Arista and Wi-Fi 7: What the Real World and Arista Lab Tests are Really Telling Us

Event: Mobility Field Day 13

Appearance: Arista Presents at Mobility Field Day 13

Company: Arista

Video Links:

Personnel: Asvin Kumar Muthrurangam

In this session, Asvin will share some insights into the Multi-Link Operation (MLO) behavior of various Wi-Fi 7 clients based on testing conducted in Arista’s lab. Through a live demo, we’ll see how MLO-enabled clients use the spectrum intelligently compared to non-MLO clients – highlighting the real-world benefits of MLO in action.


Engineering Campus-Wide Mobility: Arista’s Scalable Wi-Fi Roaming Design

Event: Mobility Field Day 13

Appearance: Arista Presents at Mobility Field Day 13

Company: Arista

Video Links:

Personnel: Ken Duda

Designing for large Wi-Fi roaming domain especially in environments like university campus has many challenges. In this video, Arista’s own founder and CTO, Ken Duda will talk about how we applied some of our learnings from large scale data center & AI cluster deployments to solve Wi-Fi roaming and thus unifying our wired and WLAN data plane fabric.


Arista’s Latest and Greatest: Innovations in Campus Since the Last Mobility Field Day

Event: Mobility Field Day 13

Appearance: Arista Presents at Mobility Field Day 13

Company: Arista

Video Links:

Personnel: Kumar Srikantan

In this video, get an update from Kumar Srikantan about the state of Arista’s Campus solution in 2025. Hear about the new Wi-Fi 7 Access Point portfolio, expanded switching portfolio, exciting features especially around AIOps and CloudVision AGNI.


Cisco SDN Strategy Discussion

Event:

Appearance: Packet Pushers Discussion of Cisco’s SDN Strategy at Cisco Live US 2012

Company: Cisco

Video Links:

Personnel: Derick Winkworth, Ethan Banks, Greg Ferro, Russ White, Stephen Foskett, Tom Hollingsworth

A group of independent thought leaders from the Networking Field Day/PacketPushers crew gathered at Cisco Live US 2012 to discuss the company’s Open Networking Environment (ONE) announcement. This announcement centered on a strategy for software-defined networking (SDN), and this was the focus of our discussion as well.

This wide-ranging discussion touched on the following topics:

  • Contrasting Cisco’s ONE strategy with SDN and OpenFlow in general
  • APIs, OpenFlow, and XML
  • What will people do with SDN in the future?
  • Distributed and autonomous versus centralized
  • Standards: IEEE vs. IETF, de facto and interoperability
  • VXLAN and the Nexus 1000V – Is 1000V SDN?
  • Operational and organizational impacts
  • Systems engineering
  • Thinking of networks as flows

The conversation will continue in July with two more “Virtual Symposium” discussions with Cisco. We will cover Network Programmability and Virtual Machine Networking. Watch for more!


Demonstration of Day 2 AI network operations, monitoring and anomaly detection with Aviz

Event: AI Infrastructure Field Day 2

Appearance: Aviz Networks Presents at AI Infrastructure Field Day 2

Company: Aviz Networks

Video Links:

Personnel: Ravi Kumar

Aviz Networks’ AI Infrastructure Field Day demonstration focused on Day 2 operations, monitoring, and anomaly detection for AI workloads. The core challenge addressed is the specialized networking requirements of AI, including multiple networks, differentiated QoS, and the need to manage compute as part of the end-to-end network topology. Aviz presented solutions for orchestrating AI fabrics based on Sonic and NVIDIA’s Spectrum-X reference architecture, showcasing a customer workflow that includes network design, Day 0 infrastructure deployment, Day 1 tenant onboarding and traffic isolation, and Day 2 operations like adding Pods, handling alerts, and troubleshooting.

The presentation demonstrated Aviz’s orchestration capabilities for Sonic-based and NVIDIA RA-based AI fabrics. For Sonic, the presenter showed how to orchestrate the fabric using YAML-based intent, validating configurations, and performing operational checks. The demonstration emphasized the ease of use of industry-standard CLI, built-in validation, and the ability to compare configurations to identify any drift. With the NVIDIA Spectrum-X platform, the presentation highlighted agentless orchestration, the use of NVIDIA AIR for simulating deployments, and config comparison.

Finally, the presentation detailed Aviz’s monitoring and anomaly detection features. The tool provides comprehensive monitoring with a bottom-up approach for networks, servers, and GPUs. The demo showed how to view various telemetry data, including traffic, queue drops, and GPU health metrics. The presentation also covered Aviz’s built-in anomaly detection system, which allows users to create custom rules and receive notifications through tools like Slack and Zendesk. The system includes curated rules, role-based access control, and configuration comparison capabilities to streamline operations and reduce potential errors.


