<html dir="ltr">
<head>
<meta http-equiv="Content-Type" content="text/html; charset=Windows-1252">
<style type="text/css" id="owaParaStyle"></style>
</head>
<body fpstyle="1" ocsi="0">
<div style="direction: ltr;font-family: Tahoma;color: #000000;font-size: 10pt;">
<div><b>Joining with the other DSN workshops, we have extended the deadline for paper submissions </b></div>
<div><b>by 1 week to March 14, 2014. Author notifications have also been extended to April 11, 2014.</b></div>
<div><br>
</div>
<div>CALL FOR PAPERS</div>
<div>4th International Workshop on Fault-Tolerance for HPC at Extreme Scale </div>
<div>(FTXS 2014)</div>
<div><br>
</div>
<div>In conjunction with</div>
<div>The 44th Annual IEEE/IFIP International Conference on </div>
<div>Dependable Systems and Networks (DSN 2014)</div>
<div>Atlanta, Georgia, USA on June 23-26, 2014</div>
<div><br>
</div>
<div>WORKSHOP MOTIVATION</div>
<div>For the HPC community, a new scaling in numbers of processing elements</div>
<div>has superseded the historical trend of Moore's Law scaling in processor</div>
<div>frequencies. This progression from single core to multi-core and</div>
<div>many-core will be further complicated by the community's imminent</div>
<div>migration from traditional homogeneous architectures to ones that are</div>
<div>heterogeneous in nature. As a consequence of these trends, the HPC</div>
<div>community is facing rapid increases in the number, variety, and</div>
<div>complexity of components, and must thus overcome increases in aggregate</div>
<div>fault rates, fault diversity, and complexity of isolating root cause.</div>
<div><br>
</div>
<div>Recent analyses demonstrate that HPC systems experience simultaneous</div>
<div>(often correlated) failures. In addition, statistical analyses suggest</div>
<div>that silent soft errors cannot be ignored anymore, because the increase</div>
<div>of components, memory size and data paths (including networks) make the</div>
<div>probability of silent data corruption (SDC) non-negligible. The HPC</div>
<div>community has serious concerns regarding this issue and application</div>
<div>users are less confident that they can rely on a correct answer to their</div>
<div>computations. Other studies have indicated a growing divergence between</div>
<div>failure rates experienced by applications and rates seen by the system</div>
<div>hardware and software. At Exascale, some scenarios project failure rates</div>
<div>reaching one failure per hour. This conflicts with the current</div>
<div>checkpointing approach to fault tolerance that requires up to 30 minutes</div>
<div>to restart a parallel execution on the largest systems. Lastly,</div>
<div>stabilization periods for the largest systems are already significant,</div>
<div>and the possibility that these could increase in length is of great</div>
<div>concern. During the Approaching Exascale report at SC11, DOE program</div>
<div>managers identified resilience as a black swan - the most difficult</div>
<div>under-addressed issue facing HPC.</div>
<div><br>
</div>
<div>OPEN QUESTIONS</div>
<div>What does the fault-tolerance community need to do in order to be</div>
<div>prepared to face the challenges of extreme scale computing? What is</div>
<div>needed to keep applications with billions of threads of parallelism up</div>
<div>and running on systems that fail tens of times per day? As models</div>
<div>predict less than 50% efficiency of traditional checkpoint/restart</div>
<div>methods on future systems, are we ready to pay the cost of full</div>
<div>redundancy, effectively performing redundant multi-threading (RMT)</div>
<div>across entire systems? Do we even have the infrastructure necessary to</div>
<div>implement an RMT strategy?</div>
<div><br>
</div>
<div>How is the supercomputing community going to efficiently isolate</div>
<div>failures on enormously complex systems? Is it realistic to understand</div>
<div>these systems in such a way that some failure could be predicted with</div>
<div>enough accuracy and anticipation to trigger useful failure avoidance</div>
<div>actions? What can the community do to protect applications from SDC in</div>
<div>memory and logic? To what extent should users and programmers be</div>
<div>involved in managing faults? What are the most promising self-healing</div>
<div>numerical methods? Is there an emerging framework for fault management</div>
<div>at extreme scale?</div>
<div><br>
</div>
<div>GOALS</div>
<div>The goals of this workshop are to consider these complex questions, to</div>
