In this bimonthly feature, HPCwire highlights newly published research in the high-performance computing community and related domains. From exascale to quantum computing, the details are here.
Enhancing HPC system log analysis by identifying message origin
Supercomputers incorporate a wide variety of systems, each with its own logs, creating a deluge of information. In this research piece, a team from Los Alamos National Laboratory examined ways to improve its processing of the hundred million system log messages that are logged each day. The team discusses how, by associating the system logs with the lines of source code they originated from, system administrators could communicate more effectively with vendors and developers. They demonstrate a prototype tool that matches source code lines and files with high confidence.
Authors: Megan Hickman, Dakota Fulp, Elisabeth Basemen, Sean Blanchard, Hugh Greenberg, William Jones, and Nathan DeBardeleben
Programming for HPC with automatic energy efficiency
Energy use is a growing concern and limitation for supercomputers, and current development processes involve painstakingly checking to see how each new version of a piece of software affects energy use. In this paper, a team of French researchers introduce an approach to automate “green programming” based on the automatic generation of code versions. They evaluate different kinds of production scenarios to illustrate the possible benefits of the automation tool.
Authors: Issam Rais, Hélène Coullon, Laurent Lefevre, and Christian Perez
Boosting edge computing performance through heterogeneous manycore systems
Supporting massively parallel workloads for large numbers of users is a growing burden for edge computing. In this paper, written by a team of researchers from the Electronics and Telecommunications Institute in South Korea, the authors highlight “how the features and capabilities of manycore servers can be leveraged to provide efficient edge computing services.” They discuss key challenges and features for supporting services on those systems.
Authors: Ramneek, Seung-Jun Cha, Seung Hyub Jeon, Yeon Jeong Jeong, Jin Mee Kim, Sungin Jung, and Sangheon Pack.
Using Agile Condor to Demonstrate HPC in Flight
Flight sensors can produce massive amounts of data — and now, for the first time, advanced machine learning was demonstrated during flight. In this article, written by a team from the Air Force Research Laboratory and SRC, Inc., the authors outline the demonstration, which utilized the recently-invented HPC architecture called “Agile Condor.” The article goes on to describe the benefits and innovations of the Agile Condor system as well as potential future applications.
Authors: Mark Barnell, Courtney Raymond, Chris Capraro, Darrek Isereau, Chris Cicotta, and Nathan Stokes.
Accelerating DNA data analysis using HPC
Next-generation sequencing technologies have sped up the sequencing of DNA, but the massive amount of data generated by those technologies has created a new bottleneck at the analysis stage. In this doctoral thesis, written by a candidate from the Delft University of Technology, the author proposes new strategies for accelerating important bioinformatics algorithms using GPUs and FPGA, discussing how these approaches can dramatically improve speed.
Author: S. Ren
Developing a methodology for the rapid development of scalable HPC data services
HPC storage workloads are growing – and with them, doubt that parallel file systems will be sufficient in the long run. In the midst of a search for a single new solution, these authors – a team from Argonne National Laboratory, Los Alamos National Laboratory, The HDF Group, and Fermi National Laboratory – argue that “custom data services should be designed and tailored to the needs of specific applications on specific hardware.” The authors discuss the benefits of this approach using three case studies.
Authors: Matthieu Dorier, Philip Carns, Kevin Harms, Robert Latham, Robert Ross, Shane Snyder, Justin Wozniak, Samuel K. Gutiérrez, Bob Robey, Brad Settlemyer, Galen Shipman, Jerome Soumagne, James Kowalkowski, Marc Paterno, and Saba Sehrish.
Using quantum computing to solve linear algebra problems and advance engineering
Quantum computing has enormous potential to speed the solving of linear algebra problems, which are of particular interest for engineering applications that use finite element and finite difference methods. The authors of this paper – Guanglei Xu and William S. Oates – explore quantum linear algebra problems using a quantum circuit that can be tested on IBM’s quantum computing hardware. They demonstrate the algorithm’s capabilities and its measurement methodology.
Authors: Guanglei Xu and William S. Oates
We also bring your attention to a group of nine papers in the October edition of the journal Frontiers of Information Technology & Electronic Engineering focusing on post-exascale supercomputing. As noted in the editorial lead-in article (authored by Zuo-ning Chen, Jack Dongarra, and Zhi-wei Xu), the Chinese Academy of Engineering organized the special edition by inviting position papers from leading experts inside and outside China.
“This special issue targets 2020–2030 supercomputing systems that go beyond the existing exascale systems under construction,” the authors write. “It focuses on innovative research ideas in systems architecture, processors, memory, storage, interconnects, operating systems, programming languages and compilers, and application frameworks. It foresees the convergence of high-performance computing (HPC) with big data computing, intelligence computing (e.g., deep learning), cloud computing, and edge computing, and encourages consideration of future HPC workloads.”
Among the perspective papers, a research team from China’s National University of Defense Technology sets out their ambitious goal to build a zettascale computer by 2035.
Do you know about research that should be included in next month’s list? If so, send us an email at [email protected]. We look forward to hearing from you.