CAnet 4 UCLP Roadmap http:www.canarie.cacanet4uclpUCLP_Roadmap.doc

1
CAnet 4 UCLP Roadmaphttp//www.canarie.ca/canet4
/uclp/UCLP_Roadmap.doc

• Bill St. Arnaud
• CANARIE Inc www.canarie.ca
• Bill.st.arnaud_at_canarie.ca

2
Why do we need optical networks?

• There is urgent requirement to do fundamental
  rethink of the value of optical networks
• What constitutes an optical network?
• Do static fiber and wavelength connections
  constitute an optical networks?
• Or does an optical network necessarily mean it
  has to be dynamic and therefore need a optical
  control plane?
• Is there a genuine need for dynamic or agile
  optical networks?
• What is advantage of optical networks versus ATM,
  IP, etc
• More bandwidth? Lower cost? Guaranteed Quality
  of Service?
• Cut through for high bandwidth flows? Better
  sharing of limited resources?

3
What constitutes optical network research
testbeds?

• General agreement we need more research in layer
  0/1 technologies
• Improved fiber, better lasers and modulators,
  detectors, dispersion control, higher bit rates,
  better FEC, integrated optics, better high speed
  interfaces, optical repeaters (EDFA, RAMAN),
  C.L.S bands etc
• Optical cryptology, optical bit and symbol
  processing, and eventually optical computation
• Wavelength conversion
• Large questions on value of layer 2/3 optical
  network research testbeds
• Agile and dynamic optical networks
• Automatic restoral protection
• OOO verus OEO add/drop/switching
• Scheduling or signaling for bandwidth for large
  data flows
• Traffic engineering for long term flows

4
Dynamic optical networks - ATM reincarnated?

• Many of the same arguments for dynamic and agile
  optical networks were made for ATM networks
• Cut through routing/switching (remember Ipsilon?)
• Automatic restoral and protection
• QoS
• Signaling (and or scheduling) for high bandwidth
  channels
• We demonstrated with ATM that only one network
  layer should be dynamic otherwise you get
  conflicts
• Restoral and protection best done at layer 3 or
  higher
• Path selection and routing best done at layer 3
  or higher
• IP networks can handle huge data flows
• The only value of a somewhat dynamic transport
  network is to support traffic engineering of IP
  network and creation of VPNs hence MPLS-TE
• This is what we learned from ATM

5
Issues with scheduling and reservation of
bandwidth

• Call blocking !!! (nobody seems to mention this
  when discussing bandwidth reservation)
• Call blocking will always occur in any circuit
  oriented architecture
• Problem is minimal on telephone system where
  there are millions of circuits and lots of spare
  capacity
• Lots of research into call blocking probability
  Erlang formulae
• We have no research or body of evidence on
  probability of call blocking when users are
  trying to reserve huge chunks of bandwidth
• If call blocking exceeds a few percent it will be
  deemed a failure

6
CAnet testbed assumptions

• IP networking will remain the predominant
  technology
• Restoral, protection, routing will continue to
  done at IP layer
• Most high bandwidth applications required
  sustained on-going bandwidth to a small number of
  known locations
• Little need for reservation or scheduling of
  bandwidth
• There is little demand for infrequent high
  bandwidth transfers to random locations
• Best done through IP network
• Will look like a DOS attack if not properly
  planned for
• The value of optical networking is that it allows
  the creation of many fine grained IP networks
  for specific communities of interest rather than
  having a common generic IP network
• Adapts lessons learned from ATM and extends
  concept of traffic engineering to the user and to
  the topology

7
CAnet 4 design principles

• Occams rule of networking
• The simplest network is the best network
• Provide users with tools to do their own traffic
  engineering including changing topology and
  bandwidth
• Articulated Private Networks
• Allow users to create IP networks for their own
  community of interest
• Most importantly allow extension of network into
  campus to specific servers and bypass campus
  firewall
• Use Service Oriented Architecture (web services
  and workflow) to allow users to do their own
  provisioning and configuration of the network
• Also allows easy integration of application

8
UCLP Objectives

• Allow institutions to integrate wavelengths and
  fiber from different suppliers and integrate with
  institution's network management domain
• And offer VPNs (or APNs) to users
• Create discipline specific re-configurable IP
  networks
• Multihomed network which bypasses firewalls with
  direct connect to servers and routers
• User controlled traffic engineering
• Active replacement for Sockeye and Route Science
• Use lightpaths as traffic engineering underlay
  versus MPLS-TE overay
• Primary purpose is NOT reservation and leasing of
  wavelength resources
• Primary purpose is NOT switched optical networks
• Primary purpose is NOT end-to-end optical VPNs
• Primary purpose is NOT inter-domain connection of
  lightpaths

9
Todays hierarchical IP network
Other national networks
National or Pan-Nationl IP Network
NREN A
NREN C
NREN B
NREN D
University
10
Tomorrows peer to peer IP network
World
World
National DWDM Network
World
Child Lightpaths
NREN B
NREN A
NREN C
NREN D
Child Lightpaths
University
Server
11
Creation of application VPNs
Note Direct connection to server on campus
University
Dept
High Energy Physics Network
CERN
Commodity Internet
Research Network
University
University
Bio-informatics Network
University
University
eVLBI Network
12
(No Transcript)
13
UCLP intended for projects like Optiputer and
CaveWave
14
DRAC/UCLP Demo Network
Nortel DRAC
The Power of Web services
Canarie UCLP
Montreal
Ottawa
App
App
Halifax
Toronto
15
Recent UCLP examples

• Over 20 UCLP lightpaths setup across CAnet 4
• Need to purchase additional wavelength in 2005
• AARnet used UCLP to setup lightpath for
  Huygens-Cassini data transfer
• 5 HDTV streams to be switched and controlled
  through UCLP at APAN in Bangkok
• 3 UCLP lightpaths for restoral/protection by
  regional networks
• 2 UCLP lightpaths for distributed backplane
  e.g. mini TeraGrid
• 7 international UCLP lightpaths 1G to 2.5 G
• 10G UCLP lightpath shared between Tokyo Data
  reservoir and HEPnet

16
GENI initiatives vs UCLP

• Possible GENI testbeds
• Xbone http//www.isi.edu/xbone/
• Network Virtualization http//www.arl.wustl.edu/n
  etv/main.html
• Hypercast http//www.cs.virginia.edu/mngroup/hyp
  ercast/
• GENI initiatives mostly overlays on an IP
  network using various tunnels between routers
• UCLP is an underlay to IP network with
  lightpaths between switches substrate in
  Network Virtualization
• UCLP uses web services and work flow to create
  virtual networks and bind virtual (and real)
  routers/switches to the APN
• UCLP assigns control over all other network
  function to the user
• Multicast, QoS, application binding etc