From the 60s to the end of the 80s, we have gone from a remarkable progress in parallel computing to a significant downturn which occurred in the early 90s. This had the leaders in the field wonder whether Parallel Computing had been a dead end. However, since the mid 2000s, we have witnessed a remarkable comeback with the advent of multicore architectures – some have called it the “second spring” of parallel computing – and the trend has continued. This means that it is a good time to stop and ponder the lessons we have already learned from the history of parallel computing. One important reason is that, although much good work on parallel models of computation is ongoing, there is no commonly accepted program execution model for general-purpose parallel computing. Consequently, the so-called von Neumann bottleneck remains a fundamental challenge for parallel architecture and system model design. The “Parallel Models and Systems: Data Flow and Beyond” STC will focus on all manners of directions derived from Dataflow, combining all aspects of technology needed and exploiting the achievements reached in areas ranging from hardware, systems, algorithms to applications, making this STC the hub where all Dataflow research can be put together so as to construct an ecosystem of Dataflow-based next generation parallel models and systems, whether in a high-performance computing system or in an energy conscious embedded mobile system.
Our STC is motivated by the challenges that are coming with the second-spring of parallel computation — including but not limited to
- Demands for high-performance parallel systems with potentially extreme-scale performance;
- Computation demands from data-intensive workloads with “big data” volumes and real-time requirements, as well as intelligent data analytic processing capability;
- Demands for high energy efficiency and strong system resiliency. This would be particularly the case for embedded mobile systems.
Our STC focuses on all manners of directions derived from Dataflow, as listed below,
- Combining all needed aspects of technology, and exploiting the achievements from areas ranging from hardware, system, algorithm to applications;
- Making this STC the hub to present all Dataflow related research together, and to construct an ecosystem of Dataflow-based next generation parallel models and systems, whether it being in a high-performance computing system or an energy conscious embedded mobile system.
- Gathering Dataflow researchers to draw a whole picture of current Dataflow research progress;
- Attracting more researchers to work on Dataflow to speedup Dataflow technology development;
- Organizing Dataflow research accomplishments to point out the key milestones of Dataflow for the future five years;
- Connecting academic and industrial partners, and provide technical support for industry to build a whole chain of Dataflow solutions/products for various fields;
- Helping developing countries, e.g. China, to advance its computer technology.
Our STC aims to be an on-line community striving to explore innovative solutions of open problems in the area of parallel models of computation and systems. The members of this STC will share a common interest on Dataflow models of computation, which are a solid foundation for the exploitation of fine-grain parallelism and asynchrony in general-purpose parallel computation