<div>discuss the unique limitations that extreme scale and complexity impose</div>
<div>on traditional methods of fault-tolerance, and to explore new strategies</div>
<div>for dealing with those challenges.</div>
<div><br>
</div>
<div>PAPER SUBMISSIONS</div>
<div>Submissions are solicited in the following categories:</div>
<div>* Regular papers presenting innovative ideas improving the state of the art.</div>
<div>* Experience papers discussing the issues seen on existing extreme-scale </div>
<div> systems, including some form of analysis and evaluation.</div>
<div>* Extended abstracts proposing disruptive ideas in the field, including </div>
<div> some form of preliminary results</div>
<div><br>
</div>
<div>Submissions shall be sent electronically, must conform to IEEE</div>
<div>conference proceedings style and should not exceed six pages including</div>
<div>all text, appendices, and figures.</div>
<div><br>
</div>
<div>TOPICS</div>
<div>Assuming hardware and software errors will be inescapable at extreme</div>
<div>scale, this workshop will consider aspects of fault tolerance peculiar</div>
<div>to extreme scale that include, but are not limited to: </div>
<div>* Quantitative assessments of cost in terms of power, performance, and </div>
<div> resource impacts of fault-tolerant techniques, such as checkpoint</div>
<div> restart, that are redundant in space, time or information</div>
<div>* Novel fault-tolerance techniques and implementations of emerging</div>
<div> hardware and software technologies that guard against silent data</div>
<div> corruption (SDC) in memory, logic, and storage and provide end-to-end</div>
<div> data integrity for running applications; Studies of hardware / software</div>
<div> tradeoffs in error detection, failure prediction, error preemption, and</div>
<div> recovery</div>
<div>* Advances in monitoring, analysis, and control of highly complex systems</div>
<div>* Highly scalable fault-tolerant programming models</div>
<div>* Metrics and standards for measuring, improving and enforcing the need </div>
<div> for and effectiveness of fault-tolerance</div>
<div>* Failure modeling and scalable methods of reliability, availability, </div>
<div> performability and failure prediction for fault-tolerant HPC systems</div>
<div>* Scalable Byzantine fault tolerance and security from single-fault and </div>
<div> fail-silent violations</div>
<div>* Benchmarks and experimental environments, including fault-injection </div>
<div> and accelerated lifetime testing, for evaluating performance of </div>
<div> resilience techniques under stress</div>
<div>* Frameworks and APIs for fault tolerance and fault management.</div>
<div><br>
</div>
<div>IMPORTANT DATES</div>
<div>Submission of papers: March 14, 2014</div>
<div>Author notification: April 11, 2014</div>
<div>Camera ready papers: April 2014</div>
<div>Workshop: June 23rd, 2014</div>
<div><br>
</div>
<div>WORKSHOP ORGANIZERS</div>
<div>Nathan DeBardeleben - Los Alamos National Laboratory</div>
<div>Franck Cappello Argonne National Laboratory and the University of</div>
<div> Illinois at Urbana-Champaign</div>
<div>Robert Clay Sandia National Laboratories</div>
<div><br>
</div>
<div>PROGRAM COMMITTEE</div>
<div>Rob Aulwes Los Alamos National Laboratory</div>
<div>Aurιlien Bouteiller University of Tennessee Knoxville</div>
<div>Greg Bronevetsky - Lawrence Livermore National Laboratory</div>
<div>John Daly - Department of Defense</div>
<div>Christian Engelmann Oak Ridge National Laboratory</div>
<div>Kurt Ferreira Sandia National Laboratories</div>
<div>Ana Gainaru University of Illinois at Urbana-Champaign</div>
<div>Leonardo Bautista Gomez Tokyo Institute of Technology</div>
<div>Hideyuki Jitsumoto The University of Tokyo</div>
<div>Zhiling Lan Illinois Institute of Technology</div>
<div>Naoya Maruyama RIKEN Advanced Institute for Computational Science</div>
<div>Kathryn Mohror Lawrence Livermore National Laboratory</div>
<div>Bogdan Nicolae IBM Research Ireland</div>
<div>Rolf Riesen IBM Research Ireland</div>
<div>Yve Robert - ENS Lyon</div>
<div>Thomas Ropars - EPFL</div>
<div>Stephen Scott Tennessee Tech University and Oak Ridge National</div>
<div>Laboratory</div>
<div>Vilas Sridharan AMD, Inc.</div>
<div>Abhinav Vishnu - Pacific Northwest National Laboratory</div>
<div>Roel Wuyts - Intel ExaScience Lab</div>
<div><br>
</div>
<div>MORE INFORMATION</div>
<div>See https://sites.google.com/site/ftxsworkshop/home/ftxs2014 and</div>
<div>http://2014.dsn.org/ for more information.</div>
<div>
<div class="BodyFragment"></div>
</div>
</div>
</body>
</html>