Transcript - NORDUnet

NORDUnet 25 years
The Future of NORDUnet
Hans Wallberg
SUNET
University of Umeå
[email protected]
Predicting the Future?

Study the history and extrapolate!
24 May 2005 - #2
The Mission of NORDUnet



Provide international network
connectivity to Research networks and
the General Internet
Network development activities
International representation of the
Nordic networks
24 May 2005 - #3
Network Capacity



56 kbit/s in 1988
10 Gbit/s today
10 Tbit/s in 2015 ~10 Mbit/s per user
10 000
9 000
8 000
7 000
6 000
5 000
4 000
3 000
2 000
1 000
0
1988
1990
1995
2000
2005
2010
2015
24 May 2005 - #4
The NORDUnet Plug of 1989
X.25
DECnet
TCP/IP
24 May 2005 - #5
The NORDUnet Plug of Today


IP – 10 Gbit/s
NorternLight
 n x GE

Provisioned as completely separate
connections
24 May 2005 - #6
The NORDUnet Plug of Tomorrow


IP ≥ 10 Gbit/s
NorthernLight
 n x GE
 n x 10 GE
 40 Gbit/s or 100 GE?


Distributed over one hybrid network based
on dark fibre
Will the national networks and campus
networks be prepared to deliver lambdas
to the individual researcher?
24 May 2005 - #7
Technical Issues


Connection oriented or connectionless
Issues the first NORDUnet conference
 X.25 and the Q-bit
 Cambridge Rings vs Ethernet



In the beginning we were connections
oriented and used X.25
From 1988 we became connectionless with
packet switching and IP
ATM was an exception and a mistake
24 May 2005 - #8
Connectionless or Connection Oriented

Today
 Connectionless (packet switched)
for general communications.
Millions of flows per second.
 Connection oriented (lambdas)
for special needs. Point-to-point
between researchers. Few flows
with long duration.
24 May 2005 - #9
Connectionless or Connection Oriented

Tomorrow
 Still connectionless for the general
use
 Many more lambdas (connection
oriented) for special needs
 Much higher bandwidth per lambda
 Will the connection oriented paradigm
scale?
 The number of point-to-point links increases
 The lifetime of the links decreases
24 May 2005 - #10
Technical Developments

DWDM systems
 Transparent
 Passive core




Automatic attenuation and dispersion compensation
ROADMs
Passive splitting instead of OADMs
All active optical equipment at customer premises
 Increased number of channels (lambdas)
 50 GHz per channel (25 GHz per ch will come?)
 Both C-band and L-band in the same system
 Higher bandwidth per channel
 Lightwave modulation: Duo Binary, ODB, DPSK, etc.

Optical Switches – Photonic Switching
 Bit-rate and protocol transparent
24 May 2005 - #11
The Reasons for Hybrid
Networks








Routers are too expensive
Switches are cheap
Optical Switches are even cheaper
Eliminates the bandwidth limitations of
the campus networks
Eliminates limitations of campus firewalls
The only way to achieve e2e performance
We shouldn’t use expensive routers for
bandwidth intensive long-lived flows
Routers cause latency
24 May 2005 - #12
Killer Applications?

Unlimited bandwidth needs in 5 years:










Astronomy - VLBI
Physics
Space Physics - Sensors
Life Sciences - Imaging
Remote Visualization
Grid Computing
Will they really show up?
Will they be financed and able to pay their marginal part of
the communication cost?
When will we see Grid Clusters with 10GE Interfaces?
If these killer apps don’t show up the networking community
will loose credibility
24 May 2005 - #13
Conclusions

The future network of NORDUnet should:
 Use dark fibre
 Use the latest generation of
DWDM equipment




Future proof
Flexible and adaptable to any needs
Compatible with the national networks
Router consolidation
24 May 2005 - #14
Network development activities





Participate in international network development
activities (GLIF, GN2, etc.)
Coordinate common Nordic network development
activities (Nordunet3, etc.)
Participate in network testing and experiments
for specific researchers (VLBI, etc.)
Most of the network development activities will
be based on a common Nordic hybrid network
Challenges:
 Participation from the Nordic national
network
 Avoid interfering with the national
networks
24 May 2005 - #15
International representation of
the Nordic networks

One person instead of five
attending international meetings
 Saves money and resources


One Nordic voice
One common Nordic view
 Makes an impact
24 May 2005 - #16
Final Conclusion

In the future NORDUnet should:
 Provide international network
connectivity for the Nordic
NRENs
 Perform network development
activities
 Represent the Nordic networks
Internationally
24 May 2005 - #17
NORDUnet 2006 Networking Conference
September
Göteborg Sweden
Göteborg University
24 May 2005 - #18
Chalmers University of Technology