We’re excited to be at NVIDIA’s GTC 2021 and hope you’re all registered to be at the show this week! Check out what we’re doing at the show, and find out about how LIqid composable disaggregated infrastructure (CDI) solutions can increase efficiency, improve flexibility, and increase time to value for AI applications that NVIDIA’s evolving ecosystem designed for AI applications for data center-scale computing whether it’s in the cloud, on the edge, or on-premises.
Meanwhile, we thought we’d update on what’s on at the show so far!
Three-chip strategy for data center-scale computing
NVIDIA announced its new Grace CPUs. The chips were designed for terabyte-scale computing and join NVIDIA’s new Bluefield-3 SmartNIC data processing unit (DPU), designed to offload data operations from the CPU for improved performance applications, and its well-established A100 GPU technologies.
Will these three chips come to dominate data centers-scale computing for AI? Unclear, but the combined data processing power and ability to offer significantly improved resource balancing for AI applications is extremely compelling for cloud-native compute and also hybrid compute environments that include edge, on-premises, and cloud instances with significant reductions in latency for AI applications.
Tighter integration with VMware
NVIDIA is also working with VMware to better orchestrate NVIDIA GPUs with the container platform Kubernetes on VMware’s newly announced vSphere 7 platform, enabling vSphere to partition individual NVIDIA A100 GPUs for use by separate applications. This enables GPU-driven AI workloads to perform on virtual machines in vSphere.
NVIDIA’s new software suite NVIDIA AI Enterprise, is core to this collaboration, and the vSphere certification is the first of its kind. The company describes it as “an end-to-end, cloud-native suite of AI and data analytics software, optimized, certified, and supported by NVIDIA to run on VMware vSphere with NVIDIA-Certified Systems™.”
The platform provides a number of tools used by data scientists, AI researchers, devops professionals and other digital professionals who seek to speed the deployment of AI-centric applications into production.
Composing GPU and more with Liqid
Liqid is capable of composing for all industry-standard data center accelerators, in virtualized or bare-metal environments, making it a universal plane through which physical hardware is invisibly managed. We’re excited to be at the show this year with sessions that highlight Liqid’s ability to speed time to value for AI-driven workloads, optimize data center footprints with the ability to share accelerator devices, and increase overall infrastructure flexibility in order to meet the demands of uneven data operations at each phase of the AI process, from data intake to inference.
The Liqid Matrix command center does this via a single pane of glass, while integrating with high-value enterprise and HPC containers and workload managers such as Slurm and Kubernetes. NVIDIA A100 GPUs, SmartNIC technologies such as Bluefield DPU, NVIDIA Grace CPU, high-performance NVMe, and other accelerators can be mapped to all possible configurations, scheduled and deployed on demand.
Our sessions at this year’s conference include:
- SS33102: Composable A100 and NVIDIA Intelligent Networking: Adaptive Architecture for AI-driven Computing (Liqid Inc.)
- SS33106: Accelerating AI from Prototype to Production (Presented by Liqid Inc.)
- SS33107: Technical Validation: Composable Disaggregated Infrastructure for GPU Performance Optimization (Presented by Liqid Inc.)
Join the Liqid team at NVIDIA GTC 21 for breakthroughs in AI, data center, accelerated computing, healthcare, intelligent networking, game development, and more. Discover the advanced technologies that are transforming today's industries.
Register for GTC 2021 today for free and sign up for Liqid sessions. Find out how you can partner with Liqid to deliver adaptive composable disaggregated solutions to your channel today, or reach out to schedule a demonstration and look into the future of AI.
