Transcript No Slide Title - National e

The evolving network
Some Science drivers
Issues and Technology today
Future evolution
Peter Clarke
University College London
[email protected]
The Evolving Network
Peter Clarke
27-May-2004 NeSC Edinburgh
Radio Astronomy: Very Long Baseline Interferometry
The Evolving Network
Peter Clarke
27-May-2004 NeSC Edinburgh
TODAY
Onsala
Sweden
Westerbork
Netherlands
JIVE
Jodrell Bank
UK
Medicina
Italy
The Evolving Network
Effelsberg
Germany
TOMORROW
Torun
Poland
Peter Clarke
27-May-2004 NeSC Edinburgh
Particle Physics : The Large Hadron Collider
The CERN Accelerator
Complex
Proton-Proton
Collisions
~ 20 PetaBytes
per year
The Evolving Network
Peter Clarke
27-May-2004 NeSC Edinburgh
Breast Screening Programmes
Current system is
“minimal”
technologically
The Evolving Network
(taken from e-DiaMoND Project)
Peter Clarke
27-May-2004 NeSC Edinburgh
Future should be
digitised and
processed images
(taken from e-DiaMoND Project)
Remote
Radiographers ?
NETWORK
2nd opinion
Remote Patient
information
• Requires ~ Gbit/s flows for remote access
• Will not be possible without scheduled guaranteed net-services
The Evolving Network
Peter Clarke
27-May-2004 NeSC Edinburgh
Issues and technology today
Client
Campus
The
“network”
Typically
2.5 – 10
Gbit/s
User gets
10 -100
Mbit/s
Regional
Issues are:
•
TCP – the internet workhorse
•
Servers
– Disk I/O
– Network Cards
– Architecture
– OS
•
Local network and firewall
•
The mindset : im getting poor
response, the network must be
to blame – so I give up
SuperJANET
(10 Gbit/s
Regional
Campus
To 1st order the wide area network
is not the problem
Server
The Evolving Network
Peter Clarke
27-May-2004 NeSC Edinburgh
Example technology issue (1) : TCP (Transport Control Protocol )
• You use this every day for almost everything you do on the internet
• Never designed long distance high bandwidth networks
– but amazing that it does so well !!!
• Ordinary TCP cant survive packet
loss
• Plot show response to 1 packet loss
per million
Rate
Time
• “High Speed TCP” is designed for
today's internet and responds sensibly
to the same packet loss
[S.Floyd http://www.icir.org/floyd/hstcp.html]
Rate
Time
The Evolving Network
Peter Clarke
27-May-2004 NeSC Edinburgh
Example technology issue (2) : Disk Input/Output
• Great difference between low and high performance servers
• Care is needed also with interface to network
“Standard” servers easily
achieve ~ 100 Mbits/s
I/O
Rate
Time
“High performance ” servers
easily achieve
~ 500 Mbits/s
The Evolving Network
Peter Clarke
27-May-2004 NeSC Edinburgh
Example technology issue (3) : Differentiated Services
•
IP packets can be
“marked”
•
At each node (router)
in chain, marked
packets
can given “business
class” treatment
Identify & Classify
Sort
Dequeue
2.5 Gbit/s
background
Priority flow
Rate
The Evolving Network
Time
Peter Clarke
27-May-2004 NeSC Edinburgh
Messages:
 Getting 1 Gbit/s out of real networks is a known solution
(actually, getting 5 Gbit/s out of network has been done)
 Applications people and network people have to bring it
together to “make it work for science”
A 24 hour High Speed TCP test from Manchester to CCLRC-RAL
940 Mbit/s
Rate
Time
The Evolving Network
Peter Clarke
27-May-2004 NeSC Edinburgh
The Future Network ?
# of users
Picture from Cees DeLaat
University of Amsterdam
A
C
B
Requirements
A -> Lightweight users, browsing, mailing
B -> Business applications, multicast, streaming, VPN’s
C -> Special scientific applications, computing, data grids
The Evolving Network
Peter Clarke
27-May-2004 NeSC Edinburgh
Total BW
A
B
C
Requirements
A -> Need full Internet routing, many to many
B -> Need VPN services and some Internet routing, several to several
C -> Doesn’t need internet routing, Need very fat pipes, limited sites, few
to few
The Evolving Network
Peter Clarke
27-May-2004 NeSC Edinburgh
… a possible Network of the Future….
