On Demand Webinars
AI systems are only as powerful as the data foundations that support them. As applications increasingly rely on vectors, tables, graphs, and long-lived datasets, object storage must evolve to do more than simply hold data—it must actively support how AI applications coherently access, move, and reuse it over time.
In this webinar we will examine the architectural patterns shaping modern AI-ready object storage. We will explore how object storage is designed to support emerging AI data types alongside traditional unstructured data, while remaining scalable, durable, and cost efficient at cloud scale. Central to this evolution is tighter integration between data lifecycle management and application workflows, enabling data to flow seamlessly from ingestion and training to inference, governance, and long-term retention.
Attendees will gain practical approaches to building AI-ready storage systems that deliver scale, efficient access, and long-term data reuse.
As power densities continue to rise in servers, thermal management is becoming a critical challenge. Traditional air cooling is increasingly strained in high-density environments, driving interest in more efficient approaches. These more efficient approaches touch all aspects of the server including the SSD.
In this webinar, Anthony Constantine and Scott Shadley examine the role of liquid cooling in modern SSD deployments. The session will cover key liquid cooling methods, their advantages in heat transfer and performance, and the emerging specification changes shaping the industry.
Join us to better understand how liquid cooling is shaping the future of SSDs and what it means for your storage infrastructure.
Ceph is synonymous with block, file, and object storage at petabyte to exabyte scale, powering high performance computing, AI/ML workloads, and enterprise IT environments. As generative AI workloads accelerate, these users increasingly need more than traditional storage—they need native support for embedding data and nearest neighbor search to enable retrieval augmented generation at scale.
Today, this capability is often delivered through standalone vector databases, introducing new infrastructure, unfamiliar interfaces, and significant operational overhead. Operations teams and developers alike face the cost and complexity of deploying, securing, and maintaining yet another system alongside their existing storage platforms.
Ceph’s journey toward nearest neighbor search aims to change that. Ongoing research into vector storage formats, libraries, and databases is bringing semantic search closer to where enterprise data already lives. By exposing vector search capabilities through S3 Vectors, Ceph enables AI practitioners to use familiar cloud native tools, SDKs, and workflows, without introducing a separate vector database.
In this webinar, we will explore how LanceDB libraries power a vector search implementation in Ceph, delivered through S3 Vectors API actions. Attendees will see how Ceph can provide out of the box, billion scale, multi tenant nearest neighbor search that meets the needs of most RAG workloads while preserving the unified, software defined storage model enterprises already trust.
In today’s evolving IT landscape, selecting the right storage architecture is critical for optimal performance, scalability, data governance, and cost-efficiency. Furthermore, AI workloads have uniquely influenced how we meet these demands from our storage infrastructure. This webinar provides a technical deep dive into three fundamental storage deployment models – on-premises, cloud, and hybrid – examining their architectures and operational trade-offs through the lens of two key concepts: indirection (accessing data through mapping layers that provide flexibility and abstraction) and redirection (rerouting data requests to enable failover, load balancing, and optimized performance).
As data demands shift from traditional enterprise workloads to massive AI/ML pipelines, the storage software stack has undergone a radical transformation. During the process of storage systems scaling from a single server to hyperscale and AI workloads, the placement and virtualization of storage software intelligence within the stack have become the dominant determinant of performance, cost, and operational agility.
In this webinar, we will try to answer the following question: Given that storage access semantics are already abstracted into files, blocks, and objects, the remaining challenge is architectural: As storage systems evolve from raw hardware to application-facing services, where in the storage software stack should virtualization be implemented to optimize performance, scalability, and resiliency across modern workloads and vastly different physical deployments?
This webinar will help the audience to:
- Have a comprehensive understanding of the storage software stack, moving from the application and logical abstraction layers all the way down to the physical interface
Gain insights on storage virtualization and networking interconnects
Understand scaling terminologies (scale-out, scale-up, scale-across)
Appreciate how the technological evolutions in the above-mentioned areas are helping in scaling storage capacity and data services, making them suitable for modern enterprise and cloud/AI applications
Participants will walk away with a clear understanding of the storage software stack, storage functionality updates, and how storage is evolving to keep pace with the advancements that cater to modern applications like HPC/AI.
Getting started with AI in the enterprise is less about training a large language model from scratch and more about selecting a model and fine-tuning it to do what you actually need. The massive compute runs that cost hundreds of millions of dollars and consume months of GPU time? That’s already been done by OpenAI, Meta, Google, Anthropic, and Mistral. For enterprise teams, the work — and the opportunity — begins after model selection.
