Skip to main content

[@ServeTheHomeVideo] Software Defined Storage Solutions - Open Storage Summit 2025 Session 9

· 7 min read
[@ServeTheHomeVideo] Summarizer
Video Bot

@ServeTheHomeVideo - "Software Defined Storage Solutions - Open Storage Summit 2025 Session 9"

Link: https://youtu.be/-kRYnydqgn4

Short Summary​

Okay, here's a breakdown of the provided YouTube transcript, focusing on the key takeaway and a concise summary:

Number One Action Item/Takeaway:

  • Consider software-defined storage solutions from Super Micro and its partners (NVIDIA, DDN, OS Nexus, and Steel Dome) to optimize storage infrastructure for AI, cloud-native, and traditional applications, emphasizing multi-tenancy, scalability, security, and simplified management. The partnerships highlighted demonstrate a commitment to providing comprehensive, validated solutions that address diverse workload requirements.

Executive Summary:

Super Micro's Open Storage Summit showcased collaborative software-defined storage solutions designed to meet the evolving needs of modern workloads. Partners like Nvidia, DDN, OS Nexus, and Steel Dome offer validated platforms focused on scalability, security, and ease of management, enabling organizations to maximize GPU utilization, accelerate AI initiatives, and reduce infrastructure complexity. The solutions emphasize flexibility and rapid time-to-value.

Key Quotes​

Okay, here are 4 quotes from the transcript that I found particularly insightful or interesting:

  1. (John Fragala, Nvidia) "The key measuring stick for the efficiencies of an AI factory is how fast tokens can be generated and the actual amount of tokens that can be generated." - This highlights a specific and measurable metric for evaluating the effectiveness of AI infrastructure.

  2. (Balaji Venatwara, DDN) "...with the partnership of Nvidia and Super Micro we've been able to service a number of happy customers at scale including um GPU deployments as large as 100,000 plus GPUs." This quote is valuable because it gives a quantifiable example of the scale DDN can achieve.

  3. (Steve Umbocker, OS Nexus) "Object storage comes with immutability and versioning. And so that provides a layer of protection against ransomware attacks because ransomware attack cannot uh delete or change that immutable those immutable objects." - This succinctly explains a critical security benefit of object storage, particularly relevant in today's threat landscape.

  4. (Jeff Slap, Steeldome) "Scaling is immediate and seamless whether increasing capacity or compute. Expansion is handled through native software automation without the need for disruptive upgrades. And because all the storage and virtualization features are included out of the box, organizations have everything that they need exactly when they need it without delays or additional licensing overhead." - This captures the key value proposition of Steeldome: simplicity, ease of scaling, and all-inclusive functionality, addressing common pain points in infrastructure management.

Detailed Summary​

Okay, here is a detailed summary of the YouTube video transcript provided, using bullet points and highlighting the key topics, arguments, and information discussed.

Overall Topic: Software-Defined Storage Solutions with Super Micro Partners

  • The video features presentations from Super Micro partners, including Nvidia, DDN, OS Nexus, and Steel Dome, discussing their software-defined storage solutions and how they integrate with Super Micro hardware.

Nvidia Cloud Partner Program (NCP) and AI Factories (John Fragala, Nvidia):

  • Key Value of AI Factory: Efficiency measured by how fast and how many tokens can be generated from prompts through training and inference pipelines.
  • Emphasis on Agentic AI: Reasoning models generating much larger token counts necessitate efficient data input, processing, and output.
  • NCP Infrastructure Pillars:
    • Scale & Performance: Amount of infrastructure (GPUs, bandwidth, software stack) for model and inference workloads.
    • Efficiency: High GPU utilization, low latency (time to first token), power efficiency, cooling for high availability.
    • Security: Multi-tenancy support (data isolation, tenant unawareness, QOS settings for GPU allocation, network isolation, storage support for multi-tenancy, QOS, and data isolation).
  • Reference Architectures: Nvidia publishes for compute and networking, but not for high-performance storage.
  • Storage Certification Program: Nvidia analyzes storage platforms based on:
    • Performance, scale, reliability, uptime, availability, data integrity.
    • Data connector tools.
    • Security & Multi-tenant features (QOS).

