The GEANT Perspectives

1
The GEANT Perspectives

• Fernando Liello
• NREN Policy Committee
• Chairman

2
Learn the Lesson

• GEANT is a success
• We have to plan for its development
• What want we to keep ?
• Where can we improve ?
• What would be better to change ?
• Which will be our next target ?

3
Europe as an Example

• More than 3100 Research Institutions
• Many Countries (currently 31)
• Different technological choices
• Different requirements in terms of service level
• Different funding and organizational schemes

4
GEANT Reach (in Europe)
5
The Key Elements

• A Layered Model
• Campus
• National
• International
• Competent, Dedicated Organizations
• The NREN Consortium
• DANTE
• Sound no-nonsense Model
• Strategy and Control at the NREN level

6
Closing a Loop

• Research Networking Started with Mission Oriented
  Networks
• Economy of Scale lead to National
  (Multidisciplinary) Research Networks
• Todays Developments can Revive Mission (or
  Experiment) Oriented Networks ?
• Is It only a Matter of Bandwidth ?

7
Infinite Bandwidth

• Can we afford to use ls as link unit ?
• We have to face the implementation lifecycle
• How long to have 40 Gbps on a l??????????????? ?
• How long to have 256 ls (or more) on one fiber ?
• When a more efficient scheme of SDH for
  multiplexing/de-multiplexing capacity will be
  commercially available ?

8
Private Networks

• The Solution for Big Science
• Require Complex Organization
• Huge Investment
• Does Big Science Really Exist ?
• Who makes the Choice ?
• The Chain Effect
• What happens if two Private Networks have a node
  in common ?

9
Virtual Private Networks

• Share the same physical infrastructure among many
  virtual ones
• Share of financial (and management) resources
• Allow Test and Production Grade services to
  coexist
• The Chain Problem is still there

10
Quality of Service

• You have good connectivity between site A and
  site B
• Your network has not guaranteed QoS
• You Know you Need some defined Level of Service
• You Introduce QoS in your Network

11
Quality of Service

• E-to-E QoS implies Coordination of (at least)
  three organizations the last mile problem
• The site A provider
• The site B provider
• The intermediate carrier
• Negotiate the needed level of Coordination (and
  how to implement it).

12
Virtual Temporary Connection

• Connect Computers to a Virtual Network on Request
• At the Required Bandwidth
• With Guaranteed QoS
• the next step beyond GRID
• and far beyond l allocation

13
An Empty Network

• Commercial Internet serves an huge number of
  relatively small capacity transactions
• ??Average occupancy can be high to achieve
  profitability
• Research Networks have a small number of large
  capacity users
• ??Average level of usage must be low to guarantee
  availability

14
The Strategy

• Expand the Capacity to keep the bandwidth
  available to the users
• Create a VPN service to fulfill the request of
  mission oriented and development communities
• Introduce automatic end-to-end bandwidth
  allocation and QoS

15
The Challenges

• Make the Structure more Hierarchical
• Regional Networks
• Encourage the creation of Pan-Continental
  networks outside Europe
• Global Connectivity
• Research is Global
• Build the Global layer
• Make Research Networks more Flexible

16
Some (incomplete) Answer

• The Federal Model looks good
• Better technical coordination is vital
• The role of organizations like DANTE and TERENA
  is vital (but NRENs are essential)
• NREN control is the only practical solution
• Policy, Strategic, Financial and Technical
  Decisions

17
Some Constraints

• The large Political and Financial Support, that
  research networking currently enjoys, brings
  Obligations as well
• We must evolve the Research Network, not to
  destroy it to build a new one flexibility must
  be the keyword
• To set the requirements too far ahead of time
  will be foolish

18
The Good Network

• The services must be visible and accessible
• The network you use to access them must be
  transparent hence invisible

19
Conclusions

• GEANT has proved its organization model
• Keep the physical network Federal but Unique and
  evolve it
• Implement technology and tools to allow the user
  to have a partitioned view of the network (if
  needed)
• Keep in mind the whole (global) picture