A-Z Index | Phone Book | Careers

InTheLoop | 05.20.2013

May 20, 2013

'No Exascale for You!' An Interview with Horst Simon

Twice during the week of May 6, Berkeley Lab Deputy Director Horst Simon gave presentations of a new talk on “Why We Need Exascale and Why We Won’t Get There by 2020.” Not only was the talk a hit with conference attendees, but it also made its way onto ExtremeTech and Slashdot. In an HPCwire exclusive, Simon talks about his presentation with Jon Bashor. Read more.

HPCwire also has an article on James Sethian and Robert Saye’s research on computing the physics of bubbles. Read more.

University of Virginia, ESnet Team Up to Win Best Paper

Networking has several definitions, including making key connections and moving information from one point to another. And a paper co-authored by engineers at ESnet and the University of Virginia proves the value of both — it was named a Best Paper at the Sixth International Conference on Communication Theory, Reliability, and Quality of Service, held April 21–26 in Venice, Italy.

The paper, “On How to Provision Quality of Service (QoS) for Large Dataset Transfers,” was written by Zhenzhen Yan, a Ph.D. student at the University of Virginia (UVA), UVA Professor Malathi Veeraraghavan, and Chris Tracy and Chin Guok of ESnet. Read more.

CRD’s Amy Chen Mentors Oakland High Girls to Honorable Mention in Global Competition

For the second year, Amy Chen, leader of the Biological Database Systems Group in CRD’s Biological Data Management and Technology Center, served as a volunteer mentor for the Technovation Challenge and helped her team of Oakland High School girls earn honorable mention. The Technovation Challenge brings together teams of girls to develop smart phone apps, then have the teams create business plans and “sell” their ideas to a panel of judges.

For 2013, the program was opened to teams around the world, Chen said, and 115 teams submitted their projects. The top 10 projects will compete in the finals, while the next 10 — including the team mentored by Chen — were awarded honorable mentions.

This year’s theme was “How to help your community,” and the Oakland High team came up with an app to encourage community gardening. The app provides users with information on temperatures, soil pH levels, and suggestions on what crops go best where. One feature allows users to seek answers to gardening problems. And when the harvest is in, users can use the app to swap surplus zucchinis for extra tomatoes.

Chen, who served as a mentor in 2012 when Berkeley Lab hosted Technovation Challenge teams from Berkeley and Albany high schools, said that there was a significant difference in the resources available. Whereas many of last year’s participants had their own laptops, the Oakland students relied on a single old laptop from their school. “But these girls were very smart and did a great job,” Chen said. “It’s great to see them develop their ideas, and it was very exciting for them to win an honorable mention.”

Other Berkeley Lab employees who served as mentors this year are Orianna DeMasi and Krishna Palaniappan, both of CRD, Katherine Copic (Physics), Michele Dunleavy (IT), and Ina Reichel (AFRD).

Sudip Dosanjh to Give Seminar at NCAR on Exascale and Big Data

NERSC Director Sudip Dosanjh will give a seminar at the National Center for Atmospheric Research (NCAR) in Boulder, Colorado, on Wednesday, May 22. Titled “Exascale and Big Data at NERSC,” the talk will give an overview of NERSC’s strategic roadmap for the next decade, with particular focus on enhancing data capabilities and services.

New Scientist: How Green Is Your Cloud?

Eager to get more computations for their watts, cloud providers and IT researchers are moving away from a one-size-fits-all approach to computing in data centers towards an era of specialization in which resources are better matched to the tasks at hand. A range of technologies, in both hardware and software, is being designed to handle the task, according to an article in the May 4 issue of New Scientist. John Shalf, NERSC’s Chief Technologist and head of CRD’s Computer and Data Sciences Department, is one of the experts quoted in the article. Read more.

What Would You Pack?

What would you need to survive for the first three days after an emergency? That’s the average time it takes for relief to arrive for natural disasters and other widespread emergency situations. This year’s Emergency Go Kit Contest allows you to show your resourcefulness and win a cash prize of up to $250. The contest is part of Berkeley Lab's Emergency Preparedness Fair and Road Safety Day on June 18. The top three finalists will display their kits, and the winners will be announced at the event. Entries are due by May 31. Read more.

This Week’s Computing Sciences Seminars

Trillion Particles, 120,000 Cores, and 350 TBs: Lessons Learned from a Hero I/O Run on Hopper

Thursday, May 23, 12:00–1:00 pm, OSF Room 238
WebEx: https://nersc-training.webex.com/, password: nerscosf
Teleconference: 866-740-1260, PIN 4866820
Suren Byna, Berkeley Lab Scientific Data Management Group

Modern peta-scale applications can present a variety of configuration, runtime, and data management challenges when run at scale. In this paper, we describe our experiences in running VPIC, a large-scale plasma physics simulation, on the NERSC production Cray XE6 system Hopper. The simulation ran on 120,000 cores using ~80% of computing resources, 90% of the available memory on each node and 50% of the Lustre scratch file system. Over two trillion particles were simulated for 23,000 time steps, and 10 one-trillion particle dumps, each ranging between 30 and 42 TB, were written to HDF5 files at a sustained rate of ~27 GB/s. To the best of our knowledge, this job represents the largest I/O undertaken by a NERSC application and the largest collective writes to single HDF5 files. We outline several obstacles that we overcame in the process of completing this run, and list lessons learned that are of potential interest to HPC practitioners.

Shared Memory Communication in MPI

Friday, May 24, 10:00–11:00 am, 50F-1647
Andrew Friedley, Center for Research in Extreme Scale Technology, Indiana University

Multi-core shared memory architectures are ubiquitous in both High-Performance Computing (HPC) and commodity systems because they provide an excellent trade-off between performance and programmability. The Message Passing Interface’s (MPI’s) abstraction of explicit communication across distributed memory is very popular for programming scientific applications. Unfortunately, OS-level process separations force MPI to perform unnecessary copying of messages within shared memory nodes.

This talk will discuss a novel approach that transparently shares memory across MPI processes executing on the same node, allowing them to communicate like threaded applications. We have developed two protocols for shared memory message passing that exhibit better performance than existing MPIs. Next, we introduce the concept of ownership passing to MPI. Rather than passing (copying) messages, we can pass ownership (access) to data among MPI ranks. The result is a “zero-touch” communication mechanism which results in further performance improvements. Ownership passing enables further shared memory communication techniques within MPI’s distributed memory model.

Algebraic Theory Based Electronic Structure Study

Friday, May 24, 3:00–5:00 pm, NERSC Video Room (50B-2222)
Aihui Zhou, Chinese Academy of Sciences

In this talk, we will talk about some electronic structure study making use of algebraic theory. Density functional theory (DFT) has become a basic tool for the study of electronic structure of matter, in which the Hohenberg-Kohn theorem plays a fundamental role in the development of DFT. Unfortunately, the existing proofs are incomplete. In the first part of this talk, we shall present a rigorous proof for the Hohenberg-Kohn theorem for Coulomb type systems using the Fundamental Theorem of Algebra. Kohn-Sham equation, a nonlinear eigenvalue problem, is the most widely used DFT model. In the second part, after using group theory to divide an eigenvalue problem into some groups of smaller ones that can be solved independently and two-levelly, we then apply the decomposition approach to electronic structure calculations of symmetric cluster systems, in which we solve successfully thousands of Kohn-Sham eigenpairs with millions of DOFs.

Link of the Week: Blinding Us from Science

Science is under attack. With corporations manufacturing uncertainty to undermine studies that hurt their bottom lines, and the sequester cutting billions in funding for scientific research, you’d think the American science community would be hunkered down in their labs avoiding outside interference at all costs.

A new project of the Union of Concerned Scientists (UCS), the Center for Science and Democracy, is encouraging scientists to do just the opposite. The center encourages scientists to speak out and help others to better understand scientific information and to distinguish evidence from political positioning. Theresa Riley of Moyers & Company spoke with the Center’s director, Andrew Rosenberg, by phone last week. An edited version of their conversation can be found here.