Dataflow Computing for Data-Intensive Applications
503-1-001 - Council Chamber
Maxeler Technologies provides Maximum Performance Computing (MPC) based on a dataflow model of computation. Combined with a multi-disciplinary approach, these dataflow solutions are in production, reducing power consumption and data center space for a wide range of applications. Typically, our dataflow solutions utilize thousands of arithmetic units on a chip to outperform top end microprocessors by 20x-40x in computations per cubic foot and computations per Watt, and our analysis suggests that there is strong potential to do a lot better in future. Dataflow is particularly suitable for highly demanding applications such as: finite element and finite difference PDE solvers on structured and unstructured grids, Monte Carlo methods, multi-dimensional optimization problems, real-time data processing, and even sparse matrix solvers. At the core of the approach, we are building computers to match the problem, rather than optimizing algorithms to standard microprocessors.
Given the success in Earth Sciences, Quantitative Finance and Electronic Trading and the availability of a user-friendly dataflow programming environment, we believe that this novel approach to computing could potentially lead to scientific breakthroughs in data-intensive branches of science.
In spite of their spectacular speed, FPGAs, GPGPUs and also multicore architectures obey to a restrictive computation paradigm that makes less attractive the numerical solution of many algebraic problems. Hence the real challenge for the developer is the reduction of a mathematical model to a sequence of computational tasks that perfectly fit the paradigm supported by these extreme architectures. In a second part, we illustrate this concept for time imaging algorithms of general use in oil industry.
About the speakers
Dr. Oskar Mencer, Maxeler Technologies
Prior to founding Maxeler, Oskar was Member of Technical Staff at the Computing Sciences Center at Bell Labs in Murray Hill, leading the effort in "Stream Computing". He joined Bell Labs after receiving a PhD from Stanford University. Besides driving Maximum Performance Computing (MPC) at Maxeler, Oskar is Consulting Professor in Geophysics at Stanford University and he is also affiliated with the Computing Department at Imperial College London, having received two Best Paper Awards, an Imperial College Research Excellence Award in 2007 and a Special Award from Com.sult in 2012 for "revolutionising the world of computers".
Dr. Ernesto Bonomi, CRS4
Director at CRS4 of three highly-motivated research groups (Imaging and Numerical Geophysics, Environmental Sciences, and Process Engineering and Combustion) which include about 20 research staff, developing and applying numerical simulation models and data analysis tools for earth exploration and seismic imaging, hydrology and territorial planning, meteorology, and clean combustion processes.
Head of the developers? team providing innovative industrial acoustic imaging tools based on original mathematical developments and on efficient data-driven parallel algorithms giving rise to the implementation of very-large-scale 3D numerical applications used routinely in production on HPC platforms and, more recently, on acceleration hardware (FPGA and GPU).