# InTheLoop | 09.04.2012

September 4, 2012

## The Computational Keys to Future Scientific Discoveries

A new camera at the hard x-ray tomography beamline of Berkeley Lab’s Advanced Light Source (ALS) allows scientists to study a variety of structures as a function of time—from bones to rocks, plants, and even metallic alloys—in unprecedented detail. The new camera generates data 50 times faster than the one it replaced, and researchers hope that all of this information may someday lead to more effective methods of storing carbon underground, treating diseases, or creating stronger materials for supersonic jets.

But before these breakthroughs can happen, ALS scientists must figure out how to manage, store, and share the torrent of data being generated. And they are not alone. From astronomy to genomics, increasingly sophisticated instruments are producing data at staggering rates, and many scientists are struggling to find the right computational and storage strategies to deal with the deluge. To help researchers in this effort, NERSC developed a Data Intensive Computing Pilot. Read more.

## Sept. 12 Webinar: Using Globus Online for ESnet Users

Staff from ESnet, DOE’s high-speed science network, and Globus Online, a fast, reliable file transfer service, will offer a free webinar on how ESnet users can more easily transfer data. The free webinar will be held from 11 am to 12:30 pm Wednesday, Sept. 12 (Pacific time). To register, go here. This webinar will provide an overview of a Science DMZ, and a demonstration of the most frequently used Globus Online features, tips and tricks, and Q&A for integrating these tools into the scientific workflow. Read more.

## LBNL W3C Rep Blog Tracks Development of Web Standards

Annette Greiner, a web developer at NERSC, has started the LBNL W3C Rep Blog to keep web developers at the Lab informed about World Wide Web Consortium (W3C) activities. W3C is an international community that develops open standards to ensure the long-term growth of the Web.

Annette has been posting monthly updates about what’s happening at the W3C. Since the Lab is a member organization, we can have a voice in its activities, and Lab employees can join working groups. A fair number of us are already involved in the High-Performance Computing Community Group, some in the Big Data Community Group, and one in the OWL (Web Ontology Language): Experiences and Directions Community Group.

As the Lab’s representative on the W3C Advisory Committee, Annette can voice opinions about W3C actions and can nominate Lab employees to W3C working groups. The Advisory Committee votes on many W3C actions, such as the chartering of new working groups, the movement of proposed recommendations from one stage to another, and changes to the consortium’s standards development process. Your input and feedback are most welcome.

## Volunteers Needed for “Science at Warp Speed” at Berkeley Lab Open House

“Science at Warp Speed: A Journey Through Space and Time in 3D” is the theme of the Computing Sciences exhibit at this year’s Berkeley Lab Open House, which takes place Saturday, Oct. 13, from 10 am to 3 pm. Here is the full description:

Travel 14 billion years back to the beginning of space and time. Soar through burner flames and supernova explosions. Explore the microscopic pores of rocks buried deep inside Earth. And if you are still not tired, zip around the world at 100 gigabits per second. Sound impossible? It’s not, researchers go on this journey every day with the ultimate scientific instruments—supercomputers and networks. Part time-machine, spaceship, microscope and more, these tools allow humans to explore environments and events that are too big, too small, too fast, too slow, or too dangerous to study otherwise.

Volunteers who work a minimum of three hours will receive a free T-shirt and a voucher for lunch. Go here to register; under “Attendee Type,” choose “Volunteer: Assigned to an Exhibit,” and send an e-mail to Computing Sciences’ Open House coordinator Linda Vu to let her know that you’ve registered.

## Time Magazine: The Great, Exploding Monty Python Star

Astronomers have for the first time observed a nova-producing system turn into a supernova, a finding that indicates the universe has more than one way to create a Type 1a supernova. The Time Magazine story credits Peter Nugent of CRD’s Computational Cosmology Group and several colleagues. The author comments, “The method used to detect the phenomenon is almost as impressive as the fact that it happened at all.” Read more.

## Sept. 14 Deadline for Contributions to Joint US-Asian R&E Networks Conference

TIP2013, the Winter 2013 technical meeting between the Joint Techs meeting and the Asia Pacific Advanced Network (APAN), has issued a call for proposals. The meeting will be held January 13–16, 2013, at the University of Hawaii and will include a full day of tutorials and workshops. Abstracts need to be submitted by Friday, September 14, 2012, using this link. Abstracts must include collaborators or intended demonstrations. Submissions will be judged by a formal program committee process based on relevance to posted focus areas and audience areas of interest.

