A-Z Index | Phone Book | Careers

InTheLoop | 08.05.2013

August 5, 2013

Shoshani, Bautista, Nugent and Maroudas to Receive Director’s Achievement Awards at Thursday, Aug. 8 Ceremony

CRD’s Arie Shoshani and Peter Nugent, NERSC’s Elizabeth Bautista and HR's Maria Maroudas will be honored at a ceremony recognizing the recipients of the 2013 Director’s Award for Exceptional Achievement and the Berkeley Lab Prize – Lifetime Achievement Awards. The ceremony will be held at 2 p.m. Thursday, August 8, in the Building 50 Auditorium. The Berkeley Lab Prize-Lifetime Achievement Award will be presented to Shoshani, head of CRD’s Scientific Data Management Group. A Director’s Award for Exceptional Achievement in the area of Diversity will be given to Bautista, leader of NERSC’s Operation’s Technology Group. Nugent, who is co-leader of the Computational Cosmology Center, will receive a Director’s Award for Exceptional Achievement in the area of science. Maria Maroudas, a member of the HR group supporting Computing Sciences, will share a Director’s Award for Exceptional Achievement in the area of operations for her role as a member of the HR Recruiting/Hiring Technology Improvement group.

New Topological Technique by CRD Researchers Helps Scientists See & Search Large Data Sets

New computational techniques developed at Berkeley Lab may help save scientists from drowning in their own data. Computational scientists in the Computational Research Division have figured out how to streamline the analysis of enormous scientific datasets using a new method called Distributed Merge Trees. The analysis uses the same techniques that make complex subway systems understandable at a glance.

Gunther Weber and Berkeley Lab postdoctoral researcher Dmitriy Morozov of CRD’s Visualization Group pioneered the distributed merge tree approach to topological data analysis. They describe their work in a paper published in PPoPP’13: Proceedings of the 18th ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming. Read more.

Internet2 Blog Recaps July 17-18 Workshop with ESnet Examining Network Issues in Life Sciences

 Life sciences researchers, bioinformaticians and network engineers met at Lawrence Berkeley National Laboratory July 17-18 to discuss biomedical research and networking in the age of Big Data. Inder Monga of ESnet and Stephen Wolff of Internet2 hosted the inaugural Focused Technical Workshop. During the two-day workshop, three keynoters and four expert panels explored advanced networking, cloud and storage architectures, workflow engines, science gateways and Big Data transport. Read the blog.

Former ESnet Summer Student Wins Best Paper Award for Research on Energy Efficiency

 Baris Aksanli, a UC San Diego Ph.D. student who worked at ESnet in the summer of 2011, received Best Student Paper honors at the 18th IEEE Symposium on Computers and Communications held July 7-10. Split, Croatia. ESnet Chief Technologist Inder Monga, who supervised Aksanli and still helps him with research, was a co-author of the paper. About his paper on “A Comprehensive Approach to Reduce the Energy Cost of Network of Datacenters,” Aksanli said “There are a lot of papers which study the job migration over WAN. However, they mostly concentrate on a single aspect of these networks of datacenters and ignore the rest. In our paper, we specifically focus on what happens if the cost of energy is not modeled accurately and how it affects the network usage as well. We also include renewable energy integration in this study to account for clean energy and propose a job migration algorithm to account for all listed.”

Other research usually focus on only a small subset of network parameters, according to the paper’s abstract and thus their results may have large errors. “We present a comprehensive approach to minimize the energy cost of networks of datacenters by modeling performance of the workloads, power contracts, local renewable energy sources, different routing options for WAN and future router technologies. Our method can reduce the energy cost of datacenters by up to 28 percent, while reducing the error in the energy cost estimation by 2.6x.”

ACM Seeking Applications from Students for Support to Attend SC13 Conference

 The Association for Computing Machinery’s (ACM) Special Interest Group on High Performance Computing (SIGHPC) is offering two opportunities for student (undergraduate or early career graduate student) travel grants to attend the SC13 conference to be held Nov. 17-22 in Denver. One program is open to Students at the undergraduate or early graduate level, enrolled in an accredited university in any country and who are members of SIGHPC. For the second opportunity, SIGHPC has teamed with ACM's Women in Computing (ACM-W) division to help undergraduate and graduate women students attend SC13. Three grants will be provided in each program -- each awarded student will be reimbursed for up to $600 for travel within North America - or $1,200 if travel originates outside of North America.

The deadline for ACM-W grants is Thursday, August 15; the deadline for SIGHPC grants is Sunday, September 15. Students awarded a travel grant under either program will be assigned a mentor to help them navigate the conference. For details see the SIGHPC web page. Send any questions to students@sighpc.org.

This Week’s Computing Sciences Seminars

Overview of Turn Data Management Platform for Digital Advertising

Monday, August 5, 2 – 3 p.m., Bldg. 50B, Room 4205
Hazem Elmeleegy, Senior Engineer, Turn Inc.

Abstract:  In this talk, I will give an overview of Turn Data Management Platform (DMP). I will explain the purpose of this type of platforms, and show how it is positioned in the current digital advertising ecosystem. I will also describe the key components in Turn DMP, which cover the functions of (1) data ingestion and integration, (2) data warehousing and analytics, and (3) real-time data activation. For all components, I will touch on the main technical and research challenges, as well as the alternative design choices. One of the main goals of this talk is to highlight the central role that big data management is playing in shaping this fast growing multi-billion dollars industry.

SDN Security and Resilience

 Tuesday, August 6, 10 – 11 a.m., Bldg. 50B, Room 1237
Dongting Yu, University of Cambridge
Abstract:  As Software Defined Networks (SDN) are becoming a commercial reality, so has become the security and resilience aspects of these networks which have so far been focusing on performance. This talk presents an overview of our project’s work in this direction. I will first discuss the different deployment scenarios for SDN and their attack surfaces and threat models, followed by outlining a hierarchical SDN architecture that will provide the needed security and resilience properties. Finally I will discuss our ongoing work on SDN for BGP and attacking openvswitch through denial of service with some preliminary results.

Performance Analysis Gap: Processor Complexity Keeps Climbing – Developers Are More Naïve than Ever

Thursday, August 8, 10 – 11 a.m., Bldg. 50F, Room 1647
Wucherl Yoo, Computer Science Department, University of Illinois
Abstract:  The performance analysis gap is widening as processor complexity keeps climbing and developers are becoming more naïve than ever. Seemingly suitable programs can run correctly, but may suffer from hidden hardware bottlenecks that can severely hinder performance. Performance Monitoring Unit (PMU) events can provide programmers with unique and powerful insights into performance problems in their programs, but interpreting these events has been a significant challenge. While the conventional performance tuning tools can measure and visualize hardware events, they lack automatic identification of dominant resource bottlenecks and significant manual effort is required from experts to interpret the hardware events. 

To overcome this performance analysis challenge, I present a system, ADP (Automated Diagnosis of Performance pathologies) that can automatically detect a program's performance problems. ADP provides automated detection of performance problems associated with a predominant hardware resource bottleneck in parallel and single-threaded applications. ADP can discover bottlenecked resources by profiling applications with the hardware events from performance monitoring Units (PMUs). I demonstrate how ADP has successfully identified performance bottlenecks in SPEC CPU 2006 and PARSEC benchmarks. Furthermore, I present our progress to extend ADP for vision applications and OpenMP applications.

Link of the Week: What Happens Every Single Minute on the Internet

A new infographic by rewarded search site Qmee offers a glimpse at the digital doings that occur globally over the span of each minute.