THE “CPU+GPU+MIC” HYBRID SUPERCOMPUTER π
Journal Title: Azerbaijan Journal of High Performance Computing - Year 2018, Vol 1, Issue 1
Abstract
The supercomputer π at Center for HPC (CHPC) of Shanghai Jiao Tong University (SJTU) was the fastest supercomputer among all the universities in China from 2013 to 2015. With a highly hybrid architecture, it is the first “CPU+GPU+MIC” supercomputer in China. It was also the largest Kepler GPU cluster in China when deployed in 2013. In addition, it firstly utilized the “FDR+DDN” combination in China. In this paper, we will introduce the design of supercomputer π.
Authors and Affiliations
James Lin, Minhua Wen
THE INFLUENCE OF EXASCALE ON RESOURCE DISCOVERY AND DEFINING AN INDICATOR
Resource discovery in distributed Exascale computing systems, in addition to managing events resulting in failures due to not finding the resource, as well as failure to perform resource discovery activities at the accep...
EXAMIG MATRIX: PROCESS MIGRATION BASED ON MATRIX DEFINITION OF SELECTING DESTINATION IN DISTRIBUTED EXASCALE ENVIRONMENTS
In traditional computing system, load balancer, interim selecting the process, determine the destination computing node based on describing Indicators process status. In distributed Exascale computing system, due to the...
APPLYING MULTIPLE MULTIDIMENSIONAL KNAPSACK PROBLEM TO DYNAMIC LOAD BALANCING IN DISTRIBUTED EXASCALE COMPUTING ENVIRONMENT
Dynamic and Interactive nature of the processes in the Distributed Exascale computing system requires the system to be able to make Load Balancing in runtime. In this paper proposed applying Multiple Multidimensional Kna...
LOAD BALANCING IN DISTRIBUTED EXASCALE COMPUTING BASED ON PROCESS REQUIREMENTS
In distributed Exascale systems, the occurrence of a dynamic and interactive nature changes the workload of the system’s computing elements. Because of this, the load balancer needs to collect information on the system s...
CONVERGENCE OF HPC AND AI: TWO DIRECTIONS OF CONNECTION PDF
This paper examines the role of HPC systems in the solution of the most AI problems, on the other hand, assesses the impact of the application of AI methods on the resolution of different tasks in distributed systems. Th...