That’s where post-training comes in — and it’s the natural area for enterprise practitioners to develop expertise. Post-training is the collection of techniques used to take a general-purpose model and make it follow instructions, align with your organization’s preferences, reason through domain-specific problems, and integrate with your tools and workflows. This is where your infrastructure decisions directly affect model quality, where storage and compute choices create real differentiation, and where your team has genuine agency.
This webinar provides a practitioner-oriented overview of the full model development pipeline, with a clear focus on the stages that matter to enterprise teams. We’ll briefly establish the complete taxonomy — training, mid-training, and post-training — so you have the right mental model, then spend the majority of the session on the techniques you’ll actually use or influence: supervised fine-tuning, reinforcement learning from human feedback, parameter-efficient adaptation, and the state-of-the-art reasoning techniques (GRPO, RLVR) behind models like DeepSeek R1 and OpenAI’s o-series.
See Post-Training in Action — Live in Your Browser
What actually happens when you fine-tune a language model? In this live demo, we take a 360-million parameter model that knows nothing about storage and walk it through the full post-training pipeline — supervised fine-tuning, preference optimization, and reinforcement learning — until it can classify storage I/O workloads on sight.
Every training curve, every weight update, every generation is real. And at the end, you'll run the models yourself — right in your browser, no GPU required — and see the difference that post-training makes with your own eyes. Built on HuggingFace SmolLM2, trained on storage I/O patterns, and running entirely client-side.
As Agentic AI moves beyond foundational theory into real-world deployment, the focus shifts to the architectural and operational enablers that make scalable intelligence possible. This session explores the emergence of Model Context Protocol (MCP) and Agent-to-Agent (A2A) communication as critical components for orchestrating distributed, autonomous systems—laying the groundwork for the Agentic Mesh, where agents collaborate, adapt, and self-optimize across domains.
We’ll examine how Agentic AI operates in, on, with, and for storage systems—transforming them from passive data repositories into active participants in decision-making and workflow execution. From client-side perspectives, we’ll discuss how agentic interactions reshape expectations around latency, autonomy, and data locality, and what this means for future-ready infrastructure.
Read the Q&A blog: https://www.snia.org/blog/2026/how-agentic-ai-transforms-role-storage-qa
As AI workloads continue to scale, the physical infrastructure supporting them must evolve to deliver the performance, reliability, and efficiency required by modern data-intensive applications. This webinar will focus on the interconnect foundations of AI infrastructure, including storage. There will be an emphasis on the form factors, connectors, cables, and transceivers standards that enable scalable and interoperable systems.
We’ll explore how SNIA’s SFF Technical Work Group is contributing to the development of physical layer standards that support high-performance interconnects and storage devices used in AI environments. Topics will include the importance of transceiver innovation and the need to evolve; the role that form factors like EDSFF play; and how innovative cabling and connector designs support the adoption of high-speed signaling technologies like PCIe 8.0 and beyond.
Read the webinar Q&A blog: "Q&A: 400G and PCIe 8.0 in Next-Gen Interconnects" https://www.snia.org/blog/2026/qa-400g-and-pcie-80-next-gen-interconnects
Storage security changes and adaptations are a fact of life to deal with the ever-changing threat and regulatory landscapes. A black swan event, new standard, or regulation often serve as catalysts for organizations to review their controls and practices. The recent publication of the NIST Special Publication 800-88 Rev. 2 Guidelines for Media Sanitization is drawing attention to storage security, but it is not the only development worth noting. SNIA, the Open Compute Project (OCP), Trusted Computing Group (TCG), and IEEE have or are developing standards and specifications that could be important going forward. One major theme underlying several of the activities is in storage sanitization (i.e., controlled eradication of data).
This session brings together a unique panel of experts who have served as editors of some of the most important storage sanitization standards. These experts also have broad knowledge of storage security. As a panel, they will provide insight into these new developments as well as observations on what organizations are experiencing.
Smarter data management is the key to unlocking the full potential of next-generation data infrastructure for AI. This webinar explores how the SNIA Cloud Data Management Interface (CDMI™) standard enables government labs, HPC centers, and organizations leveraging the power of AI/ML to reimagine and innovate their infrastructure to meet next-generation data management demands. CDMI is a mature ISO standard (ISO/IEC: 17826:2022) with over 15 years of development and implementation, and is purpose-built for cloud data management. CDMI's extensive capabilities enhance key aspects of AI operations by transforming data management challenges into streamlined, intelligent workflows.
CDMI 3.0 (in development now) will further extend the standard with:
• Automated resource discovery that accelerates AI workload efficiency
• Portable data movement that enables preservation of metadata fidelity during migrations, including graph relationships and knowledge graph
• MCP-based data management that enhances AI accuracy while improving access speed
Read the SNIA CDMI 3.0 white paper
