Demonstrating a Pre-Exascale, Cost-Effective Multi-Cloud Environment for Scientific Computing: Producing a fp32 ExaFLOP hour worth of IceCube simulation data in a single workday.
Igor SfiligoiDavid SchultzBenedikt RiedelFrank WürthweinSteve BarnetVladimir BrikPublished in: PEARC (2020)
Keyphrases
- scientific computing
- cost effective
- simulation data
- high performance computing
- low cost
- file system
- cost effectiveness
- distributed memory
- numerical simulations
- workflow management
- massively parallel
- data center
- computer systems
- fault tolerance
- computing systems
- gene expression data
- energy efficiency
- statistical analysis and data mining
- database
- data sets