Design, deploy, and monitor networks for AI with Aviz

Event: AI Infrastructure Field Day 2

Appearance: Aviz Networks Presents at AI Infrastructure Field Day 2

Company: Aviz Networks

Video Links:

Personnel: Thomas Scheibe

Thomas Scheibe, Chief Product Officer, offers solutions for designing, deploying, and monitoring networks for AI workloads. Their focus is on addressing the specialized networking needs of AI, including multiple networks, differentiated Quality of Service (QoS), and the integration of compute into the end-to-end network topology. They aim to provide automation and orchestration for faster deployment, service activation, and infrastructure expansion. Their product, ONCE, supports Sonic and Cumulus network operating systems, focusing on streamlining network management through design, modeling, deployment, and monitoring capabilities.

The Aviz presentation highlighted the evolution of networking in AI, emphasizing the shift from a single data center network to multiple networks, particularly the separation between front-end (user access) and back-end (GPU communication) networks. Aviz recognizes the importance of lossless behavior, different methods to address AI application requirements, and the integration of network settings on both the switches and the network interface cards (NICs). The company partners with hardware providers and uses reference architectures like NVIDIA Spectrum-X to automate network configuration. This allows enterprises to define networks and configure network separation.

Aviz offers comprehensive support for Sonic deployments in enterprise data centers and at the edge. They are automating deployment workflows for the NVIDIA Spectrum-X reference architecture, with the ability to configure multi-tenancy and extend the fabric. Aviz simplifies network management in AI, allowing users to deploy and manage their networks quickly and efficiently. They offer a comprehensive suite of solutions to design, deploy, and monitor networks for AI, focusing on automation and orchestration.


Wrapping up and summarizing Nutanix Enterprise AI

Event: AI Infrastructure Field Day 2

Appearance: Nutanix Presents at AI Infrastructure Field Day 2

Company: Nutanix

Video Links:

Personnel: Mike Barmonde

The Nutanix presentation at AI Infrastructure Field Day focused on enterprise AI solutions, emphasizing giving customers a solid technical understanding of Nutanix Enterprise AI (NAI) and its role in addressing key customer challenges. The discussion highlighted the curated model catalog, offering pre-configured and customizable models, and the ability to incorporate new cutting-edge models, even within air-gapped environments, easily. This approach provides control over models and data, which is particularly relevant for customers seeking sovereign AI solutions and needing to deploy AI models in their own environments.

Nutanix also emphasized the “deploy once, inference many” model, allowing for the creation of a shared service model where multiple applications can connect to deployed models via endpoints. Furthermore, the session touched upon the simplification of sizing, as NAI streamlines the deployment of models, making the process straightforward. The speaker reiterated the benefits of NAI as an application running on Kubernetes, offering flexibility and portability. The presentation concluded by discussing the future of distributed inference across multiple nodes, acknowledging its importance and status as a planned future development.

A key takeaway from the presentation was the growing demand for sovereign AI, driven by geopolitical factors and specific terms of service that restrict the use of certain models in certain regions. Nutanix recognizes and actively helps its customers address this need by providing the necessary tools and infrastructure to enable control over AI models and data within their own environments. The company’s commitment to adapting and evolving its AI solutions to meet the rapid advancements in the AI landscape was underscored, ensuring that Nutanix remains a relevant player in the enterprise AI space.


AI Inferencing Sizing Considerations on Nutanix Enterprise AI

Event: AI Infrastructure Field Day 2

Appearance: Nutanix Presents at AI Infrastructure Field Day 2

Company: Nutanix

Video Links:

Personnel: Jesse Gonzales

Jesse Gonzales, Staff Solution Architect, offers sizing guidance for AI inferencing based on real-world experience. The presentation focuses on the critical aspect of appropriately sizing AI infrastructure, particularly for inferencing workloads. Gonzales emphasized the need to understand model requirements, GPU device types, and the role of inference engines. He walks the audience through considerations like CPU and memory requirements based on the selected inference engine, and how this directly impacts the resources needed on Kubernetes worker nodes. The discussion also touches on the importance of accounting for administrative overhead and high availability when deploying LLM endpoints, offering a practical guide to managing resources within a Kubernetes cluster.

The presentation highlights the value of the Nutanix Enterprise AI’s pre-validated models, offering recommendations on the specific resources needed to run a model in a production-ready environment. Gonzales discussed the shift in customer focus from proof-of-concept to centralized systems that allow for sharing large models. The discussion also underscores the importance of accounting for factors like planned maintenance and ensuring sufficient capacity for pod migration. Gonzales explained the sizing process, starting with model selection, GPU device identification, and determining GPU count, followed by calculating CPU and memory needs.

