast  network,  which  is  highly  efficient  in  that  it  carries  one  stream  to  many  viewers.  Live  TV  over  the  Internet  would  carry  a  separate  stream  for  each  viewer.  Such  a  shift  from  multicast  or  broadcast  to  over-­‐the-­‐top  unicast  would  multiply  the  IP  backbone  traffic  by  more  than  an  order  of  magnitude   [11].   Furthermore,   the   IP   video   traffic   will   be   further   affected   by   the  adoption  of  3DTV  (three-­‐dimensional  TV),  which  is  foreseen  to  take  3  to  5  years  to  gain  momentum  [10].  

The  increase  of  the  popularity  of  IP  video  is  the  underlying  reason  for  accelerated  busy-­‐hour  traffic  growth.  Unlike  other  forms  of  traffic  that  are  spread  evenly  throughout  the  day   (such   as   web   browsing   and   file   sharing),   video   tends   to   have   a   “prime   time”.  Because  of  such  video  consumption  patterns,  the  Internet  has  a  much  busier  busy  hour  as   video   popularity   increases.   Because   video   has   a   higher   peak-­‐to-­‐average   ratio   than  data   or   file   sharing,   and   because   video   is   gaining   traffic   share,   peak   Internet   traffic  grows  faster  than  average  traffic  [10].    

Moreover,   the   use   of   social   networking   sites   has   grown  over   the   past   few   years   and  posting  messages  to  social  media  sites  or  instant  messaging  has  become  one  of  the  most  popular   activities   of   European   internet   users   with   same   levels   of   take-­‐up   as   reading  newspapers   or   internet   banking.   Social   network   services   are   also   one   of   the   driving  factors   behind   the   production   and   uptake   of   online   video   games,   also   called   cloud  games,  supported  by  online  app  stores  and  played  on  general-­‐purpose  devices  including  the   PC   browser,   smart   phones   and   tablets,   and   to   a   lesser   degree   smart   TVs   and   TV  connected  boxes.    

With  traditional  gaming,  graphical  processing  is  done  locally  on  the  gamer’s  computer  or  console.   With   cloud   gaming,   game   graphics   are   produced   on   a   remote   server   and  transmitted  over  the  network  to  the  gamer.  Currently,  online  gaming  traffic  represents  only  0.04%  of  the  total  information  content  associated  with  online  and  offline  game  play  [12].   If   cloud   gaming   takes   hold,   gaming   could   quickly   become   one   of   the   largest  Internet  traffic  categories.  

Another  significant  promoter  of  IP  traffic  growth  is  the  rapid  adoption  of  and  migration  to  cloud  architectures,  along  with  the  ability  of  cloud  data  centers  to  handle  significantly  higher   traffic   loads.   In   order   for   cloud   data   centers   to   support   virtualization,   service  mobility  and  redundancy,  data  replication  and  Virtual  Machine  (VM)  migration  needs  to  be  performed.   Furthermore,  data   center   and   cloud   federation   for   footprint   expansion  

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  20  

and  collaborative  service  provisioning  so  as  to  achieve  energy  efficiency  and   increased  availability   also   contributes   to   the   global   IP   traffic.   The   inter-­‐data   center  communication   necessary   for   the   aforementioned   activities   leads   to   increased  performance,   but   simultaneously   generates   huge   traffic   volumes   to   be   handed   by  metro/core  networks  of  telecom  operators.    

The   increase   in   the   speed   of   broadband   internet   access,   the   development   of   more  modern   devices   and   the  migration   of   multiple   services   and   applications   in   the   cloud  drive   the   increase   of   both   fixed   and   mobile   IP   traffic.   Naturally,   the   trend   toward  mobility   carries   over   into   the   realm   of   fixed  metro/core   networks   as  well,   in   that   an  increasing  portion  of  traffic  will  originate  from  portable  or  mobile  devices.  

The  increased  volume  and  volatility  of  traffic  patterns  generated  by  new  services  and  applications  accentuate  the  pressure  for  technological  development  and  innovation  on  the   part   of   service   providers   and   network   components   vendors.   Eventually,   the  identified   market   and   traffic   trends   constitute   the   necessity   for   high-­‐capacity   optical  components   that   will   support   flexibility,   dynamicity   and   programmability   such   as   the  SPIRIT  transceiver.  

2.4 Application  scenarios  

Considering  the  market  analysis  and  evolution  discussed  in  Section  2.1,  the  technology  evolution  and  the  IP  traffic  analysis  and  forecasting  presented  in  Section  2.2,  we  define  below  three  potential  application  scenarios  for  the  SPIRIT  transceiver.  

2.4.1   Convergence  of  fixed  and  mobile  networks  As  discussed  in  Section  2.1.2,  smartphones  accounted  for  more  than  half  of  all  handset  shipments  in  2013,  and  the  percentage  is  expected  to  continue  to  grow,  mainly  due  to  decreasing  prices  and   the  perceived  value  and  greater   integration  of  mobile  apps   into  everyday  life  [4].  Additionally,  the  number  of  mobile  SIM  cards  grew  by  17.4  million  in  2012,  while  more  than  30%  of  the  growth  came  from  Machine-­‐to-­‐Machine  (M2M)  SIM  cards   [2].  Moreover,   the   enhanced   speed   of   4G   and   the   decreasing   prices   for  mobile  access   by   means   of   3G   is   also   increasing   the   attractiveness   of   games,   video   and  interactive   services.   Furthermore,   the   number   of   MVNOs   grew   by   3.2   million   within  2012.   Therefore,   smartphones,  M2M,   new  mobile   broadband   services,   higher  mobile  broadband   access   speeds   and   lower   prices   are   driving   the   exponential   increase   in  mobile  data  traffic  in  both  Europe  and  worldwide.  

As  discussed  in  Section  2.1  and  Section  2.2,  IP  traffic  both  fixed  and  mobile  will  increase  significantly  within  the  next  4-­‐5  years.  Thus,  in  order  for  telecom  operators  to  efficiently  address   the   increasing   traffic  demand   from  mobile   access  networks,  we  define  as   the  first  application  scenario  for  the  SPIRIT  transceiver  the  convergence  of  fixed  and  mobile  networks.   Specifically,  we   consider   telecom  operators   selling   backhauling   to  MNOs  or  MVNOs,   while   serving   simultaneously   their   own,   business   and   residential,   fixed  customers.   Additionally,   QoS   requirements   of   the   various   types   of   traffic   and   SLAs  

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  21  

established   with   M(V)NOs   need   to   be   fulfilled   dynamically.   The   SPIRIT   optical  transceiver  in  combination  with  SDN  and  network  virtualization  techniques  will  achieve  efficient  and  flexible  management  of  traffic  originated  by  both  fixed  and  mobile  access  networks  in  the  metro  and  core  networks  of  telecom  operators.  

 

Figure  2-­‐11:  Support  of  high-­‐bandwidth  consuming  applications  with  heterogeneous  QoS  requirements  for  a  multitude  of  users  with  different  access  devices.  

2.4.2   Support  of  bandwidth-­‐intensive  end-­‐user  applications  As  estimated   in   [7],   Internet  video  traffic  will  be  69%  of  all   Internet  traffic   in  2017,  up  from  57%  in  2012,  whereas  the  aggregate  volume  of  all  forms  of  video,  i.e.,  live  TV,  VoD,  Internet  (HTTP-­‐Based)  and  P2P,  will  be   in  the  range  of  80%-­‐90%  of  global  consumer  IP  traffic.  Additionally,  it  is  expected  that  over  two-­‐thirds  of  the  world’s  mobile  data  traffic  will  be  video  by  2018  [8].  This  video  streaming  traffic  increase  is  led  by:  

• the  increasing  access  speeds,  both  fixed  and  mobile,  that  will  be  offered  to  end-­‐users  by  the  telecom  operators,    

• the  investment  and  evolution  of  the  audio-­‐visual  industry,    • the  emergence  and  adoption  smart  TV,  i.e.  larger  screen  requires  video  of  higher  

resolution  and  thus,  higher  bitrates,  and  • the  adoption  and  increase  of  mobile  devices  (see  Section  2.1.4).  

