GTC 2020 S21287
Presenters: Raj Rao,NVIDIA; Uday Kurkure,VMware
We’ll discuss how to combine the performance of GPUs with manageability and scalability features to maximize GPU utilization for
the visualization, HPC, and machine learning workflows that are emerging in modern data centers, which thrive on the virtualization features of vSphere and NVIDIA virtual GPUs. We’ll outline end-to-end ML for training, deploying for inferencing, and managing a production environment using vSphere and Pivotal Kubernetes Service. We’ll describe ways to deploy GPU workloads developed with ML frameworks like TensorFlow & Caffe2 by using VMware DirectPathIO and NVIDIA vGPU, and we’ll provide case studies that leverage NVIDIA vGPU scheduling options such as Equal Share, Fixed Share, and Best Effort and illustrate their benefits.
Watch this session
Join in the conversation below.