INTERNATIONAL ADVANCED RESEARCH WORKSHOP ON HIGH PERFORMANCE COMPUTING
June 21 - June 25
The tools and techniques of High Performance Computing (HPC) have gained broad acceptance in wide areas of research and industry due to sustained progress in computational hardware and software technologies, ranging from hybrid CPU/GPU systems, multicore and distributed architectures, and virtualization, to relatively new paradigms such as cloud computing, explosive growth of the use of Artificial Intelligence (AI) techniques in myriad applications, and advances in quantum computer realizations. At the same time, the extremely fast pace of the field introduces new challenges in technological, intellectual, ethical and even political areas that must be addressed to continue to enable wider acceptance, implementation, and ultimately societal impact of high performance computing technologies, applications, and paradigms.
The main aim of this workshop is to present and debate advanced topics, open questions, current and future developments, and challenging applications related to advanced high-performance distributed computing and data systems, encompassing implementations ranging from traditional clusters to warehouse-scale data centers, and with architectures including hybrid, multicore, distributed, cloud models, and systems targeted for AI applications. In addition, quantum computing has captured intense and widespread interest in the last two years, in large part due to the deployment of several systems with diverse architectures. This workshop will provide a forum for exploration of both challenges and synergies that might arise from exchange of ideas across the many aspects of HPC and its applications.
The rapid uptake of AI methods to tackle myriad applications has led to rethinking of the relevant algorithms and of the microarchitectures of computers that are optimized for such applications. Although machine and deep learning are the AI technologies that are in the headlines daily and flood submissions to conferences and journals, other aspects of AI are also maturing and in some cases require HPC resources.
Similarly, the growing deployment of quantum computers, some of which are accessible to the open research community, is spurring experimentation with reformulation of problems, algorithms, and programming techniques for such computers. Quantum sensing and quantum communication are also beginning to have physical instantiations.
The importance of Cloud Computing in HPC continues to grow. We are seeing more and more cloud testbeds and production facilities that are used by government agencies, industry and academia. . Commercial cloud service providers like Amazon Web Services, Bull extreme factory, Fujitsu TC Cloud, Gompute, Microsoft Azure, Nimbix, Nimbula, Penguin on Demand, UberCloud, and many more are now offering HPC-focused infrastructure, platform, and application services. However, careful application benchmarking of different cloud infrastructures still have to be performed to find out which HPC cloud architecture is best suited for a specific application.
From an application standpoint, many of the most widely used application codes have undergone many generations of adaptation as new architectures have emerged, from vector to MPP to cluster to cloud, and more recently to multicore and hybrid. As exascale systems move toward millions of processing units the interplay between system and user software, compilers and middleware, even programmer and run-time environment must be reconsidered. For example, how much resilience and fault-tolerance can, or should, be embedded transparently in the system versus exposed to the programmer? Perhaps even greater challenges arise from the complexity of applications, which are increasingly multi-scale and multi-physics and are built from hundreds of building blocks, and from the difficulty of achieving portability across traditional architectures.
Finally, discussions and presentations related to emerging and strategically challenging application areas will also be an important part of the workshop. A special emphasis will be given to the potential of computational modeling and advanced analytics related to urban systems, including the associated diverse data sources and streams. Similarly, the challenges of data integration and use for new types of data sources such as the Internet of Things, will be examined. These and other new application areas enabled by new sources of data, including IoT and sensor networks, represent an interesting new set of HPC challenges.
Summarizing, the aim of this special workshop is to shed some light on key topics in advanced high performance computing systems and, in particular, to address the aforementioned contemporary scheduling, scaling, fault tolerance, and emerging application topics. The four and a half day program of this workshop will include roughly fifty invited talks and associated panels by experts in the field.