- Thom Dunning
National Center for Supercomputing Applications
I03: The Future of Extreme-scale Computing
- A new generation of supercomputers--petascale computers--will
soon be providing scientists and engineers with the ability to
simulate a broad range of natural and engineered systems with
unprecedented fidelity. Just as important in this increasingly
data-rich world, these new computers will allow researchers to
manage and analyze unprecedented quantities of data, seeking
connections, patterns and knowledge. The impact of this new
computing capability will be profound, affecting science,
engineering and society.
The National Center for Supercomputing Applications at the University of Illinois at Urbana-Champaign is in the final stages of deploying a computing system that can sustain a million billion calculations per second on a broad range of science and engineering applications and manage and analyze petabytes of data. This computer, Blue Waters, will have hundreds of thousands of compute cores; thousands of terabytes of memory; tens of thousands of disk drives, each storing terabytes of data; and hundreds of thousands of terabytes of tape storage. Using this extraordinary computing capability, computational scientists and engineers are preparing to dramatically extend our ability to explore the natural world, e.g., the structure and evolution of the universe, the birth of tornados, the impact of earthquakes, the functioning of the nanoworld, and the underlying genetic causes of disease, to name just a few.
The presentation will describe Blue Waters along with the current generation of petascale computers, concluding with a look toward exascale computing--the next major milestone in the advance of supercomputing technology--but one that presents challenges unlike any encountered to date.
- Tim Cornwell
Australian Telescope National Facility
I02: Meeting the challenge of Square Kilometre Array data processing
- The Square Kilometre Array (SKA) has passed two very
significant milestones in 2012. First the project has moved
into a 4 year preconstruction phase running from 2012 to
2016. Second the sites have been selected - both Australian and
South African sites have been chosen to host parts of SKA. In
the first phase of construction starting in 2016, a dish
array equipped with single pixel feeds will be built in South
Africa, and in Australia, two wide field instruments will
be built - a low frequency array similar to LOFAR and a phased
array feed array similar to ASKAP. The data processing needs
for all three telescopes are complex and large, tending up to
exascale. Our confidence in the feasibility of the data
processing is based on the experience with existing SKA
pathfinders and precursors, most notably ASKAP, LOFAR,
MeerKAT, and MWA. I will describe the lessons learned and how
these translate into planning for SKA.
- Ken Glotfelty
Chandra X-ray Center
I04: One Hop -- In the Right Direction
- Accelerated by the ease of access to data triggered by growing
archive registries and common search protocols, many more
application than before are now directly interfacing with
archives. Gone and going are the days when data sat behind
proprietary archive interfaces. What were once exclusively
analysis tools designed to consume files on disk are now
becoming first-class front ends to archives to discover and
retrieve remote data-sets. More than that, whether by saving
files to disk or sending and servicing SAMP messages, these
applications are themselves now local data service providers
for other tools.
The problem is that often data that are retrieve remotely are directly cast into the applications internal data model -- likely altering the data in the process: content may be added (eg new columns), nonessential information (to the current application) may be purged, provenance replaced, or content directly modified (eg WCS conventions). This then possibly insufficient or incompatible view of the data what is presented to postliminary tools in the work-flow. The remote data have in essence become trapped in the application that retrieved it from the remote service.
This talk will discuss this "single hop" effect and the strain it imposes on application developers. Current examples of the problem that exceed simple formatting issues will be presented, as will a discussion on why a "return to source" model of sending all applications back to the original remote data service is, in general, insufficient.
- Wen-Mei Hwu
University of Illinois Urbana-Champaign
I06: Meeting the Many-core Parallel Programming Scalability Challenge
- Many top supercomputers have employed many-core GPUs to boost their energy efficiency. For example, Blue Waters at the University of Illinois adds about 1/3 of peak compute rate by replacing about 6% of the AMD Interlagos CPUs with NVIDIA Kepler K20 GPUs. It has become increasingly clear that the most important metric in programming these systems is scalability. It is not only important for an application to achieve good performance for a given parallel hardware and data set, it must also be able to scale effectively in terms of hardware parallelism and data size. Parallelism scalability allows the application to take advantage of a wide range of current and future generation hardware, significantly reducing the cost of software development. Data scalability allows the application to handle the ever increasing data size in the real world. The rise of CPU-GPU heterogeneous computing has significantly boosted the pace of progress in this field. For example, we now have scalable sparse algorithms for several types of problems than most experts foresaw in 2007. There has been rapid progress in numeric methods, algorithm design, programming techniques, and optimization tools, and hardware support for developing scalable applications. In this talk, I will first give an overview of the Blue Waters system and its intended applications. I will then discuss recent advances in programming such systems, their implications on science and engineering research, and future research opportunities.
- Thomas Bennett
I01: Square Kilometer Array South Africa
- The South African SKA project is a multifaceted effort, with a
number of sub-projects that fall under its authority playing a
prominent role in shaping the future of science in Africa over
the next few decades.
From humble beginnings, the current effort has yielded not only the siting of a large part of the SKA, but also the MeerKAT array currently taking shape in South Africa's Karoo region. In addition, the effort has driven an unprecedented human capacity build-up programme, planted the seeds for the African VLBI Network, and hosted international collaborations such as the PAPER and C-BASS instruments near the South African SKA site.
The goal of the this talk to update the ADASS community on the ongoing efforts that fall under the SKA South Africa banner as well as focus on some of the technical issues of interest. This includes, inter alia, an overview of early science results undertaken using the MeerKAT 7-dish precursor array, the current progress and development of the MeerKAT array and with the SKA site announcement, areas of focus for SKA South Africa in terms of the broader SKA development.
- Robert Hurt
NASA Infrared Processing and Analysis Center
I05: AstroPix and the Role of Metadata in Outreach Imagery
- It is hard to find a class of imagery that has a richer layer of underlying content than those found in astronomy. The "pretty pictures" of the cosmos are rendered from research datasets spanning the spectrum, and when released to the public they are accompanied by detailed descriptive context. The desire to connect the pictures with this information was the underlying goal of the Astronomy Visualization Metadata (AVM) standard, which is one element of the IAU's Communicating Astronomy with the Public working group. This standard has been embraced by a growing number of observatories (including Spitzer, Chandra, Hubble, & ESO) and is being built into existing outreach workflows. This year the AstroPix website has moved into public beta (astropix.ipac.caltech.edu), serving as an aggregator for tagged image libraries and providing an elegant one-stop shopping site for astronomical images. AstroPix serves as an example of how metadata-rich public images can greatly enhance the user experience in astronomy, and can open up novel new applications spanning the web, mobile devices, schools, and museums. AVM can easily be integrated into any software generating images for any purpose helping enrich the utility of a simple JPEG pulled from an archive site.
- JJ Kavelaars
National Research Council of Canada
I07: Moving source detection, science from the VO
- There has been a substantial increase in the number of operating and planned ground (Catalina Sky Survey, Space Watch, SDSS, Canada-France Ecliptic Plane Survey, PanStarrs, SkyMapper, LSST, etc.) and space (eg. NEOSSAT, Euclid) based surveys searching for objects within the solar system. These surveys must attempt to maximize their detection of sources that move through the images and tracking those objects until a firm orbit can be determined, while minimizing the reporting of false positives. These sorts of surveys have also lead to a growing industry of moving object search interfaces at the archive level (such as the CADC's Solar System Object Search). In this presentation I will review the general science cases for moving object surveys (from NEOs to Oort Cloud Objects) and the various innovative (and brute force) algorithms and computing systems that have been developed to crack the moving object detection problem.
- Zhong Liu
Yunnan Astronomical Observatory
I08: The Chinese Giant Solar Telescope
- In order to detect the fine structures of the magnetic field and velocity field in the photosphere and the chromosphere of the Sun, an 8-meter solar telescope has been proposed by the Chinese solar community and several overseas colleagues. Due to the advantages of the ring structure in the polarization detection and the thermal control, the basic structure designed for CGST (Chinese Giant Solar Telescope) currently is an 8-meter ring. The spatial resolution of CGST is equivalent to an 8-meter diameter telescope of full aperture, and the light-gathering power is equivalent to a 5-meter full aperture telescope. The integrated simulation of optical system and imaging ability such as optical design, MCAO, active maintenance of primary mirror will be described in this paper. Mechanical system was analyzed by finite element method. The results of simulation and analysis showed that the current design could meet the requirement for most scientific goals in infrared, near infrared band, and even visible bands. It is supported by CAS and NSFC as a long term astronomical project.
- Pedro Osuna
European Space Astronomy Center
I09: Enhancing science return with scientific archives
- Science Archives have taken on prime importance in scientific research,
enabling repeated analysis as well as analysis combining
several datasets of different projects in the same research
area. From "RAW" data to "highly processed", science data are
made available to scientists through Science Archives.
