T

R

A

N

S

P

A

C

2

Networking in Support of Research and Education

The next ten years

James Williams

TransPAC2 Principal Investigator

Indiana University

Presented at the APII Workshop 2007

March 23, 2007

Seoul, Korea

T

R

A

N

S

P

A

C

2

Congratulations

• Congratulations on this very significant achievement and expansion of bandwidth between Japan and Korea.

• This more closely unites the two of the largest economies, science communities and networked populations in Asia.

• It provides a critical infrastructure linkage as a basis for future collaborations between Japan and Korea.

T

R

A

N

S

P

A

C

2

And Thank You

• I am honored to be asked to speak at this event.

• I am very pleased to return to Seoul, where I lived for a year in 1970.

T

R

A

N

S

P

A

C

2

Subject of my talk

• Speaking from our various roles as networking professionals, discipline scientists or administrators how will/must networking evolve in the next 10 years to better serve the needs of research and education in Asia and globally?

T

R

A

N

S

P

A

C

2

Topics to be covered

• Science drivers• Backbone networks and services• Exchange Points and services• Desktops• Static and switched connections

T

R

A

N

S

P

A

C

2

Terms to Define (1)

• Static connection: A network connection that is “set up” slowly, in minutes or longer and has a lifetime of hours or longer.

• Switchable connection: A network connection that is “set up” quickly, in minutes or less, and has a lifetime of hours or less.

• Backbone: A significant aggregation of network infrastructure that moves data between relatively geographically diverse locations.

• Exchange Point (EP): An aggregation of networks on a common network fabric that allows for easy establishment of interconnectivity between participants.

• End User Exchange Point (EUEP): An EP located at an end-user site such as a university or research laboratory.

T

R

A

N

S

P

A

C

2

Terms to Define (2)

• Network Transport: The type of signal and speed used to interconnect pieces of the network. Examples include Ethernet, SONET, Frame Relay, etc.

• Shared Network Infrastructure: Any network fabric that multiplexes multiple end-users‘ traffic together.

• Virtual circuits: Individual network channels through a shared infrastructure. Examples included MPLS tunnels on IP-based networks and VLANs on Ethernet-based networks

T

R

A

N

S

P

A

C

2

The R/E Applications Environment in 10 years

• HD penetration deeply into instruction• Grid computing evolution• Science instruments producing vast

amounts of data• Sensor nets and distributed experiments• New environments for network research• Healthcare and medical teleprescense• Distributed predictive applications

(earthquake/tsunami; bird flu, …)

T

R

A

N

S

P

A

C

2

Some Current Networking Components

• 10G backbone networks• 10G static network-network

connections• vLANs of many sizes• Exchange points (EPs)• Developing End User Exchange Points

(EUEPs)• Experiments with 1G switched

connections

T

R

A

N

S

P

A

C

2

Some possible future networking components

• Increased backbone speeds (next stop 100Gbps?) What about 40Gbps??

• What about undersea cable capacity and speed?

• Increased bandwidth devoted to static circuits• Multiple backbone networks using the same

underlying infrastructure.• Switchable circuits of varying sizes• More closely interconnected mesh of EPs,

EUEPs and science project drivers.

T

R

A

N

S

P

A

C

2

Backbones and backbone features

• Supply a production level framework for connection of regional networks

• Are a shared service and provide a mechanism for shared service delivery (measurement, security, troubleshooting …)

• Do not provide direct links between regional networks, international networks or experiments and end-users.

T

R

A

N

S

P

A

C

2

The Future - Backbones

• Backbones provide the maximum amount of “network control” as reflected in security, measurement and ability to troubleshoot.

• They will continue to be important as providers/deliverers of standard production level services, including commodity traffic.

• EP and EUEP connections will become more common, drawing traffic from traditional backbone routes.

T

R

A

N

S

P

A

C

2

Exchange Points (EPs and EUEPs)

• An aggregation of networks on a common network fabric that allows for easy establishment of interconnectivity between participants.

• Internationally EPs provide cross-boarder regional connectivity

• Allow network interconnectivity, but only at the EP or EUEP.

• An EUEP is located and managed by a university or research laboratory, usually associated with a large science project or instrument.

• Becoming more common and more interconnected.

T

R

A

N

S

P

A

C

2

Static Connections to EP and EUEP

• In service today for large (and well-funded) global research projects (LHC being a prime example) and between EPs (PacificWave – StarLight)

• Currently 10G and 1G connections seem useful and the right size

• As backbones expand and speeds increase, these will multiply

• These connections bypass backbone services such as security and measurement. And, are very difficult to troubleshoot.

• They may be dedicated to an application or a shared transport between two “interested parties”.

T

R

A

N

S

P

A

C

2

Switchable connections to EP and EUEP

• No evidence of demand for this today • Research underway to determine how

cross domain switchable connections might be engineered within international groups such as the GLIF and DICE

• Complex issue from an operational, technical and a financial perspective, perhaps too complex to be solved in the near future.

