Development of CEF networks design

1
Development of CEF networks design

• Stanislav Sima
• Lada Altmannova
• Prague May 17th, 2005

2
Previous presentations available on-line I.

• Development of the CESNET2 optical network.
  TERENA Networking Conference, Limerick, June 4,
  2002.
• (Nothing-In-Line approach 189 km NIL operation
  announced)
• http//www.terena.nl/conferences/tnc2002/Slides/sl
  3c2.ppt
• Long distance fiber connections in NREN, TF-NGN
  Budapest, October 18, 2002.
• (10 Gb/s NIL possible up to about 300 km, single
  fibre bi-directional transmission in production
  network up to 125 km)
• http//www.terena.nl/tech/task-forces/tf-ngn/prese
  ntations/tf-ngn9/sima.ppt
• Fibres and advanced optical devices for a new
  networking strategy. TERENA TF-NGN Cambridge,
  September 16th, 2003.
• (Fibre is strategic asset. CEF networks concept.
  Free space optics for cost-effective first mile.
  Connect Near over Border by fibre.)
• http//www.terena.nl/tech/task-forces/tf-ngn/prese
  ntations/tf-ngn12/20030915_SS_Optical.pdf

3
Previous presentations available on-line II.

• Procurement and lighting of dark fibre. CEF
  Networks workshop, Prague May 2004.
• (Double payment of fibre infrastructure? NIL
  results. CESNET-made optical amplifiers
  deployment announced.)
• http//www.ces.net/doc/seminars/20040525/pr/CEF20
  Networks3ext.ppt
• Towards a Nation-wide Fibre Footprint in research
  and education networking. TERENA Networking
  Conference, Rhodes, June 9, 2004.
• (Dark fibres brings progress to REN design and
  subsequently to WAN design generaly. From naive
  economy to strategy. National fibre footprint in
  European NRENs. GLIF and CzechLight.)
• http//www.terena.nl/conferences/tnc2004/programme
  /presentations/show.php?pres_id97

4
Network design

• Specifications of three items are needed
• Network services,
• building elements,
• implementation (deployment, setup, )
• Top-down design process starts with network
  services specification, botom-up design process
  starts with building elements specification
• Neither of above two start points is the best
  !!!
• Results of one way design are suboptimal (i.e.
  bad), returns and corrections of decisions in
  design process are needed !!!
• Experience and theoretical knowledge of circuits
  design and computer design help to deeply
  understand above problem

5
Legacy in network design

• Buiding elements, implementation and operation of
  large scale networks are procured
• Procurement is long process (especially if public
  funding is used) e.g. 7-24 month
• Returns and corrections of decisions are very
  difficult or impossible in procurement process
• Vendors have limited possibilities to tailor
  services and equipment to user needs, so they
  offer universal solution (suboptimal for given
  case)
• Exceptions
• legal system in some countries is not so
  restrictive to research
• lower scale networks (e.g. Metropolitan) make
  implementation and operation by own staff

6
Development of network design

• Legacy design proces often gives suboptimal
  network services (costs, rate, flexibility)
  mainly transmission layer is poorly designed from
  technical and economical point of view (is the
  most hidden to user)
• In general, improvement possibilities were
  proved
• procurement of dark fibres instead services
• procurement of multivendor equipment plus
  integration
• Necessity of bad reliability and interoperability
  was not proved
• Multidomain operation is needed generaly and
  includes multivendor interoperability and
  multidomain monitoring and control
• Disadvantage design and operation are more
  complicated
• Solution procure support of design,
  implementation and operation, and save own
  ability to quick return and correct design
  (including retendering etc.)

7
New situation in network design

• Main source of innovation is optical technology
  now
• Fibres are the most important buiding element now
• Change of P2P services by P2P fibres was first
  step only
• New transmitters, receivers, amplifiers, gratings
  etc. are available, some of they even with
  MultiSource Agreements
• CEF Networking enables high availability and
  cost-efectivness of new services (e.g. Facilities
  based networking, E2E lightpath on demand, fibre
  switching)
• CEF approach starts in WAN or MAN in companies,
  corporates, military, municipalities and
  hospitals (etc.) new market segment is emerging,
  tasks of vendors are changing

8
CESNET2 fibre procurement

• Re-tendering since 2001 to 2005, goals
• decreasing costs, including single fibre lines
• shorter lines including first mile
• NIL approach
• two or more fibre pairs to network node
• reliability increase by physical diversity of
  first miles to node
• single provider of dark fibre line (spans could
  have different owners)
• first mile for international lines in Prague
• first mile for fibre connection of user premises
• cross border fibres (single provider)