Router
Routed IP Network
Router
“layer-2” split out of
wavelength
[e.g. Ethernet]
Provides
• Normal best effort IP network where appropriate
• Extended “virtual LANs” where appropriate
• Switched lightpaths where appropriate
The Evolving Network
Peter Clarke
27-May-2004 NeSC Edinburgh
This new way of networking was the rationale for
UKLIGHT
- a National Facility for Advanced Optical Networking -
NorthernLi
ght
UKLight
SunLight
Czech
Light
CERN
Pacific NW
GigaPOP
MANLAN
The Evolving Network
Peter Clarke
27-May-2004 NeSC Edinburgh
Source: Kees Neggers, SURFnet
… in a bit more detail …
10 Gbit/s
New York
MANLAN
Stockholm
NorthernLight
IEEAF
10 Gbit/s
2.5 Gbit/s
10 Gbit/s
2.5 Gbit/s
10 Gbit/s
Tokyo
WIDE
CA*net4
10
Gbit/s
IEEAF
10 Gbit/s
Chicago
StarLight
10 Gbit/s
SURFnet
10 Gbit/s
NSF
10 Gbit/s
10
Gbit/s
Amsterdam
NetherLight
2.5 Gbit/s
SURFnet
10 Gbit/s
Tokyo
APAN
10 Gbit/s
London
UKLight
The Evolving Network
2.5 Gbit/s
Peter Clarke
Geneva
CERN
Prague
CzechLight
27-May-2004 NeSC Edinburgh
UKLIGHT is:
•
A VISION
– The UK was missing from an important the global stage
– The UK needed to be there – and now it is
•
SCIENCE AND NETWORKING ENABLEMENT
– Driven by a cross disciplinary scientific case
– Applications: HEP, Radio Astronomy, HPC, TeleMedicine…
– Network R&D: to find new paradigms for the next generation network
•
INFRASTRUCTURE SUPPORTED BY KEY BODIES
– Core e-Science Programme
– JISC-JCSR and JISC-JCN
– Funded through HEFC SRIF
•
A DRIVER for SuperJANET-5:
– The Global optical network initiative has been a driver which has
influenced the way the next generation networks will be built.
– The aim is more flexible – more cost efficient provisioning to meet needs
of diverse user requirements.
The Evolving Network
Peter Clarke
27-May-2004 NeSC Edinburgh
As a result of the initiative, UKERNA have pushed through a
complementary initiative to provide a nationwide domestic R&D
network in parallel with SupetJANET
This is a UK first !
We have never had a
pervasive R&D
network before
NNW
T
UHI Network
Clydenet
FaTMAN
T
C&NL
MAN
C
AbMAN
EaStMAN
T
Glasgow Edinburgh
S
NorMAN
T YHMAN
Potentially any
institute can
connect to the
R&D network,
and to access
UKLIGHT
through it
The Evolving Network
Warrington
Leeds
T
MidMAN
Reading
London
T EMMAN
T EastNet
Kentish MAN
TVN
Bristol
Portsmouth
South Wales S
S
MAN
SWERN
T
LeNSE
Northern
Ireland
T LMN
S
Peter Clarke
London
UKLight
27-May-2004 NeSC Edinburgh
Early Adopter projects already knocking at the door
– ESLEA : proof of benefit to:
• Radio Astronomy
• Particle physics through GridPP
• High performance computation through RealityGrid
• E-Health through e-DiaMoND
– White rose Grid
– High Performance Monitoring Project
–
…needs more…
The Evolving Network
Peter Clarke
27-May-2004 NeSC Edinburgh
Conclude with a success story in the field of
Computational Science
Reality Grid performs massive
simulations of complex molecular
dynamics
ReG won Most innovative Data
Intensive application award at
SC2003 conference & recently a
European ISC 2004 award
• A comprehensive interconnection
of HPC machines in UK and USA
• Connected by a “pre-UKLIGHT”
network
The Evolving Network
Peter Clarke
27-May-2004 NeSC Edinburgh
The Evolving Network
Peter Clarke
27-May-2004 NeSC Edinburgh