• 2017 ASCR Computer Science Principal Investigators Meeting (CS PI Meeting): Resilience, SSIO, Design Space and SDMAV
    03/14/2017 - 03/16/2017
    Bethesda
    Maryland

2017 ASCR Computer Science Principal Investigators Meeting (CS PI Meeting): Resilience, SSIO, Design Space and SDMAV

The purpose for this meeting is to review the current state of research through technical talks and poster sessions. In addition to allowing the Program Manager to assess the current state of research through technical talks and poster sessions, additional goals and objectives for the Program Review Meeting include:

  • Participants will understand the evolving hardware architecture for next-generation of supercomputers and future exascale computing systems and evaluate the impact the changing architecture will have on their research;
  • Participants will understand the state of the Exascale Computing Project (ECP) ; and
  • Participants will discuss what will be required to integrate various research activities/projects to provide the data management, data analytics and visualization technology required to support scientific discovery at extreme scale.

Venue

Hyatt Bethesda
1 Bethesda Metro Center
Bethesda, MD 20814
United States

Point of Contact

Lucy Nowell

Involved Programs/Projects

Lucy Nowell
Lucy Nowell

Agenda

March 14, 2017                

7:00 – 8:00       Breakfast         

8:00 – 8:05       Introductions: Lucy Nowell and Scott Klasky

8:05 – 8:20     Welcome and ASCR Update – Barbara Helland, Director, Advanced Scientific Computing Research

8:20 – 8:30       Guidance for the Day – Lucy Nowell

8:30 – 9:15       Invited Talk: Exascale Computing Project (ECP) Applications’ Needs – Doug Kothe

9:15 – 10:20     Panel: Reducing the time for data movement to and from storage

Moderator: Lavanya Ramakrishnan

Panelists:

  • Han-Wei Shen, Extreme-Scale Distribution-Based Data Analysis
  • Scott Klasky, SIRIUS: Science-driven Data Management for Multi-tiered Storage
  • Surendra Byna, In situ Indexing and Query Processing of AMR Data
  • Kwan-Liu Ma, Supporting Co-Design of Extreme-Scale Systems with In Situ Visual Analysis of Event-Driven Simulations

10:20 – 10:50    Break

10:50 – 11:55    Panel: Exploratory visualization tools and techniques

Moderator: Adolfy Hoisie

Panelists:

  • Wes Bethel, Towards Exascale: High Performance Visualization and Analytics
  • Kenneth Moreland, XVis: Visualization for the Extreme-Scale Scientific-Computation Ecosystem
  • Hank Childs (Early Career Research Program [ECRP]): Data Exploration at the Exascale
  • James Ahrens, Exploration of Exascale In Situ Visualization and Analysis Approaches

11:55 – 1:15   Working Lunch, Invited Talk: Emerging Supercomputer Architectures – Katherine Riley

1:15 – 2:20    Panel Understanding and managing resilience

Moderator: Terry Jones

Panelists:

  • Mattan Erez (ECRP), Validating Extreme-scale Resilience with Veracity
  • Christian Engelmann (ECRP), Characterizing Faults, Errors, and Failures in Extreme-scale Systems
  • William Kramer, Holistic Measurement Driven Resilience
  • Remzi Arpaci-Dusseau, Modeling Impacts of Resilience Architectures for Extreme-Scale Storage Systems

2:20 – 3:00 Invited Talk: ECP Software Technology Update and Discussion – Rajeev Thakur

3:00 – 4:30 Break and Posters

4:30 – 5:30 Breakouts

1A: Ramakrishnan, Shen, Klasky, Byna, Ma

1B: Hoisie, Bethel, Moreland, Childs, Ahrens

1C: Jones, Erez, Engelmann, Kramer, Arpaci-Dusseau

5:30 – 6:00 Invited Talk: Introduction to DOE (Code) {} Repository – Jay Billings

Dinner on your own

 

 

March 15, 2017             

7:00 – 8:00       Breakfast         

8:00 – 8:10       Regroup and guidance for the day Lucy Nowell

8:10 – 9:10       Invited Talk: Data Provenance – Kerstin Kleese van Dam

9:10 – 10:00     Invited Talk:  The Future of Scientific Workflows Workshop – Tom Peterka

