TransLight, a Major US Component of the GLIF
An Optical Web Connecting Research Networks in North America, Europe and the Pacific Rim
Tom DeFanti, University of Illinois, Chicago
Maxine Brown, University of Illinois, Chicago
Joe Mambretti, Northwestern University
John Silvester, University of Southern California
Ron Johnson, University of Washington
CTWatch Quarterly
May 2005

The US National Science Foundation (NSF) funds two complementary efforts through its International Research Connection Networks (IRNC) program — TransLight/StarLight and TransLight/Pacific Wave — that provide multi-gigabit links and supporting infrastructure to interconnect North American, European and Pacific Rim research & education networks, as well as to supplement available bandwidth that is provided by other countries.

TransLight/StarLight’s mission is to best serve established US/European production science, including support for scientists, engineers and educators who have persistent large-flow, real-time, and/or other advanced application requirements. Two OC-192 circuits are being implemented between the US and Europe. One circuit is a 10 Gb/s link that connects Internet2/Abilene and the pan-European G√ČANT2 via a routed network connection. The second circuit is a 10 Gb/s link that connects US hybrid networks, which can provide high performance, dedicated optical channels, such as the National LambdaRail (NLR), to similar European networks at NetherLight (configured as either one 10 Gb or eight 1 Gb switched circuits, or lambdas). Considerations related to security and measurement/monitoring will carefully be addressed under this award for both circuits.1

TransLight/Pacific Wave’s mission is the development of a distributed Open Exchange along the US west coast, from Seattle to San Diego, to interconnect North American, Asian, Australian and Mexican/South American links.2

Across North America, NLR, Canada’s CA*net4, and Internet2’s Abilene and Hybrid Optical and Packet Infrastructure (HOPI) projects connect the combined TransLights, from New York (Manhattan Landing, or MAN LAN), to Chicago (StarLight), to Seattle (Pacific Northwest GigaPoP). Pacific Wave carries the connection from Seattle down the US west coast to Los Angeles and on to San Diego and Tijuana (via CalREN - the California Research and Education Network, which is operated by CENIC - the Corporation for Educational Network Initiatives in California). These locations are the sites that support the vast majority of international connections to the US and form the fabric by which most international networks peer and exchange traffic with Abilene and the US Federal Research Networks. The TransLight team is the global community of people and groups who have most advanced the art, architecture, practice, and science of Open Exchange interconnectivity among high-performance networks. TransLight’s approach is based not just on backbone connectivity, but end-to-end connectivity and activism in advanced networking and applications, with a proven track record in attracting new technologies and stimulating collaborations, especially among leading domain scientists at end sites.

TransLight enables grid researchers and application developers to experiment with deterministic provisioning of dedicated circuits and then compare results with standard, aggregated “best-effort” Internet traffic. Multi-gigabit networks are referred to as “deterministic” networks, as they guarantee specific service attributes, such as bandwidth (for researchers who need to move large amounts of data), latency (to support real-time collaboration and visualization), and the time of usage (for those who need to schedule use of remote instrumentation or computers). Only through deployment of an integrated research and production infrastructure at network layers 1 through 3 will the various technical communities be able to address the major challenges of large-scale and complex systems research in peer-to-peer systems, Grids, collaboratories, peering, routing, network management, network monitoring, end-to-end QoS, adaptive and ad hoc networks, fault tolerance, high availability, and critical infrastructure to support advanced applications and Grids.3

The current monoculture of the Internet is already being replaced by a diversity of options for interconnecting researchers and educators, enabled by scalable wavelength technologies. By the year 2010, this trend will have substantially transformed networking, enabling multiple additional capabilities. Large-scale sensor nets and huge scientific instruments will generate extraordinary amounts of data. Cheap 1000-processor clusters will serve globally distributed science projects, interconnecting at tens of gigabits per second, working on computational problems, data-intensive applications, and visualization of massive datasets — if and only if there are sufficient, affordable and predictable networks by then. New research activities like the OptIPuter and new deployments like NLR in the US, as well as similar activities in other countries, and economically affordable trans-oceanic submarine capacity (up to 10Gb) are rapidly becoming essential components of the research and education landscape.

The ability to schedule and reserve lambda networks using advanced grid services is creating an advanced cyberinfrastructure, termed the LambdaGrid. A production-class, application-centric LambdaGrid, comprised of electronically and optically switched circuits and advanced grid services, is being built by teams of programmers, networking engineers, electrical/computer engineers, computer scientists and discipline scientists who are attacking the challenging research issues and helping develop innovative solutions.

The Global Lambda Integrated Facility (GLIF)4 is an international virtual organization that supports this decade’s most advanced data-intensive scientific research and middleware development for the LambdaGrid. GLIF participants include National Research Networks (NRNs), countries, consortia and institutions that have adequate bandwidth for research and education production traffic, and that also have additional capacity they are willing to make available for use by global teams of discipline scientists, computer scientists and engineers. The GLIF community shares a common vision of building a new grid-computing paradigm, in which the central architectural element is optical networks, not computers, to support this decade’s most demanding applications. To ensure the worldwide interoperability and interconnectivity of optical networks, GLIF has taken the lead in advanced facilities innovation and is developing architectural standards, or models, for open optical exchanges, which are being adopted by NRNs worldwide. The GLIF community is pioneering the concept of creating international, national, and regional distributed facilities, based on optical technologies, which departs from the traditional concept of a dedicated network that provides limited, non-deterministic services. For example, this new approach allows Grid applications to ride on dynamically configured networks based on optical wavelengths concurrent with normal Internet paths for the remaining traffic mix.

TransLight, and all the IRNC awardees, participate in GLIF and provide connectivity between multi-gigabit international networks and US/GLIF participants, including NLR, ESnet/UltraScience Net, and Internet2/HOPI.