At  the  same  time,  apart  from  video  streaming,  another  category  of  applications  that  are  strong  contributors  to   IP  traffic  growth  is  online  personal  storage  such  as  Dropbox  [9].  The   increasing   popularity   of   such   application   is   driven   also   by   the   proliferation   of  tablets,  smartphones,  and  other  mobile  devices,  and  is  of  high  importance  to  SPIRIT  due  to  the  high   impact  of  these  applications   in  terms  of  traffic  volume  on  core,  metro  and  access   links.  Moreover,   social   networking  with  more   than  1   billion   active   users   drives  the   increase   of   interacting   applications   such   as   online   gaming,  which   is   performed   in  non-­‐dedicated  devices  such  as  smartphones  and  tablets.    

The   aforementioned   applications   are   also   responsible   for   the   increase   of   IP   traffic  especially,  at  the  metro  networks  of  telecom  operators,  but  also  at  the  core  network,  as  

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  22  

also   indicated   in   Table   2-­‐2,   which   is   the   application   field   of   the   SPIRIT   transceiver.  Therefore,   the   second   application   scenario   for   the   SPIRIT   transceiver   addresses   the  support   of   high-­‐bandwidth   demanding/consuming   applications,   such   as   video  streaming,  online  personal   storage  applications,  and  online  gaming,  and   is  depicted   in  Figure  2-­‐12.  In  particular,  this  scenario  addresses  the  efficient  and  QoS-­‐aware  support  of  bandwidth-­‐intensive  applications,  concurrently,  for  a  multitude  of  users  that  access  them  by  multiple  devices  with  different  characteristics.  The  explicit  application-­‐driven  network  requirements  are  reported  in  Section  2.5.  

 

 

Figure  2-­‐12:  Support  of  high-­‐bandwidth  consuming  applications  with  heterogeneous  QoS  requirements  for  a  multitude  of  users  with  different  access  devices.  

2.4.3   Support  of  service  and  network  virtualization  We  discussed  in  the  beginning  of  Section  2.2,  the  rapid  adoption  of  cloud  computing  and  the   evolution   of   traditional   networking   architectures   to   virtualized   ones   by  means   of  software-­‐defined   networking   and   network   functions   virtualization   paradigms.   The   so-­‐called  network  virtualization  and  the  adoption  of  cloud  computing  constitute  drivers  for  the  high   increase  of  global   IP   traffic  and  especially   traffic  crossing   the  core  and  metro  networks  of  telecom  operators.    

As   reported   in   [9],   the   amount   of   data   center   traffic   is   2.6   ZBs   per   year,  while   it  will  triple  to  reach  7.7  ZBs  per  year  in  2017.  In  particular,  in  [9],  it  is  observed  that  most  of  the   traffic,   i.e.,   up   to   75%,   remains   within   the   data   center,   whereas   traffic   that  trespasses  IP  WAN  links,  i.e.  traffic  from  the  data  center  to  the  user  and  vice  versa,  and  inter-­‐data  center  traffic,  reaches  25%  of  global  data  center  traffic.  This  25%  corresponds  to  1.8  ZBs  per  year  in  2017,  which  is  higher  that  the  global  public  IP  traffic,  i.e.  1.4  ZBs  per  year  in  2017.  

Thus,   the   third   application   field   for   the   SPIRIT   transceiver   includes   the   dynamic   and  flexible   management   of   the   traffic   generated   by   the   communication   between   data  

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  23  

center  and  end-­‐users  (and  vice  versa),  as  well  as  inter-­‐data  center  communication,  e.g.,  virtual  machines  migration  to  multiple  cloud  locations  for  boost  of  availability  and  QoS  support,   or   content/data   replication   to   alternative   locations   to   achieve   redundancy,  fault-­‐tolerance,  etc.    

Additionally,   the   communication   between   data   centers   and   end-­‐users   includes   a  multitude  of  services  and  applications  that  are  served  by  both  cloud  and  traditional  data  centers,   and   which   employ   heterogeneous   QoS   requirements   depending   also   on   the  type  of  users,   i.e.,   residential,  business  or  enterprise  ones.  For   instance,   in  the  case  of  enterprise   and   business   customers,   apart   from   performance-­‐related   requirements,  security   might   be   also   significant,   e.g.,   for   banking   institutions   or   governmental  organizations;   thus,   traffic  generated  by  the  communication  between  the  data  centers  of  business/enterprise  customers  and  their  users,  or  inter-­‐data  center  traffic  needs  to  be  served   over   dedicated   lines,   either   physically,   e.g.   leased   lines,   or,   more   commonly,  logically,  e.g.,  by  means  of  VPN.    

Therefore,   the   SPIRIT   transceiver   needs   to   be   capable   of   sufficiently   addressing   the  heterogeneous  QoS  requirements  of   these  services  and  applications  dynamically,  on-­‐demand,  and   in  a   scalable  manner,   i.e.,  without  any  performance  deterioration  while  the  user  number   increases  (dramatically).  Figure  2-­‐13   illustrates  the  case  of  a  network  simultaneously   supporting   inter-­‐data   center   communication   between   the   sites   of   an  enterprise,   the   communication   between   the   enterprise   data   centers   and   some  enterprise  users,  as  well  as  the  communication  of  residential  users  with  a  generic  cloud  operator  offering  some  cloud  service,  e.g.  online  personal  storage.  

 

Figure  2-­‐13:  Support  of  service  and  network  virtualization;  flexible  and  dynamic  management  of  data  center  to  user  and  inter-­‐data  center  communication  for  services  

with  heterogeneous  QoS  requirements  for  a  multitude  of  users.  

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  24  

2.5 Application-­‐driven  network  requirements  

Based   on   the   identified   application   scenarios   for   SPIRIT   that   have   been   described   in  Section  2.4,  in  this  section  we  present  the  application-­‐driven  network  requirements  for  the  SPIRIT  component,  which  will  be  translated  to  system  requirements  to  be  addressed  by  the  design  of  the  transceiver  in  Section  3.  

Major  qualitative  requirements  for  the  SPIRIT  optical  transceiver  are:    

• Flexibility  to  sufficiently  address  volatility  of  resources  demand;    

• Dynamicity   to   support   rate-­‐   and   format-­‐adaptation   on-­‐the-­‐fly   based   either   on  monitoring  of  some  pre-­‐defined  KPI,  or  on-­‐demand;  

• Granularity  of  adaptation  per  flow,  end-­‐point  device,  or  network  slice;  

• Programmability  to  support  rate-­‐  and  format-­‐  adaptation;  

• Monitoring  and  troubleshooting;  

• Support  of  Network  Functions  Virtualization  (NFV);  

• Performance  and  scalability.  

In   particular,   concerning   performance,   quantitative   network   requirements   in   terms  of  bitrate   for  video  streaming,  and  online  gaming  are  discussed  below.    Note  that  we  do  not  consider  here  applications  such  as  online  personal  storage  or  online  social  networks,  as  traffic  generated  by  them  can  be  considered  as  delay-­‐tolerant.  

Video  streaming  is  the  most  popular  and  bandwidth-­‐consuming  application  according  to  recent  studies  ([7],  [13]),  while   it  has  significant   impact  also  on  mobile  networks  [8].  A  number  of  online   video   streaming/on  demand  and  digital   download   services  offer  HD  video,   among   them YouTube, Vimeo, Hulu, Amazon   Video   On   Demand, Netflix   Watch  Instantly,  and  others.  Table  2-­‐4  summarizes  the  requirements  of  HTTP  video  streaming  in   terms   of   bitrate.   Note   though   that   due   to   heavy   compression,   the   image   detail  produced  by  these  formats  are  far  below  that  of  broadcast  HD.  

Table  2-­‐4:  Bitrate  of  HTTP  video  streaming.  

