Clouds, Grids, and Clusters are three different computational paradigms with the potential to support High Performance Computing (HPC) and enterprise IT infrastructure. Currently, they consist of hardware, management, and usage models particular to different computational regimes (e.g., high performance cluster systems designed to support tightly coupled scientific simulation codes typically utilize high-speed interconnects and commercial cloud systems designed to support software as a service (SAS) typically do not). However, in order to support HPC, all must at least utilize large numbers of resources and hence effective HPC in any of these paradigms must address the same issue of resiliency at a very large-scale.
Recent trends in high-performance computing (HPC) systems have clearly indicated that future increases in performance, in excess of those resulting from improvements in single-processor performance, will be achieved through corresponding increases in system scale, i.e., using a significantly larger component count. As the raw computational performance of the world's fastest HPC systems increases from today's current multi-petascale to next-generation exascale capability and beyond, their number of computational, networking, and storage components will grow from the ten-to-one-hundred thousand compute nodes of today's systems to several hundreds of thousands of compute nodes in the foreseeable future. This substantial growth in system scale, and the resulting component count, poses a challenge for HPC system and application software with respect to reliability, availability and serviceability (RAS).
The expected total component count of these HPC systems calls into questions many of today's HPC RAS assumptions. Although the mean-time to failure (MTTF) for each individual component, e.g., processor, memory module, and network interface, may be above typical consumer product standard, the probability of failure for the overall system scales proportionally to the number of interdependent components and their combined probabilities of failure. Thus, the enormous number of individual components results in a much lower system mean-time to failure (SMTTF), causing more frequent system-wide interruptions than displayed by current HPC systems. This effect is not limited to hardware components, but also extends to software components, e.g., operating system, system software, and applications. Although software components do not show less reliability with increasing age like hardware components, they do contain other sources of failures, such as design and implementation errors. Furthermore, the health of software components also involves resource utilization, such as processor, memory and network usage.
To address the issue of computing resiliency, fault tolerance and high availability have become critical research topics. The goal of this workshop is to bring together the community in an effort to facilitate resilient HPC in each of these three computational paradigms -- Clouds, Grids, and Clusters. Their respective differences in architecture, management, and usage models may lend themselves to different approaches to resiliency. Knowledge of these approaches in one may be used to enable resiliency in the others or to define new usage models to enable HPC. This workshop targets fundamental solutions and issues in resiliency for HPC.
Authors are invited to submit papers electronically in English in PDF format. Submitted manuscripts should be structured as technical papers and may not exceed 12 pages, including figures, tables and references, using Springer's Lecture Notes in Computer Science (LNCS) format at <http://www.springer.com/computer/lncs?SGWID=0-164-6-793341-0>. Submissions should include abstract, key words and the e-mail address of the corresponding author. Papers not conforming to these guidelines may be returned without review. All manuscripts will be reviewed and will be judged on correctness, originality, technical strength, significance, quality of presentation, and interest and relevance to the conference attendees. Submitted papers must represent original unpublished research that is not currently under review for any other conference or journal. Papers not following these guidelines will be rejected without review and further action may be taken, including (but not limited to) notifications sent to the heads of the institutions of the authors and sponsors of the conference. Submissions received after the due date, exceeding length limit, or not appropriately structured may also not be considered. The proceedings will be published in Springer's LNCS as post-conference proceedings. At least one author of an accepted paper must register for and attend the workshop for inclusion in the proceedings. Authors may contact the workshop program chairs for more information.
Topics of interest include, but are not limited to: