Novo‐G#: a multidimensional torus‐based reconfigurable cluster for molecular dynamics

Abstract

Summary Molecular dynamics (MD) is a large‐scale, communication‐intensive problem that has been the subject of high‐performance computing research and acceleration for years. Not surprisingly, the most success in accelerating MD comes from specialized systems such as the Anton machine. In this paper, we describe Novo‐G# (novo‐jee‐sharp), a multi‐node reconfigurable system designed for the acceleration of communication‐intensive scientific problems in general, and MD in particular. This system provides a high‐bandwidth, low‐latency 3D torus network to allow direct communication between kernels running on multiple field‐programmable gate arrays. We also present a performance model for Novo‐G# running the 3D Fast Fourier Transform (FFT) kernel that forms the core of MD simulations. We validate the model against published Anton performance data and through initial hardware experiments on Novo‐G#. Finally, through simulation studies, we show that this system at scale performs better than specialized systems like Anton and outperforms established CPU‐based clusters like Blue Gene/Q by an order of magnitude for the 3D FFT kernel, with greater flexibility and lower costs. Copyright © 2015 John Wiley & Sons, Ltd.

High Performance Communication on Reconfigurable Clusters

“…Network programmability. The Novo-G# supports table-based routing which facilitates static scheduling to balance load and minimize collisions, which is especially useful for 3D FFTs (Young et al, 2009;Sheng et al, 2014;Lawande et al, 2016) and for routing collectives in MD (Grossman et al, 2015).…”

“…The input to a 3D FFT an N 3 cube of points. For fine-grained parallel implementations (Lawande et al, 2016;Sheng et al, 2014;Young et al, 2009), the 3D FFT is decomposed into 1D FFTs which are computed successively in each dimension. Each of the three dimensions requires N 2 N-point 1D FFTs for a total of 3N 2 1D FFT calculations.…”

“…The vendor IPs are extremely efficient being maximally tuned to the particular target device. The IPs used here (Altera, 2014a) For clusters there is an analogous data ordering problem that is solved by performing a cluster-wide transpose (see, e.g., (Lawande et al, 2016;Sheng et al, 2014;Young et al, 2009)). The latency of the 3D FFT, therefore, has an additional term that is a function of communication performance, which, in turn, is a function of network and the number of processors P .…”

Heterogeneous cluster architectures and applications

“…The five papers contained in this special issue cover very different areas. The titles of the papers are the following: MPI and UPC Broadcast, Scatter and Gather Algorithms in Xeon Phi A New Auction Based Scheduler for Heterogeneous Systems with Moldable Generic Resources Support Gaspar: A Compositional Aspect‐oriented Approach for Cluster Applications Novo‐G#: A Multidimensional Torus‐based Reconfigurable Cluster for Molecular Dynamics The DEEP Project: An alternative approach to heterogeneous cluster‐computing in the many‐core era …”

“…In the fourth paper, the authors show how they use a custom cluster based on reconfigurable logic to design a highly optimized solution for certain tasks, including communication‐intensive molecular dynamics . Reconfigurable logic opens up a new degree of freedom when designing computing systems; however, this huge amount of freedom comes with significant complexity increase.…”