Resolution   Video  bitrate   Audio  bitrate   Highest  bitrate  

       

HD  (1920x1080)   3.5-­‐4.5  Mbps   256-­‐320  Kbps   5  Mbps  

HD  (1280x720)   2.5-­‐4.5  Mbps   256-­‐320  Kbps   5  Mbps  

SD  (1280x720)   2.5  Mbps   192  Kbps   3  Mbps  

Concerning   IPTV   bandwidth   requirements,   there   is   no   standard   value,   as   IPTV  bandwidth  may   be   dependent   on   other   added-­‐value   services   accompanying   the   IPTV  service,   the   number   of   simultaneous   streams,   as   well   as   compression   methods.  However,   a   typical   IPTV   service   requires   up   to   4   Mbps   downstream   for   SD   quality  (720x480),  and  up  to  8-­‐10  Mbps  downstream  for  HD  (1920x1080),  while  higher  values,  

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  25  

e.g.   13  Mbps,  have  been   reported   in   literature.  Upstream  can  be  much   lower,   as   it   is  required  only  for  EPG  handling.    

Next,  we  focus  on  online  gaming  and  especially  cloud  gaming.  Cloud  gaming  is  a  newly-­‐emerged  gaming  paradigm,  and  the  most  bandwidth-­‐demanding  form  of  online  gaming,  as   it   combines   cloud   computing   and   online   gaming.   Cloud   gaming   is   real-­‐time   game  playing  via  thin  clients.  Essentially,  cloud  gaming  moves  the  game  logic  required  to  run  a  game   away   from   the   user,   and   into   a   data   center   (the   cloud),   and   then   streams   the  entire   game   experience   to   the   user.   The   streaming   can   be   done   in   three   ways:   i)  Streaming   the   3D   graphics   and   update  messages   to   the   player   (which   is   the   classical  approach),  ii)  encoding  the  game  frames  as  video  and  streaming  the  video  to  the  player,  or   iii)   a   hybrid   approach   of   streaming   graphics   primitives   as   well   as   video  simultaneously.    

According   to   [14],   the   video   streaming   case,   called   Game-­‐as-­‐Video   (GaV),   although   a  new   paradigm,   has   been   commercially   successful   and   is   growing   rapidly.   Using   video  streaming,   the  gamer   is  no   longer  required  to  possess  high-­‐end  gaming  hardware;   the  only  requirement  for  the  client  side  is  broadband  internet  connection  and  the  ability  to  display   high   quality   video.   Therefore,   it   can   be   safely   assumed   that   the   bandwidth  requirements  for  cloud  gaming  are  similar  to  those  of  HD  video  streaming.    

Nonetheless,  cloud  gaming  is  also  highly  time-­‐sensitive,  i.e.,  latency  is  also  of  the  highest  importance,   which   is   not   the   case   for   the   video   streaming   paradigm.   In   particular,  according  to  [15],  100  ms  (milliseconds)  is  the  optimal  latency  which  is  acceptable  by  the  end-­‐users;  by  subtracting  a  20  ms  playout  and  processing  delay  from  the  target  100  ms  latency,  80  ms  is  the  threshold  network  latency  for  cloud  gaming.  

2.6 Reference  network  architecture  

The   IP  network  of  OTE  consists  of  15  BRAS  and  47  ATM  nodes;   it   is  depicted   in  Figure  2-­‐14.  Moreover,  the  backbone  network  of  OTE,  which  is  depicted  in  Figure  2-­‐15,  consists  entirely  of  optical  fibers  with  length  of  27800  km  on  land  and  6000  km  submarine.    

The  transmission  network  of  OTE  currently  has  1960  NG-­‐SDH  nodes,  16  DWDM  local  and  national   rings   with   a   total   capacity   exceeding   2.7   Terabit/s,   while   it   is   continuously  expanding   in   order   to   offer   even   higher   data   speeds   as   well   as   to   support   /   provide  sophisticated  applications  and  services.  Figure  2-­‐16  depicts  an  Ethernet  domain,  which  is  part  of  the  network  of  OTE.  

 

 

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  26  

 

Figure  2-­‐14:  IP  network  of  OTE.  

 

Figure  2-­‐15:  Optical  core  network  of  OTE.  

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  27  

 

Figure  2-­‐16:  Part  of  an  Ethernet  domain  of  OTE.  

 

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  28  

3. System  requirements  and  design    

3.1  System  requirements    

The   SPIRIT   transceiver’s   high   spectral   efficiency,   advanced   flexibility   and   software-­‐defined  capabilities  are  perfectly  suitable  for  elastic  optical  networks  and  pure  coherent  optical   infrastructures.   However,   changes   in   the   deployed   DWDM   networks   are   slow  and   costly,   so   it   is   expected   that   the   SPIRIT   transceiver   will   also   operate   in   legacy  environments.  The  following  scenarios  should  be  addressed:  

1) Scenario  1:  New  generation  coherent  elastic  networks:  These  networks  will  allow  optimal   use   of   SPIRIT   transceiver   functionalities   in   terms   of   flexibility,   spectral  efficiency,  reach  performance,  and  network  utilization.  They  are  characterized  by:  

-­‐ All-­‐coherent   line   interface   (100G,   400G,   1T,   flex-­‐rate)  with   no   legacy  OOK  10G  traffic  (10G  traffic  is  muxponded  by  high  rate  interfaces).  

-­‐ No   inline   optical   dispersion   compensation   (high   non-­‐linear   tolerance,   reduced  optical  amplification,  lower  cost).  

-­‐ Flex-­‐grid:   Spectrum   can   be   allocated   according   to   new   ITU   recommendations.  ROADMs  are  flex-­‐grid.  

-­‐ SMF  or  new  fiber.    

-­‐ Typical  ultra-­‐long  haul  (ULH)  reach  >3000  km.  

2) Scenario  2:  Fixed  Grid  coherent  networks.  These  networks  are  recently  deployed  or  upgraded   in   order   to   better   support   coherent   transmission,   but   without   the  investments  for  flex-­‐grid  operation.  Elasticity  can  still  be  exploited  in  some  way,  but  spectrum  allocation  is  rigid.  

-­‐ All  coherent  interfaces  with  no  legacy  OOK  10G  traffic.  

-­‐ No  in-­‐line  optical  dispersion  compensation.  

-­‐ Fixed,  50GHz  grid.  

-­‐ SMF  (G652),  G655  fiber  (TW,  eLeaf).  

-­‐ Typical  ULH  reach  >  3000  km.  

3) Scenario  3:  Fixed  Grid  legacy  networks.  Many  deployed  networks  are  still  based  on  10G  infrastructure  with  dispersion  compensation.  We  assume  that  no  10G  OOK  line  traffic  will  co-­‐exist  with  400G  or  1T  traffic,  assuming  all  the  legacy  OOK  traffic  will  be  muxed  in  higher  rate  line  interfaces.  

-­‐ All  coherent  interfaces.    

-­‐ In-­‐line  dispersion  compensation.  

-­‐ Fixed,  50GHz  grid.  

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  29  

-­‐ SMF  (G652),  G655  fiber  (TW,  eLeaf).  

-­‐ Typical  ULH  reach  1500-­‐3000  km.  

Recent   deployments   may   have   colorless/directionless   (C/D)   or   even   contention-­‐less  (CDC)   ROADMs,   while   legacy   networks   may   have   directionless   only   for   restoration  purposes,  or  even  simple  ROADMs  with  no  flexibility.  

3.2  System  specification    

The   functional   units   that   will   be   developed   will   ensure   the   transmission   over   the  network  scenarios  listed  in  Section  3.1.  Therefore,  modulation  formats  will  be  selected  in   order   to   guarantee   an   adequate   tolerance   to   signal   to   noise   ratio   (OSNR),   linear  distortions  due  to  optical  fiber  transmission  (GVD,  PMD,  PDL,  etc.)  and  nonlinear  effects  (XPM,   SPM,   etc.),   both   in   the   presence   and   in   the   absence   of   chromatic   dispersion  compensation.    The  following  features  are  requested:  

! Stabilization   of   the   various   circuit   blocks   that  make   up   the   transmitter   and   the  optical  receiver  (temperature,  aging,  etc.).  

! Appropriate   FEC   code   (with   Hard-­‐   and   Soft-­‐Decision   available).   To   comply   with  ULH  reach,  a  BER  <1E-­‐2  is  recommended.  

! Digital   CD   and   PMD   compensation   to   ensure   operability   in   new   generation  networks.   Non-­‐linear   compensation   can   boost   the   reach,   giving   1-­‐2   dB   more  margin:   Back-­‐Propagation   Vs   linearized   channel   inversion   methods   should   be  benchmarked.  

! Ability  to  control  the  signal  quality  and  monitoring  of  linear/nonlinear  distortions  introduced  by  the  optical  link  (GVD,  DGD,  OSNR,  etc.).  This  monitoring  information  is   used  by   the   SDN   controller   to   optimize   transceiver   configuration   and  operate  protection/restoration  switching  mechanism.    DSP  usage  can  also  be  monitored  in  order  to  perform  power-­‐saving  operation  depending  on  performance  required.  

! Compatible  with  Hybrid  Raman/EDFA  amplification.  

3.2.1 Modulation  Formats  

SPIRIT   transceivers   will   be   compatible   with   current   and   near-­‐future   standards.   In  particular,   100Gb/s   implementation   will   be   supported   with   DP-­‐QPSK   at   28/32   Gbaud  (with  appropriate  FEC  overhead).  The  emerging  standard  for  400Gb/s  (and  later  1Tb/s)  must   comply   with   relevant   evolving   standards   (ITU-­‐T,   OIF)   during   the   course   of   the  project.   SPIRIT’s   transceivers   will   explore   possible   avenues   for   400G   and   1T  implementation.  Figure  3-­‐1  illustrates  the  flexibility  of  SPIRIT  transceivers  in  terms  of  the  range   of   possible   configurations,   given   their   capability   in   terms   of   baudrate   (up   to  32GBaud)   and   dual-­‐polarization   QAM   formats,   on   a   single   wavelength   (shaded   area).  The  solid  curves  indicate  the  loci  of  bits/symbol-­‐baudrate  pairs  needed  for  generating  a  400G   channel   (448Gb/s   with   overhead)   with   1   carrier   (black   line),   2   carriers   (red),   3  carriers  (green)  and  4  carriers  (blue).  Thus,  the  intersection  of  the  shaded  area  with  the  

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  30  

loci   in   the  plot  show  the  400G  configurations  possible  with  SPIRIT  devices,   in   terms  of  carrier  number,  baudrate  and  QAM  format.  As  can  be  seen,  400G  is  possible  for  a  SPIRIT  transceiver  even  with  a  single  carrier  configuration  (or  multiple  carriers  and  lower-­‐order  QAM  formats  depending  on  the  transmission  reach).  

 

Figure  3-­‐1:  Showing  the  range  of  combinations  of  spectral  efficiency  (or,  equivalently,  M-­‐QAM  format)  and  symbol  rate  possible  with  SPIRIT  devices  (shaded  area).  

Figure  3-­‐2  quantifies  the  theoretical  transmission  performance  (in  terms  of  the  capacity-­‐reach  product)  as  a  function  of  M-­‐QAM  format  [16].  Clearly,  the  format(s)  chosen  must  be  tailored  to  the  application:   In  ULH  links   it  will  be  necessary  to  keep  the  modulation  order   low   (e.g.   DP-­‐QPSK),   in   order   to  maximize   the   reach.  Metro/access   links,   on   the  other  hand,  where  reach  can  be  sacrificed,  higher  bandwidth  efficiency  can  be  achieved  by  using  DP-­‐16-­‐QAM  and  beyond.  

SPIRIT  should  account  for  the  current  trends  in  research  and  development  of  coherent  optical  systems,  as  well  as  standardization  body  efforts,  in  order  to  select  the  formats  to  be  targeted.  One  of  the  strongest  candidates  for  the  400G  standard  is  dual-­‐carrier  DP-­‐16QAM   at   28/32   Gbaud,   which   should   be   supported   [17].   Standardization   for   1T   is  further   away,   and   SPIRIT   should   therefore   explore   a   range   of   possible   formats   on  multiple   wavelengths.   In   order   to   achieve   realistic   transmission   reaches   while   at   the  same  time  achieving  high  spectral  efficiencies,   formats  up  to  DP-­‐64-­‐QAM  are   targeted  (Table  3-­‐1   lists   the   likely  modulation   formats  and  how  many  carriers  could  be  used  to  achieve  100G,  400G  and  1T  standards).  

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  31  

 

Figure  3-­‐2:  Capacity-­‐Reach  product  vs.  Bandwidth  Efficiency  (modulation  format).  

Table  3-­‐1:  Possible  modulation  formats  to  be  targeted  for  generation  of  current  and  prospective  telecom  standards,  with  one  or  multiple  wavelengths.  

Number  of  optical  carriers/wavelengths  

Standard  1   2   3-­‐4   4-­‐5   8-­‐10  

100G  DP-­‐QPSK  

28/32  GBaud          

400G    DP-­‐16-­‐QAM  28/32  GBaud  

 DP-­‐QPSK  

28/32  GBaud    

1T      DP-­‐64-­‐QAM  28/32  GBaud  

DP-­‐16-­‐QAM  28/32  GBaud  

DP-­‐QPSK  28/32  GBaud  

3.2.2 Multi-­‐Carrier/Superchannel  Generation  

SPIRIT   transceivers   should   support   both   Orthogonal   Frequency   Division   Multiplexing  (OFDM),   as   well   as   Nyquist   WDM   for   multi-­‐carrier/superchannel   generation.   The  superchannel  structures  must  be  selected  in  order  to  comply  with  the  network  scenarios  highlighted.   In   both   cases,   the   modulation   formats   and   number   of   carriers   to   be  supported   should   be   determined   after   theoretical   and   experimental   investigation,  taking   into   account   the   constraints   imposed   by   the   resolution   of   the   multi-­‐level  functionality  of  the  developed  transmitter.    

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  32  

OFDM   and   Nyquist   WDM   may   be   suitable   for   networks   based   on   a   flex-­‐grid  infrastructure,  but  may  be  troublesome  for  use  in  50  Ghz  fixed-­‐grid  deployed  networks.  As   an   alternative,   Ericsson   has   pioneered   a   new   superchannel   approach   called   ‘Time-­‐Frequency   Packing’   [18]   which   is   an   innovative   and   highly   spectral-­‐efficient  implementation   of   the   Faster-­‐than-­‐Nyquist   paradigm.   The   subcarriers   need   not   be  orthogonal,   increasing   superchannel   flexibility   in   terms   of   channel   spacing   and  generation.   Furthermore,   the   modulation   format   used   is   QPSK   which   is   simple   and  robust.  High  spectral  efficiency  is  obtained  by  strong  filtering  and  spectral  shaping  at  the  transmitter,   in   order   to   space   the   subcarriers   closer   than   the   Nyquist   paradigm:   This  introduces   controlled   ISI   that   is   cleared   at   the   receiver   by   appropriate   DSP   based   on  channel   shortening   and   turbo   equalization   techniques.   This   is   in   agreement   with   the  Shannon   theory,   which   postulates   that   the   spectral   efficiency   of   a   communication  system   can  be   improved  by   giving   up   the   orthogonality   condition,   at   least  when   low-­‐order  constellations  are  considered.  

This   technique   also   enables   working   at   high   baud   rates,   with   low   electrical   analog  bandwidth  requirements  (due  to  narrow  filtering).  The  Frequency  Packing  approach  can  be  adapted  to  fit  both  the  flex-­‐grid  and  fixed  50GHz  grid  networks,  thanks  to  the  tuning  flexibility  of  the  subcarriers.  