Reaching Out to Broader Communities of Interest

TransLight’s goal is not only to make enough international bandwidth available to try a myriad of application-serving solutions, leveraging all nations’ and science, engineering and education needs, but also to empower access to a diversity of networking strategies. TransLight has proven that simply providing additional bandwidth with traditional networks, which provide only for a narrowly defined monolithic “best effort” service to all communities, is not a solution for long-term requirements. TransLight was the first advanced international infrastructure project that demonstrated the potential of agile optical networking in meeting the needs of these communities. TransLight has been developing powerful, sophisticated new techniques for matching specific community requirements with required infrastructure resources. TransLight continues to develop new techniques for precisely matching capability to each science and engineering research and education community served and, learning from the successes and failures of new models, continues to transfer best practices between these communities. Hybrid network services are desired by the international science community and are, in fact, required to advance science over the next decade. TransLight is using lambdas to advantage, using packets when expedient, and dedicated circuits when necessary.

International carriers have extraordinary overcapacity and they are actively seeking market development. According to an article in the Business section of the May 10, 2004 edition of the New York Times, “11 percent of available undersea bandwidth globally is being used.5” TransLight can lead the broadest science and engineering research and education communities to exploit this bandwidth to communicate and collaborate while demonstrating to the carrier community that market opportunities will emerge.

TransLight History

TransLight was an outgrowth of the NSF Euro-Link award, which funded high-performance connections between the US and Europe from 1999 through June 2005. TransLight, conceived in 2001 and started in 2003, was and continues to be a rational global network architecture that achieves great economy of scale and provides links to the largest communities of interest with the broadest services. In 2002, the goal was to give researchers who participated in iGrid 20026 as much international bandwidth as they could use. We engaged network managers, carriers, artists, network engineers, computer scientists, and domain scientists. We persisted at Supercomputing (SC) conferences in November 2002 and November 2003. Those who participated in these events published many journal and conference papers based on their results.7 By mid-2003, TransLight became the first persistent LambdaGrid.

In 2003, in partnership with SURFnet, NSF Euro-Link funds were used to purchase an OC-192 transatlantic circuit between Chicago and Amsterdam that provided both Layer 2 and Layer 3 connectivity. This hybrid network architecture provided the research and education community with both packet-switched (Layer 3) routed paths for many-to-many usage, as well as circuit-switched (Layer 2) lightpaths (or lambdas) for high-speed few-to-few usage.

TransLight quickly became a global partnership among institutions, organizations, consortia or country NRNs who wished to make additional bandwidth on their links available for scheduled, experimental use8. TransLight evolved into a two-year experiment to develop a governance model of how the US and international networking collaborators would work together to provide a common infrastructure in support of scientific research. In September 2004, as more and more countries began sharing bandwidth among one another, TransLight members dissolved the TransLight Governance body in favor of having the GLIF, an international body, assume this role.

GLIF partners are now organizing institutions of the iGrid 2005 event, to be held at the new Calit2 building in San Diego, September 26-30, 2005.9 Emphasis is on demonstrating applications research and middleware development that utilize new architectural approaches to next-generation Internet design and development using optical networking. Communities of interest will create their own private networks or share networks, creating on-demand LambdaGrids of interconnected, distributed computing, sensor and instrument resources that enable new infrastructures for advanced science.

NSF International Research Network Connections Program

The NSF IRNC Program funds international network links to connect US and foreign science and engineering communities, encourage the investigation and incorporation of advanced architectures needed to support the advanced and developing needs of science and engineering, encourage rational development and leveraging of deployed infrastructure to meet current and anticipated needs, and enable network engineers to engage in system and technology demonstrations and rigorous experimentation. The IRNC Program supports efforts to network North America to Europe, South America, the Pacific Rim, and to a global ring (via The Netherlands, Russia, Korea and China). It also supports network measurement and monitoring.

Relevant Websites
TransLight/StarLight receives major funding from National Science Foundation (NSF) International Research Network Connections (IRNC) program, award SCI-0441094 to the University of Illinois at Chicago (UIC), for the period February 2005 — January 2010; Tom DeFanti is principal investigator and Maxine Brown is co-principal investigator. Previous funding for Euro-Link, from the NSF High Performance International Internet Services (HPIIS) program, award SCI-9730202 to UIC, was for the period April 1999 – September 2005. SURFnet bv is a Key Institutional Partner of the TransLight/StarLight award, continuing an already long and successful partnership between the UIC, StarLight in Chicago, and NetherLight/SURFnet in Amsterdam.

TransLight/Pacific Wave receives major funding from NSF IRNC award SCI-0441119 to the University of Southern California (USC) for the period March 2005 – February 2010; John Silvester of USC is principal investigator and Ron Johnson, University of Washington, is co-principal investigator.

3 Aiken, R., Boroumand, J., Wolff, S. "Back to the Future," CACM, Volume 46, No. 1, January 2004.
5 Belson, K. "Technology; New Undersea Cable Projects Face Some Old Problems," New York Times Section C, Page 4, May 10, 2004.
6 and
7 DeFanti, T., Brown, M., De Laat, C., eds. "iGrid 2002: The International Virtual Laboratory," Journal of Future Generation Computer Systems (FGCS), Elsevier, Volume 19, No. 6, August 2003.
8 DeFanti, T., De Laat, C., Mambretti, J., Neggers, K., St. Arnaud, B. "TransLight: A Global Scale Lambda Grid for E-Science," Special Issue on "Blueprint for the Future of High Performance Networking," Communications of the ACM, Volume. 46, No. 11, November 2003, pp. 34-41.
9 and (coming soon)

URL to article: