The repercussions of complex data explosion are deeply felt in the field of astronomy especially radio astronomy where the science-data output of the order of petascale leading to exascale are rapidly becoming a daunting reality. The expert panel at the Supercomputing Conference 2010 in New Orleans LA recognized the computational challenges of Square Kilometer Array (SKA) radio telescope as one of the top real-world applications for baseline exascale computing. The roadmap leading to SKA and exascale computational challenge is punctuated with near-term telescopes with petascale requirements, viz., Atacama Large Millimeter/Sub-millimeter Array (ALMA), extended Very Large Array (eVLA), Low Frequency Array (LOFAR) etc. therefore making the problems current.
To avoid the impasse and successfully simulate, process and analyze unprecedentedly large volumes of data for generating important scientific results will require in depth research and development of new computational architectures, programming models, algorithms, standardizations, frameworks, and tools. Most importantly, the research and development of the eco-system will require transformative shift from the traditional in-house development methodology to close collaborations and co-design with community networks and cross-discipline expertise of high-performance computing (HPC) professionals along with radio astronomers and the industry. All these are necessary before and in parallel with telescope hardware development, and associated challenges such as efficient power provision and use.
The Exascale Radio Astronomy (ERA) conference of AASTCS provides a unique opportunity to marshal disparate expertise and focus on impending challenges in cross-disciplinary areas of radio astronomy to help define common solutions for a wider community.
Scientific Organizing Committee
(Chair and Organizer)
|Paul Shapiro||University of Texas|
|Melvyn Wright||UC Berkeley|
|Aaron Evans||University of Virginia/NRAO|
|Ilian T Iliev||University of Sussex|
|Leon Koopmans||LOFAR/Kapteyn Astronomical Institute|
|Garrelt Mellema||Stockholm University|
|Ue-Li Pen||Canadian Institute of Theoretical Physics|
|Hemant Shukla||LBNL/UC Berkeley|
|Oleg Smirnov||SKA/Rhodes University|
|Elena D'Onghia||University of Wisconsin|
|Sui Ann Mao||University of Wisconsin|
|Yuexing Li||Pennsylvania State University|
|Wen-mei Hwu||University of Illinois, Urbana-Champaign|
|Ana Varbenescu||Delft University|
The weeklong meeting will have each day devoted to specific topic for the day. The keynote speaker will set the stage followed by talks in sessions along the day. Members of the organizing committee will chair the sessions. Usual breaks will be provided between sessions. Before closing for the day, there will be a recap and discussion session.
For every session, notes will be maintained and will later be distilled for the white paper.
Day 1: Science Goals
The topics covered by the next-generation of radio telescopes will touch upon Cosmology, Dark Ages, Epoch of Reionization, Galaxy Evolution, Dark Energy, Dark Matter, General Relativity, Cosmic Magnetism, and Pulsars etc.
Day 2: Simulations
- Public Domain Codebases – List publically available codes. Explore opportunities for consolidating related and/or similar codes and make them widely available.
- Standardization – Define data format standards for large datasets. Storage and distribution mechanisms.
- Scalability – Make the codes scalable for large volume and/or high- resolution simulations.
- Reusability – The simulation codes should be modular and reusable thus avoiding redundant rewrites usurping valuable resources.
- Reproducibility – Simulations take massive resources and are hard to redo. There is clearly a need for defining mechanism third party validation of code and its results.
- Algorithms – Discuss novel ideas for simulations algorithms used in other fields. Explore new trends in adaptive mesh refinement (AMR), N-body, hydro-dynamics, radiative transfers etc.
- Modeling Telescope and Propagation Effects – Develop solutions for end- to-end simulations to help the overall design of current and future telescopes. Such simulations aid in comprehensive and realistic trade-off studies associated with noise in the data due to propagation effects.
Day 3: Instrumentation
- Digital Signal Processing (Software Correlators, Beam Formers)
- Real-time Adaptive Imaging
Day 4: Data Processing/Analysis
- Processing Pipelines
- Analysis Pipelines
- Radio Frequency Interference Excision
- Strategies for Massive Data Acquisition, I/O, Storage
Day 5: Emerging Computational Technologies
- Centralized Software Ecosystem – Standardized environments for development of peta- and exascale class simulation suites for end-to-end modeling scientific domain, instrumentation effects; data processing and analysis pipelines and software-based instrumentation.
- Hardware, Programming Models and Algorithms – Discover optimal and long-term solution for ensemble of architectures and languages/models best suited to represent the underlying algorithms.
- Exascale Risk Assessment and Mitigation – Evaluate immediate and long-term effects of disruptive technologies due to exascale initiative on the overall growth of scientific research and devise mitigation strategies.
- Power Estimate and Modeling – Measure power budgets and work with vendors to lower expenditure.
- Standardization and Cohesion of modules used for complex simulation. Validated ensembles of codebases used for high-resolution simulations of scientific domains.
- Access to domestic and international supercomputing test-beds – Expand existing mechanisms for seamless use of participating facilities.
- In-situ Visualization and Analysis. Standard and highly efficient framework to visualize and analyze exascale process at various stages and dynamically feedback decisive decisions.
Outreach and training
Continue the outreach through conferences, training and deployment of publically available components of the ERA eco-system.
The output of the meeting will be a comprehensive white paper that will cover extensive data challenges and their impact on current and future experiments in radio astronomy. The white paper will consolidate expert suggestions and solutions in wide range of areas.
ERA will invite world experts form radio astronomy, HPC and industry. A wide range of participation of students, post-docs, and young researchers will be encouraged and facilitated.
ERA will invite keynote and expert speakers and will solicit participation through talks and posters. The conference day will offer complete overview of topical challenges by setting the stage through the keynote followed by wide ranges of talks concluded by discussion.
The later part of the day will conclude with discussions followed by a summary and associated action items preferably by the keynote speaker or the chair of the session.