 

Figure  3-­‐3:  16-­‐QAM  and  Frequency  Packing  spectra  in  fixed  and  flex-­‐grid  implementations.  

Figure  3-­‐3  shows  the  Frequency  Packing  spectrum  in  both  fixed-­‐  and  flex-­‐grid  cases,  and  it   is   compared   to   the   16-­‐QAM   option.   The   spectral   efficiency   is   about   the   same   as  16-­‐QAM,  the  latter  requiring  fewer  carriers,  but  achieving  worse  performance  in  terms  of  reach,  due  to  the  poor  OSNR  tolerance  of  the  16-­‐QAM  format.  

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  33  

Frequency   Packing   can   be   a   valid   alternative   selectable   by   the   SPIRIT   transceiver   to  transport   400G,   as   opposed   to   the   currently   demonstrated   Dual-­‐Carrier   DP-­‐16-­‐QAM:  Four  QPSK  sub-­‐carriers  in  two  50  GHz  slots  can  be  employed,  with  an  expected  back  to  back   OSNR   performance   of   13.5   dB.   Furthermore   the   DAC   functionality   of   the   SPIRIT  transmitter,  as  well  as  its   linearity,  can  enable  better  signal  shaping  at  the  transmitter,  thus  improving  Frequency  Packing  performance.  

Typically,   Frequency   Packing   is   operated   at   a   fixed   baud-­‐rate   and   the   FEC   rate   is  adjusted  to  trade  off  spectral  efficiency  with  reach.  The  SPIRIT  transceiver,  by  means  of  the  SDN  controller,  can  then  select  between  M-­‐QAM  formats,  and  in  addition  configure  to  Frequency  Packing  when  reach  and  high  spectral  efficiency  are  necessary,  adjusting  the   FEC   to   optimize   the   transmission:   Number   of   subcarriers   involved   is   then   also  traded-­‐off.  

Figure  3-­‐4   shows   the   spectral  efficiency  vs.  OSNR   required   to   transport  1  TB  net   rate:  7  bits/s/Hz   can   in   principle   be   obtained.   The   various   points   of   the   plot   correspond   to  different  FEC  overhead.  

 

Figure  3-­‐4:  Spectral  Occupancy  and  Spectral  Efficiency  vs.  OSNR  required  for  1  Tb/s.  

It   is   important   for   the   SPIRIT   transceiver   to   have   flexibility   both   in   the   constellation  space  and  FEC-­‐rate  space.  SPIRIT  DSP  will   implement  both  equalization  techniques  and  FEC.  It  is  important  to  design  these  blocks  in  order  to  be  able  to  switch  on  and  off  parts  of   the   digital   processing   according   to   system   needs,   thus   enabling   forms   of   cognitive  power   saving.   For   example,  when   the   light  path  has   a  high  margin,   hard-­‐decision   FEC  can  be  used  instead  of  soft-­‐decision  FEC,    or  the  number  of  soft-­‐decision  FEC  iterations  can   be   lowered.   Chromatic   dispersion   compensation   should   also   be   selectable.   The  more  modular   the   DSP,   the   easier   it  will   be   to   control   the   power   consumption   of   its  constituent  parts.  

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  34  

3.2.3 Software-­‐Defined  Operation  

The  SPIRIT  transceiver  flexibility   in  terms  of  spectral  occupancy  and  data  rate  provides  optical  networks  with  new  optimization  operations  as  well  as  protection  schemes.  The  following  scenarios  should  be  addressed:  

! Distance-­‐adaptive  modulation:  The  transceiver  shall  enable  the  capability  to  adjust  its  modulation  format  and  FEC  overhead,  depending  on  the  reach  of  the  lightpath.  Trading  off  spectral  efficiency  with  reach  will  improve  network  utilization.  A  shorter  path  may  be  served  by  a  high  level  constellation  format  saving  spectrum  for  other  traffic.  Alternativley,  FEC  can  be  traded  off  with  payload  if  no  additional  spectrum  is  available   for  a  given   reach.     This   flexibility  will  be  managed  by   the  Transport-­‐SDN  controller.  

! Sub-­‐Rate   Protection:   A   new   protection   scheme   shall   be   enabled   by   the   SPIRIT  flexible  transceiver.  When  no-­‐protection  path  can  be  found  by  standard  protection  schemes   due   to   reach   or   spectral   constrains,   the   rate   of   the   protected   traffic   is  reduced,   losing  part  of   the  traffic   (possibly   the   lowest  priority  one),  but  still  being  able   to   protect   the   relevant   traffic.   Sub-­‐rate   protection   may   be   achieved   by  increasing  FEC,  reducing  baud  rate,   lowering  the  level  of  modulation  format,  or  by  slicing  some  of  the  subcarriers  of  the  superchannel.  

! Network  Monitoring:  SDN  functionalities  require  monitoring  of  the  network  status  to   select   the   proper   configuration.   The   coherent   receiver   enables   physical   layer  parameter  monitoring.  CD  and  PMD  monitoring   techniques  are  well-­‐known   in   the  literature,   while   OSNR   and   Non-­‐Linearity   monitoring   is   still   troublesome.  Furthermore,  techniques  to  get  local  information  from  end-­‐to-­‐end  information  are  required:   In   fact,  a  coherent  receiver  gives  estimates  over  the  whole   link,  and  not  on  a  span-­‐by-­‐span  basis.  By  combining  information  of  different  end-­‐to-­‐end  paths,  it  is   possible   to   derive   local   information   to   be   used   for   new   path   feasibility   and  assessment   (e.g.   network   kriging).   Distributed   monitoring   techniques   may   be  investigated  to  minimize  the  paths  probing.  An  accurate  assessment  of  a   lightpath  status  enables  SDN  optimizations  at  the  network,  as  well  as  the  transceiver  level.  

3.3  Test  Cases  

3.3.1   Laboratory  Testing  at  Ericsson  

The  SPIRIT  prototype’s  performance  will  be  assessed  considering  the  network  scenarios  described  above,  focusing  on  metro/core  networks.  It  is  suggested  that  a  lab  set-­‐up  be  employed,   emulating   a   small   meshed   network   (Figure   3-­‐5)   with   real   fiber   and  attenuators.  One  of  the  spans  of  the  network  is  replaced  by  a  recirculating  loop  so  that  arbitrary   distances   can   be   covered   for   many   network   lightpaths.   This   way,   distance-­‐adaptive  modulation  and  sub-­‐rate  protection  can  be  emulated,  as  well  as  other  flexible  reconfigurations.  The  node  switches  (ROADMs)  can  be  controlled  by  the  SDN  controller.  

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  35  

A  traditional  multi-­‐span  point-­‐to-­‐point  ULH  link  can  also  be  set  up  especially  for  testing  with  real  commercial  equipment,  as  shown  in  Figure  3-­‐6.  Signals  generated  by  the  SPIRIT  transceiver   will   be   transmitted   along   with   traffic   from   commercial   transponders   by  means  of  a  wavelength-­‐selective  switch  (WSS),  and  propagated  over  a  physical  link.  

 

Figure  3-­‐5:  Meshed  network  topology  laboratory  setup  for  performance  evaluation  of  the  SPIRIT  transceiver.  

 

Figure  3-­‐6:  Point-­‐to-­‐point  link  for  SPIRIT  transceiver  validation.  

 

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  36  

3.3.1.1     The  Ericsson  SPO  Packet/optical  Transport  Platform:  System  Overview  

SPIRIT  devices  will  be  tested  in  Ericsson’s  SPO  product  family,  a  packet  optical  integrated  platform  with  flexible  DWDM  capabilities  for  core  and  metro  applications.  It  supports  a  maximum  of  96  channels  in  the  C-­‐band  on  a  50GHz  channel  grid  as  well  as  flex  grid.  10G  and   100G   WDM   interfaces   are   available.   100G   is   based   on   best   in   class   coherent  technology.  Further  general  information  on  the  product  is  available  at  the  Ericsson  web-­‐site6.  