• This is a current research topic.

T

R

A

N

S

P

A

C

2

The Future – EPs and EUEPs

• The role of EPs and EUEPs will continue to increase unless there is a dramatic increase in bandwidth pricing.

• Internationally, EPs will be the standard connection mechanism due to the topology of undersea cable infrastructure.

• EUEPs will expand as research and education demand requires.

T

R

A

N

S

P

A

C

2

Desktops

• Individual workstations or groups of workstations managed by a research group and independent of central IT

• To reach a desktop you generally have to “punch through” central IT services and infrastructure…at least at present.

T

R

A

N

S

P

A

C

2

Switchable connections to the Desktop

• No evidence of demand for this• Little evidence of this being a useful

technology in the next 10 years.• Switchable connections to desktops will

require massive coordination and consensus from all aspects including operations, finance, measurement and security among network operators and equipment vendors.

• Sounds very much like ATM-SVCs…sigh

T

R

A

N

S

P

A

C

2

Static Connections to the Desktop

• Very much like switchable connections

• Little demand and many problems.• The static nature of these

connections may make them more useful and less complex.

T

R

A

N

S

P

A

C

2

The Future – Switched and Static Connections

• Static connections address a need not filled by current backbone networks. They will increase where financially prudent. Future research needs to be done on how to apply backbone services (security, measurement, troubleshooting techniques) to these connections.

• Switched connections present many difficult technical, operational and financial challenges. Research is ongoing. The outcome is uncertain.

T

R

A

N

S

P

A

C

2

vLANS and MPLS tunnels• Individual network channels through a shared

infrastructure

• Useful because they provide exclusive channels through the network that may not be subject to the same degree of scrutiny/delay as individual IP packets on a shared network; that, in turn, may have a speed up in delivery/performance, and reduce the possibility that packets are being inspected for security concerns.

• Not useful because they're still being used on a shared infrastructure (ethernet frames instead of IP packets) and because they may not be inspected for security concerns. Also not useful from the troubleshooting perspective. This is a continuing concern.

T

R

A

N

S

P

A

C

2

vLANS and MPLS tunnels

• Part of both the DICE and GLIF research.

• Research centering on automated mechanisms for creating cross-domain network paths connecting diverse equipment and different technologies (vLAN-MPLS interconnections as an example)

T

R

A

N

S

P

A

C

2

Network Based Video

• Medical teleprescense will drive demand for better end-to-end connections.

• Backbone networks seem to be able to handle most of the video demand today.

• The impact of IP-TV, especially in deeply networked countries like Korea and Japan is uncertain.

• Widely distributed network video will require better multicast support

• Widely distributed network video will require better networking at the end site.

T

R

A

N

S

P

A

C

2

Network Research

• Current efforts like PlanetLab use existing infrastructure and have limited infrastructure control.

• Future efforts like GENI will extend experimental control deeper into the infrastructure.

• Control plane research efforts such as those undertaking within DICE and GLIF will be critical to the future of switched services.

T

R

A

N

S

P

A

C

2

Summary

• I see limited interest in any type of connection (static or switched) to the desktop, now or in the immediate future.

• Backbone networks will continue to be important within governmental (and policy) domains of control.

• Dedicated static connections to an EP or EUEP are in place today and useful, but bypass measurement and security infrastructure.

• As these issues are addressed, dedicated static connections will become more widely used, where cost effective.

• Switched connections to an EP or EUEP will be difficult to engineer, finance and manage. They may prove to be useful or perhaps not…

• Ongoing network research efforts in control plane development will be critical in the effective deployment of switched services.

• vLANS and MPLS tunnels are in use today. Further research may provide more automated mechanisms for creation and linkage.

T

R

A

N

S

P

A

C

2

And, finally…

• Thank you again for this invitation and your attention

• I will be happy to answer any questions either now or at any time during the workshop

James Williams

Indiana University

williams@indiana.edu

T

R

A

N

S

P

A

C

2

URLs (1)

• The TransPAC2 Project: www.transpac2.net• The US National Science Foundation International Research Network

Connections Program: www.irnclinks.net• The APII Project: http://www.tc.apii.net/• Science facility – LHC: http://lhc.web.cern.ch/lhc/• Science instruments

– eVLBI: http://www.evlbi.org/– IceCube: http://icecube.wisc.edu/

• Sensor networks– NEON: http://www.neoninc.org/– GEON: http://www.geongrid.org/– NEES: http://www.nees.org/– ORION: http://www.orionprogram.org/

T

R

A

N

S

P

A

C

2

URLs (2)

• Grids– TeraGird: http://www.teragrid.org/– OSG; http://www.opensciencegrid.org/

• Network research– GENI: http://www.geni.net/– PlanetLab: http://www.planet-lab.org/

• Switchable connection projects– GLIF (Global Lambda Integrated Facility) http://www.glif.is/– DICE ( DANTE, Internet2, CANARIE, ESnet) collaboration

http://www.geant2.net/server/show/nav.1227