InTheLoop | 07.19.2010
July 19, 2010
LBNL Contributes Expertise to New Amazon Web Services Offering
Amazon Web Services (AWS) has launched Cluster Compute Instances for Amazon EC2, which makes high-bandwidth, low-latency high performance computing (HPC) resources available in a cloud-computing environment. To ensure that the new Amazon EC2 service will be able to handle a gamut of demanding HPC applications ranging from electronic design automation to financial services, Amazon Web Services worked closely with Berkeley Lab researchers Keith Jackson, Lavanya Ramakrishnan, Shane Canon, and others. Read more.
Berkeley Lab Hosts Weeklong Workshop on Cloud Computing
The forecast for many aspects of computing definitely calls for clouds, but cutting through the hype and getting a clear picture on the possibilities and limitations of cloud computing requires intensive hands-on sessions. To this end, Berkeley Lab’s Computational Research Division (CRD) is hosting a weeklong workshop on cloud computing this week, July 19-23.
The Center for Information Technology Research in the Interest of Society (CITRIS) at UC Berkeley is helping host the workshop. Participating organizations helping Berkeley Lab provide content include UC Berkeley, UC Davis, Yahoo!, Amazon, Eucalyptus Systems, the University of Virginia, and Microsoft. The Department of Energy’s National Energy Research Scientific Computing Center (NERSC) contributed both workshop content and computing resources. Participation was limited to 45, and the limit was quickly met.
“There is a lot of information floating around on cloud computing, but it’s not easy to know where to start,” said organizer Deb Agarwal, head of CRD’s Advanced Computing for Science Department. “Fortunately, we have a number of research projects, testbeds and collaborations in the cloud, so we were in a good position to marshal the necessary resources for this workshop.”
Among the topics to be covered during the workshop are an overview of cloud computing, software and systems training, data management and analysis, and experiences in using research and commercial cloud services.
The Role of Exascale Computing in Energy Security
How will the United States satisfy energy demand in a tightening global energy marketplace while, at the same time, reducing greenhouse gas emissions? Exascale computing — expected to be available within the next eight to ten years — may play a crucial role in answering that question by enabling a paradigm shift from test-based to science-based design and engineering.
An expanded version of the “Exascale for Energy” article published earlier this year in SciDAC Review is now available. It includes a new section on the electric power grid and gives additional examples of energy-related computational research at Berkeley Lab. Read more.
The Green Corollary to Moore’s Law
There’s a little-known but very important parallel movement to Moore’s Law recently elucidated by Jonathan Koomey, an energy specialist at Stanford and Berkeley Lab’s Environmental Energy and Technologies Division. Koomey went back into the historical record and found that the electrical efficiency of computing has been increasing at the roughly the same pace as computing speed. Let’s call it Koomey’s Corollary to Moore’s Law. Read more.
Berkeley Lab Staff Contribute to SIAM Annual Meeting
The Society for Industrial and Applied Mathematics (SIAM) held its annual meeting last week, July 12–16, in Pittsburgh, PA. As usual, Berkeley Lab staff were well represented among the speakers. Their contributions are listed below. (All staff are members of CRD unless otherwise noted.)
- David Bailey and Juan Meza co-authored “A Machine Learning Approach to Intrusion Detection for High Performance Computing” and co-organized (with Ali Pinar of Sandia) a session on “Mathematics of Complex Distributed Interconnected Systems.”
- Phil Colella co-authored “Local Mesh Refinement for Plasma Physics” with Bei Wang and Greg Miller of UC Davis.
- Inez Fung of Earth Sciences gave an invited talk on “Mathematical Challenges in Climate Change Science”
- Bakytzhan Kallemov, David Trebotich, and Greg Miller co-authored “A Multi-Scale Model for Polymer-Laden Flows Coupling Stochastic Particle Dynamics with Continuum Fluid Mechanics.”
- Xiaoye Sherry Li gave an invited talk on “Factorization-Based Sparse Solvers and Preconditioners.” She also organized a session on “Architecture-Aware Algorithms.”