 

 

Figure  3-­‐7:  SPO  main  subrack.  

The  photonic   infrastructure   is   implemented   in  a  unit   called  Photonic  Attachment  Unit  (PAU)  which   can   be   configured   to   operate   as   an  Optical   Terminal   (OTA),   Optical   Line  Amplifier  (OLA)  or  RROADM  (see  Figure  3-­‐8).  This  unit  is  a  separate  subrack  that  can  be  used   in  any  network  node  as   standalone  network  element,  or   in   conjunction  with   the  main  POTP  (Packet  Optical  Transport  Platform)  subrack  which  houses  the  transponders,  muxponders,  packet  card,  OTN  units,  etc.  

 

Figure  3-­‐8:  SPO  Photonic  Attachment  Units  (PAU).                                                                                                                  6  http://www.ericsson.com/us/ourportfolio/products/spo-­‐1400-­‐family  

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  37  

3.3.1.2     Multiplexing  and  de-­‐multiplexing  

SPO  has  a  wide  range  of  multiplexing  options  ranging  from  thin-­‐film  filters  combining  a  pair  of  channels  (2  skip  0  filters),  to  48  channel  100GHz-­‐multiplexers  based  on  AWG.  The  latter,  when  combined  with  a  group  inter-­‐leaver  units  (GIU),  provide  the  most  efficient  method  to  multiplex  and  de-­‐multiplex  96  channels  on  a  50GHz  grid.    

Additionally,   SPO   supports   meshed   network   topologies   with   a   range   of   Wavelength  Selective  Switch  (WSS)  solutions  to  enable  Reconfigurable  Optical  Add-­‐Drop  Multiplexer  (ROADM)  nodes  (see  Table  3-­‐2).  Flex-­‐grid-­‐ready  WSSs  are  also  available  with  SPO.  

Table  3-­‐2:  SPO  RROADM  configurations.  

RROADM  type   3  dB  BW  (GHz)  

2  ways  RROADM  version  house  a  2x1  WSS   -­‐  

4  ways  RROADM  version  house  a  4x1  WSS   32  

9  ways  RROADM  version  house  a  9x1  WSS   35  

9  ways  RROAD  Flexgrid  (N  slices)   12.5*N-­‐15  

3.3.1.3     Optical  amplifiers  

SPO  has  a  range  of  EDFA  amplifier  options  with  varying  operating  gain  to  cover  different  fiber  span  lengths.    A  single-­‐stage  amplifier  (SSA)  with  variable  gain  and  output  power,  but  no  mid-­‐stage  access  for  Dispersion  Compensating  Modules  (DCMs),  and  a  dual-­‐stage  amplifier   (DSA)   with   up   to   10.5dB   loss   budget   for   DCMs   are   available.   Each   of   these  EDFAs  has  an  embedded  Variable  Optical  Attenuator  (VOA)  to  allow  for  deployment  on  a   range   of   spans   with   different   losses,   and   to   compensate   for   variation   of   span  attenuation  over  the  lifetime  of  the  system.    

 In   the  DSA,   the   first   stage   contains   the   embedded  VOA  and  operates  with   a   variable  gain,  such  that  its  output  power  is  significantly  lower  than  that  of  the  second  stage,  to  suppress   any   nonlinear   effects   in   the   DCF.   The   SPO   product   family   also   provides   a  Raman  amplification  solution:  A  700mW  pump  power  capable  of  14.5  dB  gain  on  G.652  fibers.The  EDFA  parameters  of  the  SSA  and  DSA  modules  are  summarized   in  Table  3-­‐3  and  Table  3-­‐4,  respectively:    

Table  3-­‐3:  EDFA  parameters  of  SSA  modules.  

Amplifier  type     SSA-­‐20/20   SSA-­‐27/20  

Max  power   dBm   20.5   20.5  

Gain  range   dB   15-­‐25   22-­‐32  

NF  @  G  min   dB   8.4   6.2  

NF  @  G  nom   dB   6   5.3  

NF  @  G  max   dB   5.4   5.3  

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  38  

Table  3-­‐4:  EDFA  parameters  of  DSA  modules.  

Amplifier  type     DSA-­‐20/20   DSA-­‐27/20  

Max  power   dBm   20.5   20.5  

Gain  range   dB   15-­‐25   22-­‐32  

NF  @  G  min   dB   10.6   7.2  

NF  @  G  nom   dB   7.6   6.1  

NF  @  G  max   dB   6.5   5.7  

3.3.1.4     Dispersion  Compensation  Modules  

SPO   allows   for   the   introduction   of   DCMs   to   provide   dispersion   compensation   for  systems  operating  at  10Gbit/s  and  40G.    DCMs  are  available  for  SMF,  TW-­‐RS  and  eLEAF  fibre  types  with  a  range  of  lengths.    SMF  DCMs  are  available  up  to  a  maximum  length  of  150km   with   a   granularity   of   10km.   TW-­‐RS   and   eLEAF   modules   are   available   with  maximum  lengths  of  240km  and  200km,  respectively,  both  with  granularities  of  40km.      

Dispersion  compensating  fibre  parameters  are  given  in  Table  3-­‐5:  

Table  3-­‐5:  EDFA  parameters  of  DSA  modules.  

Fibre   Dispersion  Dispersion  

slope  Dispersion  curvature  

Loss  Effective  Area  

  ps/nm/km   ps/nm2/km   ps/nm2/km   dB/km   μm2  

SMF   -­‐113.7   -­‐0.41   0.0   0.47   19  

TW-­‐RS   -­‐150   -­‐1.41   -­‐0.007   0.65   13  

eLEAF   -­‐110   -­‐2.375   -­‐0.04   0.8   13  

3.3.2   Laboratory  Testing  at  ICCS/NTUA  A  high-­‐capacity   transmission   and   coherent   reception   test-­‐bed  will   be   implemented   at  ICCS/NTUA   and   will   be   used   in   the   evaluation   of   the   SPIRIT   transceivers.   This   will  support   testing  of  dual-­‐polarization  QAM  signals  at   symbol   rates  up   to  32GBaud.  Two  options  will   be   pursued   for   generating   the   high-­‐speed   binary   data   signals   needed   for  multi-­‐level   QAM   operation:   (a)   Using   FPGA   development   boards   equipped  with   serial  transceivers  capable  of  up  to  28Gbit/s  operation  (Altera  Stratix  V  GT  and/or  Xilinx  Virtex  7),  and  (b)  Using  high-­‐speed  pulse  pattern  generator   (PPG)  outputs.   In   the  case  of   (a),  the  symbol  rate  possible  is  limited  to  28  GBaud,  while  case  (b)  allows  much  higher  rates.  The   use   of   an   FPGA,   however,   enables   programmable,   flexible   operation   and  transmitter-­‐side   DSP,   whereas   available   PPGs   at   these   speeds   only   provide   PRBS  outputs.   An   option   for   overcoming   this   rate   limitation   of   the   FPGA   is   the   use   of  commercial  2:1  or  4:1  RF  multiplexers.  

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  39  

The   test   signals   will   be   transmitted   through   a   recirculating   fiber   loop,   in   order   to  emulate   propagation   through   multiple   spans,   and   several   hundreds   of   kilometres  (mimicking  typical  metro/core  network  scenarios).  The  signals  will  then  be  received  and  digitized   with   a   33   GHz,   80   GSa/s   real-­‐time   oscilloscope   for   further   offline   DSP  processing.   ICCS/NTUA  will  utilize   its  existing  expertise  and  suite  of  DSP  algorithms,  as  well   as   those  developed  during   the   course  of   the  project,   and   required   specifically  by  SPIRIT   for  modulation/demodulation  of   the  high-­‐order,   single/multi-­‐carrier  M-­‐QAM  or  arbitrary  constellations  that  will  be  employed.  

