In this bimonthly feature, HPCwire highlights newly published research in the high-performance computing community and related domains. From parallel programming to exascale to quantum computing, the details are here.

Applying HPC to document and research rock art

This article – written by James Andrew Dodd – explores the application of HPC to remote processing of rock carving models. Dodd discusses his investigations into Southern Tradition rock art in Denmark, Norway and Sweden. He outlines prior work on the Danish e-Infrastructure Collaboration National High Performance Computing Abacus 2.0 and describes his test implementation using an AWS EC2 HPC cluster. He proposes a workflow, arguing that its implementation could break “new ground in the fields of rock art research, archaeology, and high performance computing in equal measure.”

Author: James Andrew Dodd

Using video games to design new proteins

In a new development led by a team from the Institute for Protein Design at the University of Washington School of Medicine, researchers have successfully encoded their knowledge into the computer game “Foldit,” allowing citizen scientists to design synthetic proteins. “One hundred forty-six Foldit player designs with sequences unrelated to naturally occurring proteins were encoded in synthetic genes,” wrote the authors. “56 were found to be expressed and soluble in Escherichia coli, and to adopt stable monomeric folded structures in solution.”

Authors: Brian Koepnick, Jeff Flatten, Tamir Husain, Alex Ford, Daniel-Adriano Silva, Matthew J. Bick, Aaron Bauer, Gaohua Liu, Yojiro Ishida, Alexander Boykov, Roger D. Estep, Susan Kleinfelter, Toke Nørgård-Solano, Linda Wei, Foldit Players, Gaetano T. Montelione, Frank DiMaio, Zoran Popović, Firas Khatib, Seth Cooper and David Baker.

Analyzing cost-performance tradeoffs of HPC network designs using simulations

Optimizing the network topology and its parameters is critical to HPC system design. In this paper – written by a team from Lawrence Livermore National Laboratory, Nvidia, and Argonne National Laboratory – the authors “identify different practical constraints that determine the number of nodes, routers, and links, and in turn, influence dollar costs and impact network design.” Using these constraints, the authors identify network topologies that perform best under different configurations and compare their performance.

Authors: Abhinav Bhatele, Nikhil Jain, Misbah Mubarak and Todd Gamblin.

Using a high-performance distributed object store for exascale weather and climate prediction

As the ambitions and needs of weather and climate prediction grow ever greater, the size and number of output data elements have grown by orders of magnitude. This paper – written by a team from the European Centre for Medium Range Weather Forecasts (ECMWF) – discusses novel hardware and software approaches to workflow and data management, including ECMWF’s meteorological object store, which acts as a “hot-cache” and supports multiple storage technologies. The paper presents extensions to the object store that allow it to support a wider range of hardware.

Authors: Simon D. Smart, Tiago Quintino and Baudouin Raoult.

Assessing the financial returns on investments for cyberinfrastructure facilities

Cyberinfrastructure is a costly enterprise – and more frequently, attention is being given to assessment of the value of cyberinfrastructure investments. In this paper – written by a team from a number of universities and the Roswell Park Comprehensive Cancer Center – the authors survey current methods for assessing the financial returns on investment (ROIs) of cyberinfrastructure. They conclude with a discussion of future research directions and challenges.

Authors: Craig A. Stewart, David Y. Hancock, Julie Wernert, Thomas Furlani, David Lifka, Alan Sill, Nicholas Berente, Donald F. McMullen, Thomas Cheatham, Amy Apon, Ron Payne and Shawn D. Slavin.

Using HPC frameworks for large-scale genome assembly

Genome sequencing data has exploded in size in recent years, necessitating increasingly heavy-duty hardware to handle the data load. In this dissertation – written by Sayan Goswami of Louisiana State University – Goswami presents a next-generation sequence assembler (“Lazer”) that achieves scalability while improving memory efficiency. He then presents a distributed GPU-accelerated assembler (“LaSAGNA”) that uses a single GPU to assemble large sequence datasets. Finally, he presents the first third-generation sequence assembler, which is able to assemble a ~150 GB human genome dataset in just over half an hour. Author: Sayan Goswami

Developing a scientific collaboration workspace for distributed HPC datacenters

Increasingly, networks are working to improve big data transmission between geographically disparate HPC datacenters. In this paper, a team from Sogang University proposes “SciSpace,” which would provide “a global view of information shared from multiple geo-distributed HPC data centers under a single workspace” to enable scientific collaborations across HPC datacenters. The authors evaluated their proposal by testing it on two geo-distributed small-scale HPC datacenters, demonstrating SciSpace’s feasibility.

Authors: Awais Khan, Taeuk Kim, Hyunki Byun and Youngjae Kim.

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.