10:00 – 10:30    Break

10:30 – 11:35    Panel: Scientific Workflow

Moderator: Galen Shipman

Panelists:

  • Lavanya Ramakrishnan, Usable Data Abstractions for Next-Generation Scientific Workflows
  • Patrick McCormick, A Unified Data-Driven Approach for Programming In Situ Analysis and Visualization
  • Tom Peterka, High Performance Decoupling of Tightly-Coupled Data Flows
  • Sriram Krishnamoorthy (ECRP), Whole-program Adaptive Error Detection and Mitigation

11:35 – 1:00     Working Lunch, Invited Talk: Workshop on Management, Analysis and Visualization of Experimental and Observational Data – Wes Bethel             

1:00 – 2:05       Panel: Understanding, Optimizing, and Programming for Power

Moderator: Remzi Arpaci-Dusseau

Panelists:

  • David Rogers, Optimizing the Energy Usage and Cognitive Value of Extreme Scale Data Analysis Approaches
  • Hank Childs (ECRP), Optimizing Power Usage For Data-Intensive Workflows and Algorithms on Modern Computing Architectures
  • Adolfy Hoisie, Beyond the Standard Model
  • Rami Melhem, Lazy Shadowing - An Adaptive, Power-Aware Resiliency Framework for Exascale Computing

2:05-3:10          Panel: Modeling and Understanding   

Moderator:  Kenneth Moreland

Panelists:

  • Greg Eisenhauer, Performance Understanding and Analysis for Exascale Data Management Workflows
  • Kevin Barker, Evaluating Exascale Execution Models
  • Chris Carothers, Co-Design of Exascale Storage and Science Data Facilities (CODES)
  • Nicholas Wright, A Framework for Holistic I/O Workload Characterization

3:10 – 4:40       Break and Posters

4:40 – 5:40       Breakout

2A: Shipman, Ramakrishnan, McCormick, Peterka, Krishnamoorthy

2B: Arpaci-Dusseau, Rogers, Childs, Hoisie, Melhem

2C: Moreland, Eisenhauer, Barker, Carothers, Wright

5:40 – 6:15       Program Management Expectations – Lucy Nowell

 

 

 

March 16, 2017

7:00 – 8:00       Breakfast 

8:00 – 8:10     Regroup and guidance for the day – Lucy Nowell

8:10 – 9:15       Panel: Data Analysis

Moderator:  Christian Engelmann

Panelists:

  • Surendra Byna, Proactive Data Containers for Scientific Storage
  • Daniela Ushizima (ECRP), Images Across Domains, Experiments, Algorithms and Learning
  • Chandrika Kamath, IDEALS: Improving Data Exploration and Analysis at Large Scale
  • Alok Choudhary, Scalable, In-situ Data Clustering Data Analysis for Extreme Scale Scientific Computing

9:15 – 9:45       Break  

9:45 – 10:50     Panel: In situ visualization and infrastructure

Moderator: William Kramer

Panelists:

  • Patrick McCormick, Domain-Specific Languages for in situ Data Analysis and Visualization on Emerging Architectures
  • Wes Bethel, Scalable Analysis Methods and In Situ Infrastructure for Extreme Scale Knowledge Discovery
  • Florin Rusu (ECRP), Scalable and Energy-Efficient Methods for Interactive Exploration of Scientific Data
  • Matthew Wolf, RSVP: Runtime System for I/O staging in support of Voluminous in-situ Processing of extreme scale data

10:50 – 11:55    Panel: Exascale storage challenges

Moderator:  Greg Eisenhauer

Panelists:

  • John Wu, Scientific Data Services (SDS) – Autonomous Data Management on Exascale Infrastructure
  • Rob Ross, A Software Defined Storage Approach to Exascale Storage Services
  • Elena Pourmal, ExaHDF5: Advancing HPC I/O to Enable Scientific Discovery
  • Terry Jones, UNITY: Unified Memory and Storage Space

11:55 – 1:30     Working Lunch & Breakout 3

3A: Engelmann, Byna, Ushizima, Kamath, Choudhary

3B: Kramer, McCormick, Bethel, Rusu, Wolf

3C: Eisenhauer, Wu, Ross, Pourmal, Jones

1:30 – 3:00       Break and Posters

3:00 – 4:45       Moderators report back from breakouts

  • 1A Ramakrishnan , 1B Hoisie, 1C Jones, 2A Shipman, 2B  Arpaci-Dusseau, 2C: Moreland, 3A Engelmann, 3B Kramer, 3C Eisenhauer

4:45 – 5:00       Wrap-up and closing remarks – Lucy Nowell

Adjourned – Safe Travels!