SPIRIT   devices   will   be   tested   for   suitability   in   future   software-­‐defined   flexible   optical  networks,  including  dynamic  adjustment  of  the  baudrate  and  QAM  format,  performance  monitoring  in  DSP  and  determination  of  Quality  of  Service.  System  parameters  that  will  be   evaluated   will   include   OSNR,   transmission   reach,   extinction   ratio   and   dispersion,  polarization  crosstalk  and  BER  for  variable  modulation  formats.  Back-­‐to-­‐back,  as  well  as  transmission   performance   will   be   assessed,   considering   the   trade-­‐off   between  modulation  type  and  reach.  

3.3.3   Field  Trial  Testing  

The  SPIRIT  transceiver  will  be  tested  in  field  trials  in  order  to  investigate  its  performance  under   real   transmission   conditions   and   identify   performance   trade-­‐offs.   The   primary  objective   of   the   field   trials   is   to   test   the   provided   optical   platform   in   a   real   DWDM  system   configuration,   focusing   on   interoperability   of   existing   10,   40   and   100  Gbit/s  channels  with  high  capacity  channels  generated  by  SPIRIT  devices.  This  will  be  done  by  inserting   them   into  a  DWDM  system   installed   in   the   regional  OTE  network,  employing  G.652  fiber.  This  network  will  be  used  for  testing  the  transceiver  unit  in  a  deployed  plant  environment.   The   interoperability   trial   will   be   crucial   to   clearly   evaluate   network  migration   scenarios   and   share   the   benefits   of   the   new   technology,   as   well   provide  clearer  ideas  of  potentiality  and  limits.  

The   channels   provided   by   the   developed   optical   platform   will   be   tested   under   real  impairments  (optical  noise  from  amplifiers,  CD,  PMD,  non-­‐linear  interaction,  reflections,  filter  cascading  distortion,  high  insertion  loss  for  optical  splices  and  connectors).  

 

 

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  40  

4. Conclusions  

In  this  deliverable,  we  have  defined  the  major  application  scenarios  of  interest  to  SPIRIT;  we   identified   the   application-­‐driven   network   requirements  mainly   in   terms   of   bitrate  and   latency   and   described   the   reference   network   architecture   of   OTE,   which   will   be  considered   in   the   evaluation   and   demonstration   of   the   functionality   of   the   SPIRIT  component.  

In  particular,  as  discussed  in  Section  2,  the  increased  Internet  usage  due  to  a  multitude  of   factors   e.g.   emergence   of   new   applications   such   as   cloud   gaming,   increasing  penetration  of  broadband  (fixed  and  mobile),  increasing  number  of  mobile  devices  and  M2M,  etc.,  coupled  with  faster  speeds  are  accelerating  recent  trends  in  Internet  traffic  volumes,   as   well   as   they   are   inducing   new   and   different   behaviour   traffic   patterns.  Moreover,   consumers’   increased  demand   for   services,   information   and  entertainment  "anytime,   anywhere",   is   putting   pressure   not   only   on   service   providers,   but   also   to  components   vendors   for   constant   innovation  and   specifically,   for  high-­‐capacity  optical  components   that   will   support   flexibility,   dynamicity   and   programmability   such   as   the  SPIRIT  transceiver.  

Major  application  scenarios  described  in  Section  2,  include  the  flexible  management  of  fixed  and  mobile  traffic  in  the  core/metro  network  of  network  operators,  the  support  of  bandwidth-­‐intensive  applications,   including  both  end-­‐user   application  as  well   as   inter-­‐data  center  services,  and  the  support  of  service  and  network  virtualization,  

Furthermore,   the   requirements   from   a   system   perspective   have   been   presented   in  Section   3,   and   appropriate   test   cases   for   performance   evaluation   of   SPIRIT’s   devices  have  been  identified.  Operation  in  a  fixed-­‐grid  WDM  environment  is  envisaged,  both  in  a   mixed   coherent/legacy   traffic   scenario,   as   well   in   an   all-­‐coherent   channel   setting.  Moreover,  the  requirement  for  programmable  and  flexible-­‐format  operation  will  ensure  compatibility  with   future   elastic   optical   networks,  with   an   emphasis   on   the   emerging  flex-­‐grid  standard.  

Evaluation  of  SPIRIT’s  devices  will  take  place  both  in  the  laboratory,  as  well  as  in  a  field  trial  with  deployed  fiber  and  co-­‐propagating  traffic  from  commercial  transponders.  The  test  cases  identified  aim  to  evaluate  component  performance,   in  addition  to  validating  SPIRIT’s  targeted  system  concepts  in  realistic  network  scenarios.  

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  41  

5. List  of  Figures  Figure  2-­‐1:  Fixed  broadband  coverage;  end  of  2012.  (Source:  Point  Topic) ................................... 9  

Figure  2-­‐2:  NGA  coverage;  end  of  2012.  (Source:  Point  Topic) ..................................................... 10  

Figure  2-­‐3:  Fixed  broadband  penetration;  2004-­‐2013.  (Source:  Communications  Committee)... 10  

Figure  2-­‐4:   Fixed  broadband  subscription  per  100   inhabitants;  EU  &  OECD;   July  2012.   (Source:  Communications  Committee  and  OECD  [3]) ................................................................................. 11  

Figure  2-­‐5:  HSPA  coverage;  end  of  2012.  (Source:  Point  Topic).................................................... 12  

Figure  2-­‐6:  LTE  coverage;  end  of  2012.  (Source:  Point  Topic)....................................................... 12  

Figure   2-­‐7:   Mobile   broadband   penetration   at   EU   level;   2009-­‐2013.   (Source:   Communications  Committee) ................................................................................................................................... 13  

Figure  2-­‐8:  Mobile  IP  traffic;  2013-­‐2018.  (Source:  Cisco  VNI  Mobile,  2014) ................................. 16  

Figure  2-­‐9:  Mobile  video  traffic;  2013-­‐2018.  (Source:  Cisco  VNI  Mobile,  2014) ........................... 16  

Figure   2-­‐10:   Personal   content   locker   traffic   growth;   2012-­‐2017.   (Source:   Cisco   Global   Cloud  Index,  2013)................................................................................................................................... 18  

Figure   2-­‐11:   Support   of   high-­‐bandwidth   consuming   applications   with   heterogeneous   QoS  requirements  for  a  multitude  of  users  with  different  access  devices........................................... 21  

Figure   2-­‐12:   Support   of   high-­‐bandwidth   consuming   applications   with   heterogeneous   QoS  requirements  for  a  multitude  of  users  with  different  access  devices........................................... 22  

Figure  2-­‐13:  Support  of  service  and  network  virtualization;  flexible  and  dynamic  management  of  data  center  to  user  and  inter-­‐data  center  communication  for  services  with  heterogeneous  QoS  requirements  for  a  multitude  of  users. ......................................................................................... 23  

Figure  2-­‐14:  IP  network  of  OTE. .................................................................................................... 26  

Figure  2-­‐15:  Optical  core  network  of  OTE. .................................................................................... 26  

Figure  2-­‐16:  Part  of  an  Ethernet  domain  of  OTE. .......................................................................... 27  

Figure  3-­‐1:  Showing  the  range  of  combinations  of  spectral  efficiency   (or,  equivalently,  M-­‐QAM  format)  and  symbol  rate  possible  with  SPIRIT  devices  (shaded  area)........................................... 30  

Figure  3-­‐2:  Capacity-­‐Reach  product  vs.  Bandwidth  Efficiency  (modulation  format). ................... 31  

Figure  3-­‐3:  16-­‐QAM  and  Frequency  Packing  spectra  in  fixed  and  flex-­‐grid  implementations. ..... 32  

Figure  3-­‐4:  Spectral  Occupancy  and  Spectral  Efficiency  vs.  OSNR  required  for  1  Tb/s. ................ 33  

Figure  3-­‐5:  Meshed  network  topology  laboratory  setup  for  performance  evaluation  of  the  SPIRIT  transceiver..................................................................................................................................... 35  