9
(No Transcript)
10
(No Transcript)
11
(No Transcript)
12
(No Transcript)
13
(No Transcript)
14
(No Transcript)
15
DF Lines

• 3 500 km of fibre pair operational in CESNET2
• 350km of single fibre within CESNET2
• 660 km of fibre pair in CzechLight testbed
• More than 760 km of special fibres
  on the spools in laboratory

16
Example of lambdas and fibres costs(fibres are
blue, road distances of cities are from 20 to 210
km)
17
Cross border fibre (CBF, NoB)

• Viena-Bratislava-Brno-Ostrava-BialskoBiala
• Austria-Slovakia-Czech Rep-Poland
• Our presentation of experience with CBF on TF-NGN
  Cambridge, September 2003 (we call it NoB, Near
  over Border)
• SERENATE study in 2003 we request analysis
• Having NoB fibre connection, you could implement
  some lambdas for GEANT2 (or for Abilene in US.)
• Cost rule for fibres is KIS (keep it short).
• GN2 tender allowing fibres continues (we will use
  examples, not results)
• NoB is supported by SEEFIRE study

18
(No Transcript)
19
2 years after Analysis still needed

• SERENATE study, p.24, published April 22., 2003
• http//www.serenate.org/publications/d9-serenate.
  pdf
• Prices of equipment are relatively important for
  the economics of transmission systems, but the
  overall costs heavily depend on the network
  architecture and topology. The basic approach is
  to connect geographically neighbouring
  universities by fibres. NRENs should carefully
  analyse the pros and cons of any solution going
  beyond this, i.e. using an overlay fibre
  structure, because such a solution is usually
  more expensive. It would be feasible and cheap,
  albeit not simple, to independently cross
  national borders to build regional fibre
  networks.
• This situation requires a new and careful
  analysis of the architecture and topology of the
  transmission layer (typically a transmission
  structure overlaying the NRENs topology) at the
  European level.

20
Overlaying fibre lines when are needed??(lines
are examples only)
21
Facility based networking GLIF and CzechLight

• Global Lambda Integrated Facility (GLIF)
• Environment for co-operation
• networking, infrastructure, network engineering,
  system integration, middleware, applications
• GLIF was established by invited participants at
  the 3rd Global Lambda Grid Workshop, held August
  27, 2003 in Reykjavik, Island, www.glif.is
• Dark fibres are often used for Gigabit or 10
  Gigabit access to cities and university premises

22
Facility and production networking

• Facilility and production networking should be
  complementary
• Prague-Amsterdam 10 Gb/s lambda since September
  1, 2004 (Cisco-Lucent-Nortel)
• We are studying/testing possibilities of lambda
  connection to FermiLab, RAL, Karlsruhe, Taipei
  etc.
• We will use E2E lambdas prepared by
  GEANT2NRENMAN, if (when) available for such
  connections

23
GLIF World Map December 2004
24
CzechLight connections (to GLIF, to GN2 testbed
and to cz premises)
25
(No Transcript)
26
CzechLight intercity connections

• Prague Amsterdam lambda (GLIF facility), first
  mile in Prague on dark fibre, OC-192c
  transmission, grey 1550nm
• Prague Brno, dark fibre, 298.3 km, 66,5 dB,
  including 257.3 km of G.655 fibre, DWDM OC-192c
  transmission, NIL test
• Prague Plzen, dark fibre, 159.4 km, 36.7 dB, GE
  transmission by CzechLight EDFA 2in1

27
CzechLight connections in Prague region

• Connections of Particle physics workplaces
• - Institute of Physics of the Academy of
  Sciences of the Czech republic, dark fibre, GE
  transmission
• - Faculty of Mathematics and Physics of Charles
  University, dark fibre, GE transmission
• - Nuclear Physics Institute of the Academy of
  Sciences of the Czech Republic in Re, dark
  fibre, GE transmission
• - Faculty of Nuclear and Physical Engineering
  of Czech Technical University, lambda on PASNET
  dark fibres by CWDM, GE transmission
• We prepare connections for medical research

28
City dark fibre lines to the of CzechLight node
in Prague
29
CzechLight configuration

• Core node in Prague
• Cisco 15454 with 10 G DWDM SONET, 4xGE
• Cisco C3550 12G
• Cisco C6503 with 10 GE and GE
• Core node in Brno
• Cisco 15454 with 10 G DWDM SONET, 4xGE
• Cisco 6506 with 10 GE and GE prepared

30
Acknowledgement

• Jan Gruntorad for support
• Miroslav Karasek, Jan Radil and Josef Vojtech for
  advanced lighting of CESNET2 and CzechLight
  fibres
• Comment presented ideas and opinions are result
  of our ongoing RD activities and are opened to
  improvement