Invited Talks

Abstract: ASCR  CS Program Guidelines for PIs
Abstract: ECP Applications
Abstract: ECP Software Technology Update
Abstract: Emerging Supercomputer Architectures
Abstract: How to Build a DOE Office of Science Highlight Slide
Abstract: Virtually all modern computational and experimental science projects rely on software technology for gaining insight into data of increasing size and...
Abstract: The talk will discuss the basics of provenance, and its traditional and novel applications in support of high end computational science. A particular...
Abstract: The traditional approach for scientific visualization and analysis is to first write data to persistent storage for subsequent, post hoc use. This...
Abstract: Workshop on the Future of Scientific Workflows

Resources

Quad Charts

Quad chart for the Whole Program Adaptive Error Detection and Mitigation (AEDAM) project.  
In this project, we are developing a holistic AMR-capable analytics framework capable of supporting simulation analysis toward extreme-scale. Our...
This project sets the stage for establishing an integrated approach to modeling performance and power, which will take us “beyond the standard”...
Abstract Department of Energy (DOE) leadership computing facilities are in the process of deploying extreme-scale high-performance computing (HPC)...
Cinema Quad Chart
The goal of the CODES project is use highly parallel simulation to explore the design of exascale interconnect and storage architectures and...
The objective of this project is to support the co-design of extreme scale system architectures for compute- and data-centric science through...
The challenges of extreme scale computing systems exist across multiple dimensions including architecture, energy constraints, memory scaling,...
DataExplor Quad Chart for March 2017 CS PI meeting
Decaf quad chart
Quad-chart describing the challenges, novel ideas, impact, and significant outcomes for the Evaluating Execution Models project.  This project aims...
Modern computational and experimental sciences face a major challenge in coping with the sheer volume and complexity of data produced. Storing,...
This quad chart summarizes our accomplishemnts for the 2017 period 
Quad Chart for Holistic I/O March 2017 PI meeting
DOE research across a myriad of science domains are increasingly reliant on image-based data from experiments, but domain scientists continue to...
The inherent instability of extreme-scale computing systems puts into question the viability of traditional fault-tolerance methods and calls for a...
The objective of this project is to develop and demonstrate an approach for composing storage services that provides new levels of functionality,...
Performance Understanding and Analysis for Exascale Data Management Workflows
Quad chart for PI meeting
OptPowerUsage Quad Chart for March 2017 CS PI Meeting.
Targeting at new paradigms for SSIO in the exascale era, we are investigating novel “object-centric” data abstractions and storage mechanisms that...
This project is tackling a set of R&D research problems to produce in situ software technology and tools that will help DOE computational science...
DOE mission critical applications are expected to generate exabytes of data per second in a few years. Capturing, storing, and accessing the most...
SEEM Quad Chart Mar 2017
A quad-chart on our project about data durability.
Virtually all modern computational and experimental science projects rely on software technology for gaining insight into data of increasing size and...
The goal of this project is to facilitate next-generation data analyses by developing new tools that make it easier to process large amounts of...
Impact: •Enables domain scientist to focus on their specialty rather than requiring them to become experts on memory architectures. •Enables target...
As semiconductor manufacturing technology continues to scale to smaller dimensions and as power-consumption of processor chips continues to increase...
The XVis project brings together the key elements of research to enable scientific discovery at extreme scale. Scientific computing will no longer be...