Figure  3-­‐6:  Point-­‐to-­‐point  link  for  SPIRIT  transceiver  validation. .................................................. 35  

Figure  3-­‐7:  SPO  main  subrack........................................................................................................ 36  

Figure  3-­‐8:  SPO  Photonic  Attachment  Units  (PAU). ...................................................................... 36  

   

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  42  

6. List  of  tables  Table  2-­‐1:  Global  IP  traffic;  2012-­‐2017.  (Source:  Cisco  VNI,  2013) ............................................... 15  

Table  2-­‐2:  Metro  and  Long-­‐Haul  traffic;  2012-­‐2017.  (Source:  Cisco  VNI,  2013) ........................... 15  

Table  2-­‐3:  Global  data  center  traffic;  2012-­‐2017.  (Source:  Cisco  Global  Cloud  Index,  2013) ....... 17  

Table  2-­‐4:  Bitrate  of  HTTP  video  streaming. ................................................................................. 24  

Table  3-­‐1:  Possible  modulation  formats  to  be  targeted  for  generation  of  current  and  prospective  telecom  standards,  with  one  or  multiple  wavelengths................................................................. 31  

Table  3-­‐2:  SPO  RROADM  configurations. ...................................................................................... 37  

Table  3-­‐3:  EDFA  parameters  of  SSA  modules. ............................................................................... 37  

Table  3-­‐4:  EDFA  parameters  of  DSA  modules. .............................................................................. 38  

Table  3-­‐5:  EDFA  parameters  of  DSA  modules. .............................................................................. 38  

 

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  43  

7. Abbreviations  

BoD     Bandwidth-­‐on-­‐Demand  

CAGR     Compound  Annual  Growth  Rate  

CDN     Content  Delivery  Network  

DA     Digital  Agenda  

DOCSIS     Data  over  Cable  Service  Interface  Specification  

FTTB     Fibre-­‐To-­‐The-­‐Building  

FTTH     Fibre-­‐To-­‐The-­‐Home  

FTTP     Fibre-­‐To-­‐The-­‐Premises  

GaV     Game-­‐as-­‐Video  

HSPA     High  Speed  Packet  Access  

ISP     Internet  Service  Provider  

LTE     Long  Term  Evolution  

M2M     Machine-­‐to-­‐Machine  

MNO     Mobile  Network  Operator  

MVNO     Mobile  Virtual  Network  Operator  

NGA     Next  Generation  Access  

OECD     Organization  for  Economic  Co-­‐operation  and  Development  

OFDM     Orthogonal  Frequency  Division  Multiplexing  

SIM     Subscriber  Identity  Module  

VDSL     Very-­‐high-­‐bit-­‐rate  Digital  Subscriber  Line  

VoD     Video-­‐on-­‐Demand  

VPN     Virtual  Private  Network  

 

 

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  44  

8. References  [1] www.spirit-­‐project.eu  

[2] Digital   Agenda   Scoreboard   2013   –   Chapter   2:   Broadband   markets,  http://ec.europa.eu/digital-­‐agenda/sites/digital-­‐agenda/files/DAE%20SCOREBOARD%202013%20-­‐%202-­‐BROADBAND%20MARKETS%20_0.pdf  

[3] Organization  for  Economic  Co-­‐operation  and  Development:  http://www.oecd.org/  

[4] Digital  Agenda   Scoreboard  2013  –  Chapter   5:  Online   content,   http://ec.europa.eu/digital-­‐agenda/sites/digital-­‐agenda/files/DAE%20SCOREBOARD%202013%20-­‐%205-­‐ONLINE%20CONTENT_0.pdf  

[5] McKeown,  Nick,  Tom  Anderson,  Hari  Balakrishnan,  Guru  Parulkar,  Larry  Peterson,  Jennifer  Rexford,   Scott   Shenker,   and   Jonathan   Turner.   "OpenFlow:   enabling   innovation   in   campus  networks."  ACM  SIGCOMM  Computer  Communication  Review  38,  no.  2  (2008):  69-­‐74.  

[6] Chiosi,   M.,   Clarke,   D.,   Willis,   P.,   Reid,   A.,   Feger,   J.,   Bugenhagen,   M.,   ...   &   Sen,   P.   (2012,  October).  Network   functions  virtualisation:  An   introduction,  benefits,  enablers,   challenges  and  call  for  action.  In  SDN  and  OpenFlow  World  Congress.  

[7] Cisco   Visual   Networking   Index:   Forecast   and   Methodology,   2012   –   2017   (white   paper):  http://www.cisco.com/en/US/solutions/collateral/ns341/ns525/ns537/ns705/ns827/white_paper_c11-­‐481360.pdf  

[8] Cisco   Visual   Networking   Index:   Global   Mobile   Data   Traffic   Forecast   Update,   2013-­‐2018,  white   paper,   February   5th,   2014:  http://www.cisco.com/en/US/solutions/collateral/ns341/ns525/ns537/ns705/ns827/white_paper_c11-­‐520862.pdf  

[9] Cisco   Global   Cloud   Index:   Forecast   and   Methodology,   2012-­‐2017,   white   paper,   2013:  http://www.cisco.com/en/US/solutions/collateral/ns341/ns525/ns537/ns705/ns1175/Cloud_Index_White_Paper.pdf  

[10] Cisco   Visual   Networking   Index:   The   Zettabyte   Era   –   Trends   and   Analysis,   May   2013:  http://www.cisco.com/en/US/solutions/collateral/ns341/ns525/ns537/ns705/ns827/VNI_Hyperconnectivity_WP.pdf  

[11] Gerber,  Alexandre,  and  Robert  Doverspike   (AT&T).   "Traffic   types  and  growth   in  backbone  networks."  Optical  Fiber  Communication  Conference  and  Exposition  (OFC/NFOEC),  2011  and  the  National  Fiber  Optic  Engineers  Conference.  IEEE,  2011.    

[12] Bohn,   Roger   E.,   and   James   E.   Short.  How  Much   Information?:   2009   Report   on   American  Consumers.  University  of  California,  San  Diego,  Global  Information  Industry  Center,  2009.  

[13] Sandvine:   Global   Internet   Phenomena   Report   –   2H   2013:  https://www.sandvine.com/downloads/general/global-­‐internet-­‐phenomena/2013/2h-­‐2013-­‐global-­‐internet-­‐phenomena-­‐report.pdf  

[14] Hemmati,  Mahdi,  Abbas   Javadtalab,  Ali  Asghar  Nazari  Shirehjini,   Shervin  Shirmohammadi,  and   Tarik   Arici.   "Game   as   video:   bit   rate   reduction   through   adaptive   object   encoding."  

SPIRIT  –  D2.1:  Report  on  application  test  cases,  system  requirements  and  design  for  400G/Terabit  transmission     Page  45  

In  Proceeding  of   the  23rd  ACM  Workshop  on  Network  and  Operating  Systems  Support   for  Digital  Audio  and  Video,  pp.  7-­‐12.  ACM,  2013.  

[15] Choy,   Sharon,   Bernard   Wong,   Gwendal   Simon,   and   Catherine   Rosenberg.   "The   brewing  storm  in  cloud  gaming:  A  measurement  study  on  cloud  to  end-­‐user  latency."  In  Proceedings  of  the  11th  Annual  Workshop  on  Network  and  Systems  Support  for  Games,  p.  2.  IEEE  Press,  2012.  

[16] Kuang-­‐Tsan   Wu   et   al.,   “The   Age   of   Optical   Coherent   Communication,”   CLEO   Technical  Digest  2012,  paper  CF1F.1.  

[17]  “Path   to   400G,”   Fujitsu   presentation   (available   online:  http://www.fujitsu.com/downloads/TEL/fnc/whitepapers/Pathto400G.pdf).    

[18] L.   Poti   et   al.,   “Casting   1   Tb/s   DP-­‐QPSK   Communication   into   200   GHz   Bandwidth,”   ECOC  Technical  Digest  2012,  paper  P4.19.