- Kamesh Madduri presented “Hybrid Parallel Programming for Massive Graph Analysis.”
- Brian Van Straalen presented “An Embarrassingly Parallel Benchmark to Study Architecture Heterogeneity.”
At the SIAM Conference on the Life Sciences, held in conjunction with the general meeting, Kirsten Fagnan of NERSC presented “Computational Modeling of Extracorporeal Shock Wave Therapy”; and Jonathan Tang of Life Sciences presented “Identifying the Rules of Engagement Enabling Leukocyte Rolling and Adhesion using a Synthetic Model.”
More Conference Presentations by NERSC and CRD Staff
Katie Antypas of NERSC will give a tutorial course on “Parallelization and Parallel I/O” tomorrow through Friday at the Astrosim Summer School on Computational Astrophysics, being held July 11–25 at the Nicolaus Copernicus University in Toruń, Poland. This year’s theme is “Simulating the Universe: Towards petascale supercomputing in astrophysics.”
Daniela Ushizima of CRD and NERSC will give a talk on “spmR: An R package for fMRI Data Analysis Based on the SPM Algorithms” at useR! 2010, the R user conference being held at the Gaithersburg, Maryland, campus of the National Institute of Standards and Technology (NIST) July 20-23. R is a programming language and software environment for statistical computing and data analysis.
Early Career Research Program Town Hall Meeting on Friday
The Department of Energy’s Office of Science has announced the FY 2011 Early Career Research Program. Now in its second year, the program supports the development of individual research programs of outstanding scientists early in their careers. Pre-proposals are now being accepted in Advanced Scientific Computing Research and five other fields. It is anticipated that up to $10 million will be available for 15-35 awards in FY 2011. Mandatory pre-applications are due August 13. Director Paul Alivisatos will host an informational meeting on Friday, July 23, 11 a.m. to noon. Details on eligibility, program rules, and applying are available here, or visit the Lab’s informational website.
Access to Yahoo! Cloud Cluster Is Available
The Berkeley CSE/CITRIS Cloud Computing Initiative will continue to be granted access to Yahoo!’s cloud computing cluster M45 in the upcoming academic year for systems software and applications research. Students, faculty, and staff scientists are welcome to apply for time on the cluster; the application deadline is August 6. Go here for more information, or contact Masoud Nikravesh.
Par Lab Boot Camp Scheduled for August 16–18
The 2010 Pab Lab Boot Camp—Short Course on Parallel Programming is intended to offer programmers a practical introduction to parallel programming techniques and tools on current parallel computers, emphasizing multicore and manycore computers. Three all-day sessions will be held at Wheeler Auditorium on the UC Berkeley campus, as well as online. The course is free to students, faculty, staff, and affiliates (including Berkeley Lab), but registration is required. Go here for details.
This Week’s Computing Sciences Seminars
Maintainable and Reusable Scientific Software Adaptation
Thursday, July 22, 11:00 am–12:00 pm, 50B-2222
Pilsung Kang, Virginia Tech
Scientific software must be adapted for different execution environments, problem sets, and available resources to ensure its efficiency and reliability. Although adaptation patterns can be found in a sizable percentage of recent scientific applications, the traditional scientific software stack lacks the adequate adaptation abstractions and tools. As a result, scientific programmers manually implement ad-hoc solutions that are hard to maintain and reuse. In this talk, I present a novel approach to adapting scientific software written in Fortran. Our approach applies a C++ Aspect-Oriented Programming (AOP) extension to Fortran programs. Specifically, our approach expresses the adaptability functionality as abstract aspects that implement known adaptation patterns and can be reused across multiple scientific applications. Application-specific code is systematically expressed through inheritance. The resulting adaptability functionality can be maintained by any programmer familiar with AOP, which has become a staple of modern software development. We validated the expressive power of our approach by refactoring the hand-coded adaptability functionality of a real-world computational fluid dynamics application suite. The refactored code expresses the adaptability functionality in 27% fewer lines of code on average by removing duplication and leveraging aspect inheritance.
