Dell Technologies, VMware, and NVIDIA are working collaboratively to develop solutions to improve the state-of-the-art practices for implementing AI in the enterprise. Flexibility and ease of use continue to be key tenets and desirable traits for artificial intelligence (AI) platforms that are used to train and host models for AI. This white paper describes the Dell Validated Design for Virtualizing GPUs for AI with VMware and NVIDIA, a solution that the three companies jointly engineered and validated. We provide background information and recommendations for how to implement a wide variety of NVIDIA graphics processing unit (GPU) acceleration products using several types of servers from Dell Technologies along with VMware vSphere and Tanzu technology. This solution can match the needs of almost any AI application requirements for enterprise customers who want to standardize on VMware for systems deployment and operations. The validated design also shows how to incorporate NVIDIA AI Enterprise, a comprehensive software suite of AI tools and frameworks, which enables organizations running VMware vSphere to virtualize and containerize AI workloads on NVIDIA-Certified Systems.
One lesson learned over the last decade is that AI platforms with custom architectures—whether on-premises or in a cloud service provider—introduce integration and operational challenges that in turn result in higher costs. With this validated design, VMware professionals can use their existing skill set for deployment and operations of this virtualized AI platform. The ability for an organization to maintain and evolve its existing operational centers of excellence allows for lower total cost of ownership and better overall business continuity.
Enterprises are embracing AI in every aspect of their business. Human resources are using AI for talent acquisition, marketing is using AI for pricing and demand forecasting, IT is using AI for cyber security, and customer services are using AI for chatbots. Such wide adoption of AI across an enterprise requires seamless integration of AI capabilities into its data center operations. Through NVIDIA AI Enterprise on VMware vSphere, this validated design enables enterprises to deploy, modernize, manage, and operate AI workloads along with their existing applications using the same tools with which they are familiar.