InTheLoop | 08.12.2013
Aug. 19 Town Hall Meeting on 2014 Office of Science Early Career Research Program Awards
For those who are interested in competing for the 2014 Office of Science Early Career Research Program awards, please be reminded that the Lab Director will host a Town Hall Meeting from 10 – 11:30 a.m. Monday, August 19, in the Bldg. 50 Auditorium.
The program supports the development of individual research programs of outstanding scientists early in their careers. Mandatory pre-proposals will be accepted until Thursday, Sept. 5 for research ideas that foster Berkeley Lab programs funded by the Office of Science. Visit the Lab’s program website for more on the proposal process. CS employees thinking about applying for the award and haven't yet done so should inform Esmond Ng (and your group lead) of your intention by COB on Monday, August 12.
New Security Procedures at OSF – Must Show Lab Badge When Entering
Effective Tuesday, August 13, when coming to OSF, be prepared to show your badge to the security guard. While at OSF, it’s suggested that you wear your badge so that it is visible; however, Security will only need to see it when you enter the building. Signs will be posted at the entrances to remind you.
Regarding visitors and unbadged personnel, please remember the following:
- Visitors must check in with security at the front desk and sign the visitor's log.
- If you are hosting visitors, please provide the names to the administrative staff.
- Hosts must escort guests to and from the security desk. Alternatively, hosts may request visitors be sent to the second floor admin area after check-in, but the host must meet and escort the visitor beyond that area.
LBNL staff without badges must be hosted by badged employees, usually their immediate supervisors. Staff who work primarily at OSF may check out a loaner badge at the security desk once a badged employee has agreed to host them for the day.
ESnet, Globus Online to Host Free Aug. 22 Webinar on Managing Scientific Data
ESnet, DOE’s high-speed network, and Globus Online invite scientists and IT staff to join a free online session on cutting-edge data transfer and best practices for managing data. The webinar will be held at 2 p.m. Thursday, Aug. 22. The session will include presentations by two scientists who have used these technologies to accelerate their science workflow and discovery. Globus Online is used by both ESnet and NERSC and provides easy-to-use services and tools for research data management.
Applications for CS’ Alvarez Fellowship Now Being Accepted
The Computational Research Division, the National Energy Research Scientific Computing Center (NERSC) and the Energy Sciences Network (ESnet) are now accepting applications for the Luis W. Alvarez Fellowship in Computational Science. The fellowship supports recent graduates with a Ph.D. (or equivalent) with a strong emphasis on computing or computational science. The successful applicant will be compensated with a competitive salary and excellent benefits. Applications are due November 26, 2013, for Fall 2014.
Past Alvarez fellows among CS staff include: Lavanya Ramakrishnan, Rollin Thomas and Ayd?n Buluç. Current Alvarez Fellows are Alexander Kemper, Lin Lin, Robert Saye and Didem Unat.
This Week’s Computing Sciences Seminars
Toward Smart-tuned Krylov-based Methods for Future Extreme Computing
Thursday, August 15, 11 a.m. – 12 noon, Bldg. 50F, Room 1647
Serge G. Petiton, Maison de la Simulation/CNRS and University Lille 1, Sciences et Technologies
Exascale hypercomputers are expected to have highly hierarchical architectures with nodes composed of lot-of-core processors and accelerators. The different programming levels (from clusters of processors loosely connected to tightly connected lot-of-core processors and/or accelerators) will generate new difficult algorithm issues. New methods should be defined and evaluated with respect to modern state-of-the-art of applied mathematics and scientific methods.
Krylov linear methods such as GMRES and ERAM are now heavily used with success in various domains and industries despite their complexity. Their convergence and speed greatly depends on the hardware used and on the choice of the Krylov subspace size and other parameters which are difficult to determine efficiently in advance. Moreover, hybrid Krylov methods would allow reducing the communications along all the cores, limiting the reduction only through subsets of these cores. Added to their numerical behaviours and their fault tolerance properties, these methods are interesting candidates for exascale/extreme matrix computing. Avoiding communication strategies may also be developed for each of the instance of these methods, generating complex methods but with high potential efficiencies. These methods have a lot of correlated parameter which may be optimized using auto/smart-tuning strategies to accelerate convergence, minimize storage space, data movements, and energy consumption.
In this talk, we first will present some basic matrix operations utilized on Krylov methods on clusters of accelerators, with respect to a few chosen sparse compressed formats. We will discuss some recent experiments on a cluster of accelerators concerning comparison between orthogonal, incompletely orthogonal and non-orthogonal Krylov Basis computing. Then, we will discuss some results obtained on a cluster of accelerators to compute eigenvalues using the MERAM method with respect to the restarting strategies. We will survey some auto/smarttunning strategies we proposed and evaluated for some of the Krylov method parameters. As a conclusion, we will propose auto-tuning strategies for future hybrid methods on post-petascale computers, on the road to exascale hybrid methods.
Parallel and Distributed Application Paradigms
Parallel and Distributed Application Paradigms
Daniel S. Katz, University of Chicago & Argonne National Laboratory
Friday, August 16, 2013m, 10:00am - 11:30amm, Bldg. 50B, Room 4205
This talk will use Montage (http://montage.ipac.caltech.edu/), an astronomical image mosaicking application that is a toolbox of independent components, to explore various application paradigms on parallel and distributed systems, as the Montage components can be used in a variety of settings, including on a single system, on a parallel system, or on a set of distributed systems, including grids and clouds. Montage, which was built to use MPI in parallel, and Pegasus/DAGman on distributed systems, has also been used as an exemplar many-task computing (or workflow) application by a number of other tool and system developers. In this talk, a variety of work with Montage will be discussed, including the use of multiple types of infrastructure/middleware, the use of scripting to allow a user to easily customize their use of the Montage components, and overcoming data management issues.
Link of the Week: Moore's Largesse to Benefit Basic Physics Research
Gordon Moore, the former CEO or Intel who is known for his oft-cited "law" that the number of transistors on an integrated circuit roughly doubles every two years, is planning to spend $90 million over five years to support basic research in physics. Specifically, the money will fund research in condensed matter physics, pushing to enhance understanding of a class of systems called quantum materials. The funding from the Gordon and Betty Moore Foundation comes at a time when basic research budgets are shrinking. »Read more.