Many organizations have been running high-performance computing and artificial intelligence workloads for years, almost exclusively using on-premises infrastructure to meet their various needs. However, organizations that want to use a public cloud provider for their HPC and AI workloads have typically been stymied by the lack of scalability, since demands can vary widely over short periods of time. This paper looks at how these systems are evolving, how technologies like InfiniBand are enabling new levels of cloud elasticity to meet the myriad emerging use cases, and how a unique partnership between a cloud provider and silicon vendor is stepping up to meet this long unfilled need.
Cette ressource est disponible en English.
Date de publication : 2021-06-25