R&E Networks As Instruments for Discovery, Not Just Infrastructure

Greg Bell

May 28, 2012

This interview was originally published on the ALICE project's (América Latina Interconectada Con Europa) website.

Written by Tania Altamirano López

Q: What comes to your mind when you hear that a researcher is talking about collaboration?
Greg Bell: ESnet is a science network, dedicated to accelerating the process of discovery. Because modern science depends upon large-scale teams, we think about collaboration every day. When I hear about a new collaboration, I ask myself: ‘how could ESnet improve the productivity of that collaboration?’ For instance, we routinely work with scientists to facilitate the transfer of large data sets, assure their remote instrumentation can be accessed reliably, or consult regarding a new distributed data model. Each time we engage with a new collaboration, our team learns about new network requirements and emerging challenges that scientists are facing. In turn, this conversation helps us develop new services and best practices can be applied to other collaborations.

Q: What would you identify as the main importance of research and education networks?
GB: Innovation. Research and education networks are different from the commercial Internet: they face different challenges, offer different services, and are growing far faster. ESnet for instance is projected to carry over 100 Petabytes of science traffic per month by 2015 – a 72% increase per year. Our exponential growth results from the explosion in data generated at experimental facilities around the world. While experiments like the Large Hadron Collider have been known for creating massive data sets, this trend is emerging across nearly all science collaborations including those involved in climate, genomics, and materials discovery. New detectors being deployed at X-ray synchrotrons are generating data at unprecedented resolution and refresh rates, for example. The current generation of instruments can produce 300 or more megabytes per second, and the next generation will produce data volumes many times higher.
In order to respond to challenges posed by the worldwide data revolution, R&E networks are investing in partnerships to support applied research, development, and innovation activities with the aim of delivering new capabilities that work across multiple domains for Labs and universities around the globe.

How would you describe the role of ESnet at both a regional and a global level?
GB: ESnet has the mission of accelerating scientific discovery for projects funded by the US Department of Energy’s Office of Science, which supports 27,000 PhDs, graduate students, and engineers at 300 institutions (and claims credit for roughly 100 Nobel Prizes). Besides interconnecting the DOE’s National Laboratory system, its supercomputing sites and its experimental facilities, ESnet links the National Labs to 140 research and commercial networks around the world. We actively contribute to the community by sharing our best practices and innovations in joint forums like the GLIF (The Global Lambda Integrated Facility) collaboration as well as partner in developing new interoperable services, capabilities and standards that mutually benefit the science mission. Prime examples of our community contributions include development around inter-domain virtual circuit and performance measurement technologies as well as the associated standards development within the Open Grid Forum (OGF) and DICE collaborations. Within the U.S., we have a very close and complementary partnership with Internet2, which has also yielded many important collaborations - most recently around emerging 100 Gigabit Ethernet and software-defined networking, among others.

How important for ESnet is collaboration with other regional networks and in what ways do you collaborate at a global level?
GB: Modern science depends upon advanced R&E networks to connect scientists to each other and to research facilities, wherever in the world those might be located. One of the most well-known examples of such a facility is the Large Hadron Collider at CERN, which depends on high-performance networking to deliver data to thousands of researchers distributed globally. We expect that in the coming years, more and more facilities – in a range of disciplines – will adopt the same data model. As a result of this rapid paradigm shift, it is imperative that regional and national networks collaborate to deliver services that work seamlessly across multiple states, regions, countries and continents. Partnership is absolutely vital, because science discovery depends on it. To this end, ESnet collaborates with peer networks in many domains including network operations, engineering, software development, standards work, emerging services, and long-term strategy. We have been active in global collaborations that have led to the development of standardized, open-source tools for performance measurement, multi-domain virtual circuits, and other services and tools.

How do you think global collaboration among regional networks will change in the coming years?
GB: Without a doubt, we will need to step up our global collaborative efforts. Over the past several decades, we’ve worked as a community to successfully develop tools and services that are tailored to the multi-domain needs of global science. These tools are now moving from prototype deployments into hardened, production-ready services. Over the next decade, these services will continue to be adopted and refined, leading to an even more seamless and consistent end-to-end experience for our users. At the same time, I think it’s critical that we focus attention on outreach and education. As the data revolution continues to unfold, many scientists who have never used the network before will be forced to do so as their datasets grow too large to ship via portable media. Many of these scientists participate in small collaborations without the level of IT expertise of larger collaborations like the LHC. The R&E network community needs to come together and develop models and best practices that any researcher can easily adopt as part of their science workflow.

Could you describe your vision of R&E networks in the future?
GB: In the future, we’ll think of R&E networks as instruments for discovery, not just infrastructures. These instruments will be programmable, and they will offer a rich services interface to meet the needs of any collaboration. R&E networks will constantly communicate with each other over simple web-service interfaces, coordinating the lifecycle of service requests, brokering competing demands, and optimizing network services based on the specific requirements of individual workflows. They will do all this in a global context, while coping with a massive yearly growth in traffic. Networks will consume much less energy in general, and the energy they consume will be proportional to the work performed, which is currently not the case. ESnet has been a pioneer in the area of programmable networks, advanced capability development, and network energy efficiency - and we expect to continue in those roles for many years to come.