The following topics have been identified as being of particular interest, although proposals in other areas relating to research and education networking are also invited:

• Network Research and Emerging Technologies, including software defined networking/OpenFlow; security in R&E networking; network performance and measurement; and “Future Internet.”
• Network Architecture and Operations, including 100G implementation experiences; IPv6; network performance and utilization; R&E network infrastructure and connectivity updates; and Pacific Islands connectivity.
• R&E Network Applications, including astronomy; ocean observatories; climate and environment; disaster management; e-medicine and e-health; e-culture and performance; global collaboration; and e-learning.

ESnet’s Inder Monga is a member of the program committee.

## IEEE IPDPS 2013 Issues Call for Papers

The 27th IEEE International Parallel and Distributed Processing Symposium, to be held May 20–24, 2013, has issued a call for papers. Authors are invited to submit manuscripts that present original unpublished research in all areas of parallel and distributed processing, including the development of experimental or commercial systems. Work focusing on emerging technologies is especially welcome. Topics of interest include, but are not limited to:

• Parallel and distributed algorithms, focusing on issues such as: stability, scalability, and fault-tolerance of algorithms and data structures for parallel and distributed systems, communication and synchronization protocols, network algorithms, scheduling, and load balancing.
• Applications of parallel and distributed computing, including web applications, peer-to-peer computing, parallel crowd sourcing, social network analysis, management of big data, cloud and grid computing, scientific applications, and mobile computing. Papers focusing on applications using novel commercial or research architectures, or discussing scalability toward the exascale level are encouraged.
• Parallel and distributed architectures, including architectures for instruction-level and thread-level parallelism; petascale and exascale systems designs; special-purpose architectures, including graphics processors, signal processors, network processors, media accelerators, and other special purpose processors and accelerators; impact of technology on architecture; network and interconnect architectures; parallel I/O and storage systems; architecture of the memory hierarchy; power-efficient and green computing architectures; dependable architectures; and performance modeling and evaluation.
• Parallel and distributed software, including parallel and multicore programming languages and compilers, runtime systems, operating systems, resource management including green computing, middleware for grids, clouds, and data centers, libraries, performance modeling and evaluation, parallel programming paradigms, and programming environments and tools.

Abstracts are due September 24, 2012, and full manuscripts must be received by October 1, 2012. Go here for more information.

CRD’s Aydin Buluç, Xiaoye Sherry Li, and Esmond Ng are members of the program committee.

## This Week’s Computing Sciences Seminars

Design Methodology for Domain-Optimized Computer System Architectures
Tuesday, Sept. 4, 10:00–11:00 am, 50B-4205
Daniel R. Burke, Berkeley Wireless Research Center

Project ISIS is an effort toward application-driven hardware design, which seeks to develop tools and methodologies to foster rapid evolution of computing systems that are better tuned to the application requirements of demanding scientific applications and result in more cost-effective and efficient future HPC systems.

The goal is to develop and demonstrate a design methodology for domain-optimized computer system architectures; this talk will present the status of our emerging techniques and tools for rapid prototyping and design-space exploration including the Chisel language, RAMP simulation, and Chiplets manycore 2-D substrates.

Par Lab Seminar: A Graphics Performance Wishlist: What We Would Like Hardware to Be Good At
Tuesday, Sept. 4, 1:00–2:30 pm, 430 Soda Hall (Wozniak Lounge), UC Berkeley
Geoffrey Irving

Algorithms in computer graphics and numerical simulation often possess high intrinsic parallelism, and should be well suited to hardware acceleration on GPUs or similar devices. To explore how well current hardware achieves this goal, I will discuss several representative problems taken from these two fields. For each one, I’ll present the structure and characteristics of state of the art massively parallel algorithms for the problem, the limiting factors controlling performance on current GPUs, and what ideal hardware might look like for each problem. In terms of GPU performance in particular, the conclusion is that most parallel algorithms map across fairly well (at least given the recent dynamic parallelism extension) given sufficient programmer time and expertise, saturating either bandwidth or compute depending on the problem. Therefore, the hard challenge (unsurprisingly) is to lower the effect required to perform this mapping. Caveat: My background is in algorithms, not GPU acceleration per se, so this talk depends heavily on knowledge collected from more knowledgeable colleagues. All mistakes are my own.

Block Gram-Schmidt Algorithms with Reorthogonalization: LAPACK Seminar
Wednesday, Sept. 5, 12:10–1:00 pm, 380 Soda Hall, UC Berkeley
Jesse Barlow, The Pennsylvania State University

A Gram-Schmidt step can be thought of as attempt to solve the following problem: given a nearly left orthogonal matrix $U \in \mat{m}{t}$ and $B \in \mat{m}{p}$ where $t+p \leq m$ find $Q_B \in \mat{m}{p}$ left orthogonal, $S_B \in \mat{t}{p}$ and $R_B \in \mat{p}{p}$ such that $Q_B R_B + U S_B = B$ and $U^T Q_B =0$. As written, this problem can be ill-posed if $\bear{cc} B & U \enar$ does not have full column rank. In this talk, a general approach to the solution of this problem is given which leads to some useful block algorithms.

Two contexts for these algorithms are important, the Q-R decomposition necessary for Krylov space methods and the modification of Q-R decompositions when columns are added or rows are deleted. Block Gram-Schmidt algorithms with reorthogonalization are given for both of these problems. Some of this work is joint with Alicja Smoktunowicz of Warsaw University of Technology, and some of the more recent work was done here at UC Berkeley’s Par Lab.

EECS Colloquium: Cybercasing the Joint: On the Privacy Implications of Multimedia Retrieval
Wednesday, Sept. 5, 4:00–5:00 pm, 306 Soda Hall, UC Berkeley
Gerald Friedland, International Computer Science Institute (ICSI)

In this talk, I present recent case studies that highlight the potential for multimedia retrieval of online (social network) data to support real-world attacks. Multimedia Retrieval, i.e., the task of matching and comparing multimedia content across databases, has rapidly emerged as a field with highly useful applications in many different domains. Researchers from different areas in signal processing and computer science (including the presenter) have invested significant effort into the development of convenient and efficient retrieval mechanisms. While retrieval speed, flexibility, and accuracy are still research problems, this talk will demonstrate that they are not the only ones.

This talk aims to raise awareness for a rapidly emerging privacy threat that we termed “cybercasing”: leveraging information available online to mount real-world attacks. Based on the initial example of geo-tagging, I will show that while users typically realize that sharing information, e.g., on social networks, has some implications for their privacy, many users 1) are unaware of the full scope of the threat they face when doing so, and 2) often do not even realize when they publish such information. The threat is elevated by recent developments that make systematic search for information (either posted by humans or by sensors) and inference from multiple sources easier than ever before. However, even with relatively high error rates, multimedia retrieval techniques can be used effectively for different real-world attacks by using “lop-sided” tuning; for example by favoring low false alarm rates over high hit rates when scanning for potential victims to attack.

This talk presents a set of scenarios demonstrating how easy it is to correlate data with corresponding publicly available information for compromising a victim’s privacy.

Memory and Performance Issues in Sparse Direct Solvers
Friday, Sept. 7, 10:00–11:00 am, 50F-1647
François-Henry Rouet, ENSEEIHT-IRIT, Université de Toulouse, France

We consider the solution of large sparse linear systems on parallel architectures using direct solvers. We are interested in memory issues in the two computationally intensive phases of these solvers: the numerical factorization and the triangular solution step. In the first part we consider the memory scalability of the multifrontal factorization and suggest a “memory-aware” mapping and scheduling algorithm that maximizes performance while enforcing a given memory constraint. In the second part we consider problems where sparse right-hand sides and sparse solution vectors appear; in particular, we focus on the computation of inverse entries. We show how exploiting sparsity can reduce the memory footprint and the computational cost. Many applications exhibit a large set of right-hand sides that cannot be processed simultaneously and must be divided into blocks; this entails different partitioning problems that we present. Throughout the talk we present experimental results on large matrices from real-life applications using the MUMPS solver. In a last part we also mention some work related to the domain decomposition-based solver PDSLin.

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.