Enterprises worldwide now have access to a variety of NVIDIA-validated storage options for NVIDIA OVX computing systems.
To help overcome these hurdles, NVIDIA has introduced the Storage Partner Validation Program for the NVIDIA OVX computing system. Leading the way in completing NVIDIA OVX storage validation are high-performance storage systems including DDN, Dell PowerScale, NetApp, Pure Storage, and WEKA.
The NVIDIA OVX servers combine high-performance GPU-accelerated computing with fast storage access and low-latency networking, catering to a variety of complex AI and graphics-intensive workloads. For instance, chatbots, summarization, and search tools demand vast amounts of data, and high-performance storage is crucial for maximizing system throughput.
To assist enterprises in pairing the right storage devices with NVIDIA-certified OVX servers, the new program offers partners a standardized process for validating their storage devices. They can utilize the same framework and tests required for validating storage with the NVIDIA DGX BasePOD reference architecture.
For validation, partners must complete a set of NVIDIA tests that measure storage performance and input/output scaling across multiple parameters, representing the stringent requirements of various enterprise AI workloads. This includes combinations of different I/O sizes, numbers of threads, buffered and direct I/O, random reads, re-reads, and more.
Each test is run multiple times to verify results and collect necessary data, which is then reviewed by the NVIDIA engineering team to determine if the storage system passes the test.
The program offers normative guidance to ensure enterprise AI workloads using NVIDIA OVX systems achieve optimal storage performance and scalability. Yet, its overall design remains flexible, so customers can customize system and storage choices to fit their existing data center environments and apply accelerated computing anywhere data resides.
Generative AI use cases have different requirements from traditional enterprise applications, thus IT teams must carefully consider their choices of computing, networking, storage, and software to ensure high performance and scalability.
NVIDIA-certified systems are tested and verified to deliver enterprise-grade performance, manageability, security, and scalability for AI workloads. Compared to building from scratch independently, these systems' flexible reference architectures facilitate faster, more efficient, and cost-effective deployments.
OVX servers, powered by NVIDIA GPUs, include NVIDIA AI Enterprise software along with NVIDIA Quantum-2 InfiniBand or NVIDIA Spectrum-X Ethernet, and NVIDIA BlueField-3 DPU. These servers are optimized for generative AI workloads, including training for small LLMs (like Llama 2 7B or 70B), fine-tuning existing models, and high-throughput, low-latency inference.
NVIDIA-validated OVX servers are now available, with global system providers including Gigabyte, HP, and Lenovo offering support. Collaborating with NVIDIA, system manufacturers can provide comprehensive enterprise-level support for these servers.
The validated NVIDIA-certified OVX server storage solutions are now available, with various storage and system suppliers set to release reference architectures in the coming weeks.
Discover more about NVIDIA OVX systems.