Algorithmic and Numerical Techniques for Atomistic Modeling
Friday, July 23, 11:00 am–12:00 pm, 50F-1647
Hasan Metin Aktulga, Purdue University
Molecular dynamics (MD) modeling has provided a powerful tool for simulating and understanding diverse systems — ranging from materials processes to biophysical phenomena. Many instances of this class of methods rely on a static bond structure for molecules, rendering them infeasible for reactive systems. Recent work on reactive force fields has resulted in the development of ReaxFF, a novel bond order potential that bridges quantum-scale and classical MD approaches by explicitly modeling bond activity (reactions) and charge equilibration. These aspects of ReaxFF pose significant challenges from a computational standpoint, both in sequential and parallel contexts. Evolving bond structure requires efficient dynamic data structures. Minimizing electrostatic energy through charge equilibration requires the solution of a large sparse linear system with a shielded electrostatic kernel at each sub-femtosecond long timestep. In this context, reaching spatio-temporal scales of tens of nanometers and nanoseconds, where phenomena of interest can be observed, poses significant challenges.
In this talk, first we present the design and implementation details of a sequential ReaxFF realization, SerialReax. SerialReax incorporates efficient dynamic data structures, algorithmic optimizations, numerical techniques and effective linear solvers to achieve excellent per-timestep time, linear time scaling in system size, and a low memory footprint. Upon the SerialReax platform, we build the Purdue Reactive Molecular Dynamics code, PuReMD. PuReMD inherits the highly desirable features of SerialReax and through use of scalable parallel algorithms, it extends current spatio-temporal simulation capability for reactive atomistic systems by over an order of magnitude. PuReMD is comprehensively validated for performance and accuracy on up to 3K cores on a commodity cluster (DoE/LLNL/Hera). Potential performance bottlenecks to scalability beyond our experiments have also been analyzed. PuReMD is available over the public domain and has been used to model diverse systems ranging from strain relaxation in Si/Ge/Si nanobars, water-silica surface interaction, and oxidative stress in lipid bilayers (biomembranes).
Link of the Week: The Periodic Table of Periodic Tables
Where can you find Seaborg’s extended periodic table and Segré’s chart of elements and isotopes — not to mention the periodic tables of visualization methods, rejected elements, Europeans, condiments, beer styles, comic books, and many more? Why of course, it’s the Periodic Table of Periodic Tables. Warning: Be sure tongue is planted firmly in cheek when viewing most of the links.
About Computing Sciences at Berkeley Lab
The Lawrence Berkeley National Laboratory (Berkeley Lab) Computing Sciences organization provides the computing and networking resources and expertise critical to advancing the Department of Energy's research missions: developing new energy sources, improving energy efficiency, developing new materials and increasing our understanding of ourselves, our world and our universe.
ESnet, the Energy Sciences Network, provides the high-bandwidth, reliable connections that link scientists at 40 DOE research sites to each other and to experimental facilities and supercomputing centers around the country. The National Energy Research Scientific Computing Center (NERSC) powers the discoveries of 6,000 scientists at national laboratories and universities, including those at Berkeley Lab's Computational Research Division (CRD). CRD conducts research and development in mathematical modeling and simulation, algorithm design, data storage, management and analysis, computer system architecture and high-performance software implementation. NERSC and ESnet are DOE Office of Science User Facilities.
Lawrence Berkeley National Laboratory addresses the world's most urgent scientific challenges by advancing sustainable energy, protecting human health, creating new materials, and revealing the origin and fate of the universe. Founded in 1931, Berkeley Lab's scientific expertise has been recognized with 13 Nobel prizes. The University of California manages Berkeley Lab for the DOE’s Office of Science.
DOE’s Office of Science is the single largest supporter of basic research in the physical sciences in the United States, and is working to address some of the most pressing challenges of our time. For more information, please visit science.energy.gov.