DDN Exoscaler Storage (Balaji Venatwara, DDN):

  • Partnership: Partnership with NVIDIA and Super Micro enables DDN to service AI customers at scale.
  • AI Factory Support: Exoscaler fulfills requirements of AI factory deployments.
  • Scalability and Performance: The new AI 400X3 platform delivers top performance, ultra-low latency, multi-tenancy, multi-protocol support, and encryption.
  • Validation: Platform is fully validated with NVIDIA DGX Superpod, NCP, and other platforms.
  • Modular Design: Enables starting small and scaling linearly without upfront CAPEX.
  • Open Interfaces: Seamless integration with Super Micro and Nvidia hardware and the Nvidia software platform.
  • Multi-Tenancy: Ensures tenant isolation for maximum infrastructure utilization.
  • Data Consolidation: Efficiently handles structured/unstructured data, different protocols, and on-prem/cloud/multi-cloud environments to maximize GPU utilization.
  • Key Takeaways: Start small and scale, support multi-tenancy and multi-protocol services, maximize GPU utilization.

Super Micro's Role (Vince Chen, Super Micro):

  • Platform and Infrastructure Provider: Super Micro offers a comprehensive portfolio of compute and storage platforms.
  • Data Center Building Block Solutions: Compute, storage, networking, data center design, power, cooling, integration, deployment.
  • Partnership with DDN and NVIDIA: Provides pre-validated and fully integrated solutions.
  • Serving Customers at Any Scale: From edge and enterprise to large-scale AI GPU data centers.
  • Customer-Focused: Aims to understand customer needs and provide infrastructure solutions for optimal deployment time and value.

OS Nexus and Quantistore (Steve Umbocker, OS Nexus; William Lee, Super Micro):

  • OS Nexus Quantistore Overview: A software-defined storage platform used across various verticals (healthcare, education, government, enterprise). Focused on advanced security and multi-tenancy.
  • Security Features: Meeting compliance standards (HIPPA, SEIS) related to NIS standards.
  • Multi-Tenancy: Sharing and dividing storage across organizations with showback/chargeback accounting.
  • Grid Technology: Manages storage across multiple sites using built-in grid technology.
  • Scale-Up and Scale-Out: Supports both ZFS (scale up) and Ceph (scale out) file systems.
  • Super Micro AI Ready Nodes: For AI or flash use cases, offers rack mount, hyperscale, and martino families.
  • Design Utilities: Web-based tool to design bespoke storage configurations, including power consumption, performance, and BOM (Bill of Materials). Accessible through the Super Micro website.
  • Success Stories: Growth in higher education, financial services, government, and healthcare. Object storage adoption is increasing due to security features (immutability) for ransomware protection.
  • Tiered Storage Solutions: GPU solutions for applications supported by OS Nexus supporting Flash and HDD storage tiers. AI workloads benefit from All Flash and Hybrid deployments.

Steel Dome and Super Micro (Jeff Slap, Steel Dome; Sher Lynn, Super Micro):

  • Steeldome's Strata System: Infrastructure operating system that includes storage virtualization and orchestration services.
  • Simplified Enterprise Infrastructure: Eliminates complexity for faster adoption, with no steep learning curve.
  • Key components: Stratastore (block, file, and object storage), Strataserve (Hypervisor and container management services) and Hyperserve (fully integrated hyperconverged stack).
  • Licensing: Per-node licensing, independent of core count, memory, or storage.
  • Flexible Integration: Modular and independently deployable services.
  • HyperServe Illustration: Combines storage services (StrataStore) with virtualization services (StrataServe) in a unified deployment.
  • Unified Platform: Eliminates disparate storage silos by consolidating diverse workloads.
  • Hardware Flexibility: Runs seamlessly on any Super Micro server or storage system.
  • Total Cost of Ownership (TCO) Reduction: Straightforward licensing, modular architecture, operational efficiency (simplified management, seamless scaling).
  • Steeldome Delivers Architectural Freedom Without Compromise.
  • Minimum Requirements: Single node to start, can grow to many nodes.
  • Scaling: Can be done on an individual node basis or in groups. Automation for adding/decommissioning nodes.
  • Super Micro's Role: Provides validated solutions tailored for different use cases, pre-shipment software installations, and burning tests. Diverse platforms for various needs (compact IoT, multi-node systems, high-performance all-flash NVMe, GPU platforms for AI/ML). One-stop-shop service for L10 and above and rack scales.