InTheLoop | 01.06.2014
NERSC Supercomputers Capture Turbulence in the Solar Wind
Scientists currently do not have the ability to accurately predict the severity of a space weather event or where it will have the most impact. But a team of researchers led by University of California, San Diego's (UCSD's) Homa Karimabadi is hoping to change that. Karimabadi and his colleagues managed to simulate all the scales of solar wind turbulence at once—for the first time ever. To make sense of this massive dataset, they tapped Lawrence Berkeley National Laboratory (Berkeley Lab) visualization specialist Burlen Loring, who developed custom analysis tools using supercomputers at the National Energy Research Scientific Computing Center (NERSC). Loring's work allows researchers to study turbulence in unprecedented detail, and the results may hold clues about some of the processes that lead to destructive space weather events. This work was published in Physics of Plasmas (PDF | 32MB). »Read more.
ESnet's Bell and Monga Featured Speakers at SDN Workshop
ESnet's Greg Bell and Inder Monga will speak in Tucson at a Jan. 7 workshop on Software Defined Networking (SDN) sponsored by the University of Arizona and University of New Mexico. Bell will deliver a keynote called "Seven New Year's Resolutions for Software Defined Networking," challenging the SDN community to think across network layers and organizational boundaries in addressing the needs of large-scale science collaborations. Bell will also participate in a panel discussion of the non-technical, human dimensions of SDN, a set of technologies that are changing the ways networking professionals will be trained. Monga will give the meeting's final presentation, summarizing results of a recent OSTP-sponsored, multi-agency SDN workshop in Washington which he organized and led. »Read more about the Arizona workshop.
Tierney, Zurawski of ESnet Co-Edit Journal Section on Network Measurement and Monitoring
In the November issue of IEEE Communications, ESnet's Brian Tierney and Jason Zurawski served as co-editors (with four others) of a special section on network measurement and monitoring. According to the authors, writing in a guest editorial, even though more and more services (including the cloud) are increasingly important for both the scientific and corporate worlds, network performance has not kept up. And the network problems are becoming more subtle and detrimental, while also being more challenging to troubleshoot across domains. Addressing the problem requires both sophisticated measurement and monitoring tools that are interoperable and also meet policy requirements.
To examine novel approaches and standardization efforts, the team of guest editors solicited research papers and received 18 submissions, of which eight were selected for publication in two issues of IEEE Communications, with four appearing in the November 2013 issue and four in May 2014. In addition to Tierney and Zurawski, the editors were Prasad Calyam, University of Missouri—Columbia; Constantine Dovrolis, Georgia Tech; Loki Joergenson, Lionsgate Technologies; and Raj Kettimuthu, Argonne National Laboratory and University of Chicago. »Read more.
ESnet Co-Leads Workshop on Developing Prototype SDN Network
About 100 networking experts from academia, industry, national labs and federal agencies met for a two-day workshop at the National Science foundation to plan a path forward to develop, deploy and operate a prototype SDN network. SDN, or Software Defined Networking, is an upcoming technology paradigm aimed at making it easier for software applications to automatically configure and control the various layers of the network to improve flexibility, predictability and reliability.
ESnet Chief Technologist Inder Monga was the lead organizer of the workshop and ESnet network engineer Erich Pouyoul gave a talk on science drivers for SDN. Monga also led a breakout session on "Technology and Operational Gap Analysis." ESnet has been a pioneer in developing and deploying SDN technology in support of data-intensive science for almost a decade, starting with research on virtual network circuits that eventually culminated in the facility's OSCARS project, recipient of a 2013 R&D100 award.
The invitation-only workshop was held at the National Science Foundation in Arlington, Va., and included speakers from the White House Office on Science and Technology Policy (OSTP), Google, DARPA, Internet2, SRI and Brocade, as well as ESnet. Among the areas covered were transparency and interoperation among SDN domains, security and identity management, and the participation of equipment vendors to advance technology transfer.
The workshop was organized after the OSTP directed federal agencies participating in the Networking and Information Technology Research and Development (NITRD) Subcommittee's Large Scale Networking (LSN) Coordinating Group to plan and hold an LSN workshop. The goal was to have participation by representatives from federal agencies, the commercial sector, researchers, and other networking and distributed systems research community participants to explore and report on the need for a prototype SDN network.
The workshop participants will draft a report documenting recommendations for needed R&D, resources and collaboration to deploy and operate the prototype SDN network and to identify future SDN research needs.
On the eve of the workshop, Federal Computer Week magazine published an article about federal agencies looking into SDN. Monga was among the sources interviewed for the article, which describes SDN as the next major architectural change looming for the IT community.
This Week's Computing Sciences Seminar
NERSC Many-Core Application Readiness Efforts and Experiences
Tuesday, January 7, 2014, 11:00am - 12:30pm, Bldg. 70A, Room 3377
Jack Deslippe, Chris Daley, Brian Austin, Harvey Wasserman & Hongzhang Shan, National Energy Research Scientific Computing Center (NERSC), Lawrence Berkeley National Laboratory
Ensuring that the NERSC workload can transition to future NERSC machines is a key ingredient in the success of the center. The application readiness team at NERSC has been evaluating application performance on potential future architectures. This discussion will focus on manycore performance in applications representing common user and third-party codes run at NERSC. In particular, we will discuss expected application performance on manycore architectures if codes remain unchanged; sources of good and bad performance; and planned steps to optimize applications. We've collected a lot of experience on good (and bad) practices for optimizing code and will share our "lessons learned."
Link of the Week: Open Science Grid Gives Opportunism a Good Name
The Open Science Grid (OSG) promotes more efficient use of smaller clusters at universities by sharing excess, "opportunistic" cycles. »Read more.