Heterogeneous Computing Systems for Complex Scientific Discovery Workflows
Christoph Hagleitner1,a, Dionysios Diamantopoulos1,b, Burkhard Ringlein1,c, Constantinos Evangelinos2,d, Charles Johns2,e, Rong N. Chang2,f, Bruce D'Amora2,g, James A. Kahle2,h, James Sexton2,i, Michael Johnston3, Edward Pyzer-Knapp4,j and Chris Ward4,k
1IBM Research - Europe Ruschlikon, Switzerland
ahle@zurich.ibm.com
bdid@zurich.ibm.com
cngl@zurich.ibm.com
2IBM Research Yorktown Heights, USA
dcevange@us.ibm.com
ecrjohns@us.ibm.com
frong@us.ibm.com
gdamora@us.ibm.com
hjakahle@us.ibm.com
isextonjc@us.ibm.com
3IBM Research - Europe Dublin, Ireland
michaelj@ie.ibm.com
4IBM Research – Europe Warrington, Hursley UK
jEPyzerK3@uk.ibm.com
ktjcw@uk.ibm.com
ABSTRACT
With Moore's law progressively running out of steam, heterogeneous computing architectures have been powering the top supercomputers in the world for many years and are now finding broader adoption across the industry. The trend towards sustainable computing also requires domainspecific heterogeneous hardware architectures, which promise further gains in energy efficiency. At the same time, today's high performance computing applications have evolved from monolithic simulations in a single domain to multidisciplinary complex workflows. In this paper, we explore how these trends affect system design decisions and what this means for future computing system architectures.
Keywords: Heterogeneous Computing, High-Performance Computing, Accelerators, Workflow, Disaggregation.