Posters

DOE mission critical applications are expected to generate exabytes of data per second in a few years. Capturing, storing, and accessing the most...
DOE mission critical applications are expected to generate exabytes of data per second in a few years. Capturing, storing, and accessing the most...
AEDAM proposes a comprehensive approach to error detection and mitigation for scientific applications that combines configurable error detectors, a...
In this project, we are developing a holistic AMR-capable analytics framework capable of supporting simulation analysis toward extreme-scale. Our...
Department of Energy (DOE) leadership computing facilities are in the process of deploying extreme-scale high-performance computing (HPC) systems...
Cinema Poster March 2017
The goal of the CODES project is use highly parallel simulation to explore the design of exascale interconnect and storage architectures and...
The objective of this project is to support the co-design of extreme scale system architectures for compute- and data-centric science through...
The challenges of extreme scale computing systems exist across multiple dimensions including architecture, energy constraints, memory scaling,...
DataExplor poster for March 2017 CS PI meeting
Decaf poster
This poster describes significant outcomes of the Evaluating Execution Models project, including methodologies for empirical analysis and modeling of...
Modern computational science faces a number of challenges on the path to exascale. Simulation codes are producing massive amounts of data that need...
This poster updates our reserch progress for 2017. 
Poster for the HMDR Project
DOE research across a myriad of science domains are increasingly reliant on image-based data from experiments, but domain scientists continue to...
The inherent instability of extreme-scale computing systems puts into question the viability of traditional fault-tolerance methods and calls for a...
The objective of this project is to develop and demonstrate an approach for composing storage services that provides new levels of functionality,...
Performance Understanding and Analysis for Exascale Data Management Workflows
Poster for March 2017 PI meeting
OptPowerUsage poster for March 2017 CS PI Meeting.
Targeting at new paradigms for SSIO in the exascale era, we are investigating novel “object-centric” data abstractions and storage mechanisms that...
RSVP poster
The traditional approach for scientific visualization and analysis is to first write data to persistent storage for subsequent, post hoc use. This...
SEEM Poster Mar 2017
sirius poster
A poster on the many faces of data durability.
Virtually all modern computational and experimental science projects rely on software technology for gaining insight into data of increasing size and...
The goal of this project is to facilitate next-generation data analyses by developing new tools that make it easier to process large amounts of...
The objective of this research is to design, implement and evaluate a new distributed data storage paradigm that unifies the traditionally distinct...
As semiconductor manufacturing technology continues to scale to smaller dimensions and as power-consumption of processor chips continues to increase...
The XVis project brings together the key elements of research to enable scientific discovery at extreme scale. Scientific computing will no longer be...

Handouts

Handout for the AEDAM project, prepared for use during the ASCR CS PI meeting (14-16 March 2017).  
In this project, we are developing a holistic AMR-capable analytics framework capable of supporting simulation analysis toward extreme-scale. Our...
This project sets the stage for establishing an integrated approach to modeling performance and power, which will take us “beyond the standard”...
Department of Energy (DOE) leadership computing facilities are in the process of deploying extreme-scale high-performance computing (HPC) systems...
The goal of the CODES project is use highly parallel simulation to explore the design of exascale interconnect and storage architectures and...
The objective of this project is to support the co-design of extreme scale system architectures for compute- and data-centric science through...
The challenges of extreme scale computing systems exist across multiple dimensions including architecture, energy constraints, memory scaling,...
DataExplor  handout for March 2017 CS PI meeting.
Decaf handout
The Evaluating Execution Models (EEM) project aims to develop a methodology for the quantitative assessment of novel execution models targeting...
Modern computational science faces a number of challenges on the path to exascale. Simulation codes are producing massive amounts of data that need...
Handout for March 2017 PI Meeting
Handout for the Holistic Measurement Driven Resiliency Project
One pager for PI meeting March 2017
Holistic I/O one pager
DOE research across a myriad of science domains are increasingly reliant on image-based data from experiments, but domain scientists continue to...
The inherent instability of extreme-scale computing systems puts into question the viability of traditional fault-tolerance methods and calls for a...
The traditional approach for scientific visualization and analysis is to first write data to persistent storage for subsequent, post hoc use. This...
Virtually all modern computational and experimental science projects rely on software technology for gaining insight into data of increasing size and...
The objective of this project is to develop and demonstrate an approach for composing storage services that provides new levels of functionality,...
Performance Understanding and Analysis for Exascale Data Management Workflows
OptPowerUsage handout for March 2017 CS PI Meeting
Targeting at new paradigms for SSIO in the exascale era, we are investigating novel “object-centric” data abstractions and storage mechanisms that...
Building on  the  early successes  of the ADIOS   framework,  the   design   and  realization   of   the  our envisioned  process  flows   will  be ...
DOE mission critical applications are expected to generate exabytes of data per second in a few years.  Capturing, storing, and accessing the most...
SEEM Handout Mar 2017
one page handout for the Sirius project
A short paper handout on our work on data durability.
The goal of this project is to facilitate next-generation data analyses by developing new tools that make it easier to process large amounts of...
The UNITY Project is developing smart runtime system technology to enable worry-free application development in an age characterized by increasingly...
As semiconductor manufacturing technology continues to scale to smaller dimensions and as power-consumption of processor chips continues to increase...
The XVis project brings together the key elements of research to enable scientific discovery at extreme scale. Scientific computing will no longer be...

