TeraGrid ’09: OSG and TeraGrid Collaboration
Paul Avery, a recognized leader in advanced grid and networking for science, delivered the first keynote address at the recent TeraGrid ’09 conference in Arlington, Va. A professor of physics at the University of Florida, Avery is co-principal investigator and founding member of the Open Science Grid (OSG). Avery talked about the history of OSG, some of the projects that leverage its resources, and OSG’s relationship with TeraGrid.
“The OSG is a multicultural consortium — half technology and half sociology,” said Avery.
While the majority of OSG users are from American institutions, 2,500 from 60 countries accessed OSG resources in the past year. From the current rate of usage, it is estimated that researchers will have used tens of petabytes of storage by 2012 and one hundred by 2013. By engaging with the Large Hadron Collider (LHC) at the European Organization for Nuclear Research, also known as CERN, researchers could need an exabyte of storage by the end of the next decade. Currently there are 5,000 physicists from hundreds of institutes who conduct research at CERN.
The OSG consortium includes science and research communities, national labs, international infrastructures, multidisciplinary (virtual) organizations (VO), technologists, HPC professionals, LHC/CERN experiments, regional and campus infrastructures, education communities as well as computer science and DOE-funded SciDAC projects.
OSG has a distributed facility of 85 resources with approximately 50,000 CPU cores of processing power and about 10 petabytes of disk storage. Its milestones and deliverables are a reflection of its science-driven charter. The leadership of OSG hails jointly from American national laboratory and university involvement where most science is conducted and collaborations are formed. OSG’s leadership is collaboratively focused with international and multidisciplinary interactions. The major stakeholders include the ATLAS and CMS experiments at the LHC and the Laser Interferometer Gravitational Wave Observatory (LIGO).
Jointly funded and overseen by the National Science Foundation (NSF) and the US Department of Energy (DOE), the 2006-2011 OSG project was funded with $30M — split roughly equally between NSF and DOE. There are 35 full time employees who are directly funded by the OSG project while many more are leveraged via OSG’s relationship with consortium members.
Avery offered a brief history of how OSG formed, starting about a decade ago with several NSF- and DOE-funded grid projects. The projects joined forces and in a bottom-up process began deploying test beds and a prototype grid that grew in size and complexity as resources were added from universities, campus grids and national laboratories. At the same time strong links were forged with CERN and the European grid projects to jointly form the Worldwide LHC Computing Grid (WLCG). He spoke of OSG’s developing partnerships with regional grids such as New York State Grid and SuraGrid and its collaborations with optical network organizations that provide critical links for high speed data movement.
OSG’s campus engagement model was developed by the Renaissance Computing Institute (RENCI) in North Carolina. By working with campus Cyberinfrastructure (CI) Days and through academic outreach organizations like EDUCAUSE, OSG is communicating its value to colleges and universities. “It seems to be working. Usage stats have doubled since 2008,” said Avery.
“Collaboration between TeraGrid and OSG is an integral component to achieving the nation’s vision for CI,” said Avery as he began to talk about the future of OSG. This vision was articulated in a “dear colleague” letter from the NSF Office of Cyberinfrastructure on June 9, 2009.
“As an overarching theme, OCI will promote the development of collaborative computational science — defined broadly to encompass research and development of comprehensive CI in all areas described below, as well as the application of CI to solve complex problems in science and engineering — as one of OCI’s primary missions. OCI will work to provide stewardship for computational science at NSF, in strong collaborations with other offices, directorates, and agencies.”
Explosive growth in processing and storage reflects the increasing sophistication and demands of 21st century research and engineering enterprises. However, more effective middleware and management tools are required for applications to scale so that they can efficiently use rapidly growing OSG and TeraGrid resources. OSG is moving to a bridging model to connect heterogeneous resources to one another and to the communities that utilize them in new and interesting ways. These include large digital libraries and public databases, computing clouds, massive distributed datasets, GPUs, visualization, instrument steering, collaborative environments, etc.
Avery noted that a closer OSG and TeraGrid relationship builds on several existing joint activities and substantial overlap in resources and services. Common interests such as communication, campus engagement, training, student workforce development, software licensing, security, science gateways, cloud development, virtualization and administrative tools offer additional opportunities for working together. Another mutually-beneficial goal is the development of a common software stack.
Many are aware of the fundamental differences between OSG and TeraGrid as they originated from distinctly different cultures. TeraGrid sprang from the NSF-funded PACI partnerships, while OSG grew out of a collaborative physics environment. OSG currently doesn’t have an allocation process, though future demand may cause it to examine TeraGrid’s mechanism. Both TeraGrid and OSG recognize the importance of monitoring trends to help them develop systems that will adapt to changing conditions and persist over time.
Even though OSG and TeraGrid have both contributed to the national CI which serves multiple communities, challenges remain in accommodating the expected growth and heterogeneity of the “third leg of science.” According to Avery, “A long-term OSG and TeraGrid partnership offers outstanding opportunities to advance a coherent national CI that is persistent over decades, is international in outlook, encourages federal agencies to adopt consistent strategies, and trains the workforce that will utilize it and continue its development.”
Avery’s presentation is available at http://www.teragrid.org/tg09/files/avery_teragrid.pdf.