Name: Tuning Virtualized GPUs for Optimal Performance on ML/AI Workloads
Start: 2022-07-27T17:00:00Z
End: 2022-07-27T17:00:46.000Z
Location: BrightTALK
Rating: 0

Kubernetes, an open-source container orchestration system, is becoming the consensus API for infrastructure for IT professionals. For data scientists, the once-onerous task of environment and package management is made tremendously easier by containers. And Kubernetes brings a whole new set of benefits for data scientists, including making models portable and reproducible, handling bursty compute requirements of AI workloads, and future-proofing infrastructure. In this panel discussion moderated by Chris Yang, CTO and co-founder of Domino Data Lab, Craig McLuckie, VP of R&D at VMware and Kubernetes Project co-founder, and Chris Lamb, vice president of GPU computing platforms at NVIDIA, will discuss challenges in scaling data science — and how virtualized, containerized data science workloads set the foundation for AI adoption in the enterprise.

A Vision for Kubernetes as the Foundation for Enterprise MLOps

Today's challenges for the modern data center include the computation scalability required for complex, data-intensive workloads such as AI/machine learning, 3D modeling, and big data. NVIDIA DPUs, integrated with VMware infrastructure, are revolutionizing the data center, providing applications and workloads with a unified and integrated infrastructure solution on which the capabilities of the processor are unlocked, enabling secure, parallelized, and accelerated computing by offloading network and storage processing needs from the CPU. This new transformative technology is democratizing AI/ML for all enterprises, providing the modern data center with unbounded scalability and huge compute capabilities. New workloads requiring huge processing power and 3D rendering capabilities are powered by NVIDIA DPUs on VMware. Join Marc Fleischmann, VMware cloud CTO, and Michael Kagan, CTO NVIDIA, and learn about how VMware and NVIDIA have partnered to unlock the power of today’s data center across diverse industries.

Unleashing AI for Every Enterprise

Learn how customers can implement modern AI/machine learning applications on NVIDIA-certified platforms from Dell with VMware vSphere 7 Tanzu and NVIDIA AI Enterprise. We'll demonstrate how vRealize Automation can be used for self-service of AI/ML workloads that utilize Ampere A100 and A30 GPUs with MIG profiles for deep learning and inferencing within the same solution. We'll also show how IT shops can deliver native K8s pods with VMware Tanzu that are GPU-enabled and use NVIDIA AI Enterprise to foster the adoption of accelerated resources that can be harnessed by the entire business. VI admins will learn how to design GPU resources that data practitioners want to consume in a flexible and agile manner with cloud templates. We'll help customers understand the value of mainstream AI while providing an inclusive approach that ensures all business personas can tap into the value and innovation that NVIDIA, Dell, and VMware are delivering.

A Modern Approach to End-to-end AI/ML: Learn How to Deliver Self-service MLOps

b. Machine learning apps are on the move to a new destination — migrating from ad-hoc specialized silos to mainstream IT managed by the enterprise. This is made possible by NVIDIA and VMware's jointly developed AI-Ready Enterprise Platform. Enterprise IT managers want to act as cloud providers to their internal customers. These enterprise IT managers can now leverage the NVIDIA AI Enterprise software suite on VMware, making the infrastructure (including GPUs and accelerators) easily consumable, avoiding low-level administrative tasks, both on-premises and in VMware's Cloud Provider data centers.

 NVIDIA's suite of containerized AI platforms and tools is now fully supported on VMware vSphere with Tanzu, with tight integration of Kubernetes into the hypervisor's control plane. Learn the details of the GPU enablement of virtual machines and using those VMs as nodes in K8S clusters, providing Kubernetes clusters on-demand, with GPUs attached, and the capability to share that GPU power with other users.

Enabling Enterprise Machine Learning with Kubernetes

b. VMware and NVIDIA have partnered to democratize AI for every enterprise by combining NVIDIA AI software and GPUs with virtualization. We’ll present performance data for the two modes of operation that the Ampere architecture supports using Multi Instance GPU (MIG) and vGPU. Learn to load balance, and improve latency and throughput with vGPUs. 

VMware vSphere and NVIDIA AI Enterprise enables multi-GPU configuration with NVLink in a virtualized environment. We’ll demonstrate how NVLink with NVIDIA A100 GPUs can enhance the performance of machine learning/AI workloads in vSphere, and show how to scale performance of NVIDIA GPUs with NVLink with manageability and scalability features of Kubernetes container services and virtual machines.

Tuning Virtualized GPUs for Optimal Performance on ML/AI Workloads

VMware vSphere

Machine Learning

Data Center

Nvidia

Virtualization

Kubernetes

Welcome to the big data and data management community on BrightTALK. Join thousands of data quality engineers, data scientists, database administrators and other professionals to find more information about the hottest topics affecting your data. Subscribe now to learn about efficiently storing, optimizing a complex infrastructure, developing governing policies, ensuring data quality and analyzing data to make better informed decisions. Join the conversation by watching live and on-demand webinars and take the opportunity to interact with top experts and thought leaders in the field.

Big Data and Data Management

As an IT professional, many of the problems you face are multifaceted, complex and don’t lend themselves to simple solutions. The information technology community features useful and free information technology resources. Join to browse thousands of videos and webinars on ITIL best practices, IT security strategy and more presented by leading CTOs, CIOs and other technology experts.

Tuning Virtualized GPUs for Optimal Performance on ML/AI Workloads

Presented by

About this talk

More from this channel