Presentations

Slides on the AEDAM project to be presented at the CS PI meeting (14-16 March 2017).  
In this project, we are developing a holistic AMR-capable analytics framework capable of supporting simulation analysis toward extreme-scale. Our...
ASCR  CS Program Guidelines for PIs
Department of Energy (DOE) leadership computing facilities are in the process of deploying extreme-scale high-performance computing (HPC) systems...
This presentation provides an update for the Cinema project.
The goal of the CODES project is use highly parallel simulation to explore the design of exascale interconnect and storage architectures and...
The objective of this project is to support the co-design of extreme scale system architectures for compute- and data-centric science through...
DataExplorer presentation for March 2017 CSPI meeting
DataExplor presentation for March 2017 CS PI Meeting (no animations)
Decaf highlight slides
ECP Applications
ECP Software Technology Update
Emerging Supercomputer Architectures
The Evaluating Execution Models (EEM) project aims to develop a methodology for the quantitative assessment of novel execution models targeting...
Modern computational science faces a number of challenges on the path to exascale. Simulation codes are producing massive amounts of data that need...
Update of Exploration of Exascale In Situ Visualization and Analysis Approaches project
This is the presentation slides for the 3/14-16 meeting.
HMDR Presentation
How to Build a DOE Office of Science Highlight Slide
DOE research across a myriad of science domains are increasingly reliant on image-based data from experiments, but domain scientists continue to...
The inherent instability of extreme-scale computing systems puts into question the viability of traditional fault-tolerance methods and calls for a...
The inherent instability of extreme-scale computing systems puts into question the viability of traditional fault-tolerance methods and calls for a...
Virtually all modern computational and experimental science projects rely on software technology for gaining insight into data of increasing size and...
The objective of this project is to develop and demonstrate an approach for composing storage services that provides new levels of functionality,...
A short talk describing the MONA project, presented at the PI meeting in Bethesda, MD - Mar 2017.
The talk will discuss the basics of provenance, and its traditional and novel applications in support of high end computational science. A particular...
Highlight presentation for PI meeting
OptPowerUsage Presentation March 2017 CSPI Meeting
Targeting at new paradigms for SSIO in the exascale era, we are investigating novel “object-centric” data abstractions and storage mechanisms that...
Mathew Wolf's presentation of RSVP, and how the research ideas turn into software used in the HPC community.
DOE mission critical applications are expected to generate exabytes of data per second in a few years. Capturing, storing, and accessing the most...
SEEM Presentation Mar 2017
The traditional approach for scientific visualization and analysis is to first write data to persistent storage for subsequent, post hoc use. This...
Science-driven Data Management for Multi-tiered Storage
A talk about our work on data durability.
A talk describing our work on data durability.
•Problem –The simple bifurcated memory hierarchy of the 1950s has evolved into a much more complex hierarchy while interfaces have remained...
As semiconductor manufacturing technology continues to scale to smaller dimensions and as power-consumption of processor chips continues to increase...
Workshop on the Future of Scientific Workflows
The XVis project brings together the key elements of research to enable scientific discovery at extreme scale. Scientific computing will no longer be...

Breakout Notes

Posted by David Rogers on 04/23/2017
Summary of the discussions from panels and breakout sessions, including executive summary