The prime importance of Science Archives is illustrated by several cases, among which the Hubble Space Telescope and the XMM-Newton projects are good examples. With varying sizes in their respective science community (bigger for the HST than for XMM-Newton due to the wavelength range covered by both) the amount of presently registered users in their respective science archives amounts to nearly 8000 for HST and around 3500 for XMM-Newton.
The abstract of the paper "Lessons from a High-Impact Observatory: The Hubble Space Telescope's Science Productivity between 1998 and 2008" (see ref.) states in a sentence:
The overall productivity, however, is still steadily increasing due to the increasing number of archival articles...
Similar study is being conducted currently for the XMM-Newton data, for which the number of publications has recently reached the 3000 number in the refereed literature.
Benchmark activities within ESA and other agencies (e.g. NASA) have shown that partners attach a high criticality to the proper operation of the archive facilities and their front end that is the main interface for general astronomers or scientists within the mission. It is of crucial importance for the science return of space based missions to produce powerful, flexible and usable scientific archives.
With the upcoming huge data production missions, like GAIA (around 1PetaByte of data estimated), Euclid (recently approved ESA mission, with tens of PetaBytes estimated) or Solar Orbiter (also ranging int he PetaBytes), archiving technologies have to be reassessed and state of the art technology is under strong pressure to find the best ways to cope with the data production capabilities of new missions.
This talk will describe the experience gained int he last 15 years building ESA Science Archives, the lessons learned and the plans to address the challenges for the upcoming huge data missions.
- Matthew Schneps
I10: Using an iPad App Reverses Tenacious Misconceptions About the Scale of Space
- A quarter-century ago the video "A Private Universe" showed
that graduates of Harvard frequently fail to understand
fundamental ideas about the seasons and the phases of the moon
taught in grade school. The video showed that ideas children
develop early in life hold a deep-seated logic that can make
their nonscientific ideas remarkably resistant to instruction.
In astronomy, many such misconceptions can be traced to ideas
that critically depend on an accurate understanding of the
scale of space.
Few opportunities in life prepare children with experience they need to build scientifically accurate intuitions about the scale of space. Nonscientific notions about scale are often reinforced by instruction, where drawings and models of the solar system invariably exaggerate the sizes of the planets in relation to their orbits. Today, however, students have access to multi-touch tablet computers that are able to render interactive, photorealistic models of the solar system in their true scale. As a result, students for the first time can have access to rich experiences about the scale of space, unimaginable at the time "A Private Universe" was produced.
In this presentation we critically examine whether experience afforded by such technology can successfully address previously intractable learning challenges related to scale. We describe preliminary results from a study examining learning gains using Apple iPads. Here, 152 high school students manipulated a virtual simulation of the solar system rendered by the Solar Walk app in a true-to-scale (TTS) display. The research design compared the TTS rendering to that of a more traditional display that exaggerated the scale (Orrery). Gains in learning were measured using a validated multiple-choice instrument, comprised of items strongly confounded by well-studied astronomical misconceptions.
Preliminary findings show that students using the TTS displays are significantly more successful in reversing their misconceptions overall, when compared to those who use an Orrery display. Furthermore, as expected, these gains were especially strong for questions where the concept of scale was an important factor. A surprising exception was the famous question about the causes of the seasons, featured in "A Private Universe." Only 1% of the students responded to this question correctly prior to instruction. However, iPad intervention produced no gains for this concept.
Overall, these preliminary findings support the assertion that experience with true-to-scale simulations enabled by multi-touch tablet computers can successfully address otherwise challenging scale-related misconceptions in astronomy. However, certain ideas (such as the seasons) appear to be shaped by such deep-seated notions, that even exposure to the rich experience afforded by this new technology has no influence on these misconceptions.
- Alexandar S. Szalay
Johns Hopkins University
I11: Extreme Data-Intensive Computing in Astrophysics
- Astronomical discoveries always had their origins in data. With new surveys already in progress and soon to be started, the amount of data is growing faster than ever. The analysis of these data sets requires substantial changes in how we approach the analysis. These changes are reflected in system architectures, algorithms, toolkits and statistical techniques. The talk will explore the current state of the art of data-intensive computing in astrophysics and will explore near term trends, with focus on both observational data and large numerical simulations.