Throughout the presentation, Gonzales addresses critical aspects like FinOps and cost management, highlighting the forthcoming integration of metrics for request counts, latency, and eventually, token-based consumption. He addressed questions about the deployment and licensing options for Nutanix Enterprise AI (NAI), offering different scenarios for on-premises, bare metal, and cloud deployments, depending on the customer’s existing infrastructure. Nutanix’s approach revolves around flexibility, supporting various choices in infrastructure, virtualization, and Kubernetes distributions. The presentation demonstrates how the company streamlines AI deployment and management, making it easier for customers to navigate the complexities of AI infrastructure and scale as needed.


Nutanix Enterprise AI Demonstration

Event: AI Infrastructure Field Day 2

Appearance: Nutanix Presents at AI Infrastructure Field Day 2

Company: Nutanix

Video Links:

Personnel: Laura Jordana

As presented by Laura Jordana, Nutanix Enterprise AI (NAI) is designed to simplify the process of deploying and managing AI models for IT administrators and developers. The presentation begins by demonstrating the NAI interface, a Kubernetes application deployable on various platforms. The primary use case highlighted is enabling IT admins to provide developers with easy access to LLMs by connecting to external model repositories and creating secure endpoints. This allows developers to build and deploy AI workflows while keeping data within the organization’s control.

The demo showcases the dashboard, which offers insights into active endpoints, request metrics, and infrastructure health. This view is crucial for IT admins to monitor model usage and impact on resources. The process involves importing models from various hubs like Hugging Face and creating endpoints that serve as the inference engine connection. The presenter emphasized the simplicity of this process, with much of the configuration pre-filled to ease the admin workload. They also highlighted the platform’s OpenAI compatibility, allowing integration with existing tools.

While focusing on inferencing, not model training, the platform provides a secure and streamlined way to deploy and manage models within the organization’s infrastructure. The key takeaway from the presentation is the simplification of AI model deployment, focusing on day 2 operations and ease of use. The platform leverages Kubernetes’ ability to run on Nutanix, EKS, and other cloud instances. It also provides API access and monitoring capabilities for IT admins, and easy access to LLMs for AI developers.


Lets take a look at Nutanix Enterprise AI

Event: AI Infrastructure Field Day 2

Appearance: Nutanix Presents at AI Infrastructure Field Day 2

Company: Nutanix

Video Links:

Personnel: Ashwini Vasanth

Ashwini Vasanth presented Nutanix Enterprise AI, which simplifies the complexities of adopting and deploying GenAI models and addresses common customer challenges. The product, launched in November 2024, focuses on providing a curated and validated approach to model selection, deployment, and security. The presentation highlighted the “cold start” problem, acknowledging the overwhelming number of available models and the need for a user-friendly starting point for IT or AI admins.

Nutanix Enterprise AI offers a curated list of validated models through partnerships with Hugging Face and NVIDIA to address these challenges, providing a “small, medium, and large” selection. This approach aims to simplify model selection and ensure reliable operation. Additionally, the platform handles GPU selection, inference engine choices, and security complexities, incorporating dynamic endpoint creation to streamline the deployment process. Key to Nutanix’s offering is the integrated security, where Nutanix security experts perform scans for vulnerabilities, eliminating the need for customers to manage their security efforts.

Beyond the mechanics of model deployment, Vasanth discussed the need for on-premises deployment, choice of environments, and addressing the “shadow IT” problem through centralized resource management and monitoring dashboards. The presentation underscored Nutanix’s strategic move into the AI space, leveraging its existing infrastructure expertise, including its Kubernetes platform, storage solutions, and the core principles of simplifying infrastructure. The company’s approach has evolved from a solutions-based offering to a full-fledged product based on the need for a pre-integrated AI platform.


Company Overview and AI Challenges we address with Nutanix

Event: AI Infrastructure Field Day 2

Appearance: Nutanix Presents at AI Infrastructure Field Day 2

Company: Nutanix

Video Links:

Personnel: Mike Barmonde

GenAI’s rapid advancement and impact present a significant challenge for enterprises seeking to leverage its potential. Nutanix helps businesses transition from GenAI possibilities to production with its Nutanix Enterprise AI (NAI) solution, a full-stack AI infrastructure designed specifically for IT needs. The NAI offering provides a standardized inferencing solution centered around a model repository, allowing for creating secure endpoints with APIs for GenAI applications, spanning from edge to public clouds.

Mike Barmonde, the Sr. Product Marketing Manager for Nutanix AI products, presented an overview of Nutanix and its approach to addressing AI challenges. The presentation focused on how Nutanix simplifies AI inferencing for IT, highlighting that many organizations struggle to scale their AI initiatives. Nutanix Enterprise AI provides a four-step process to deploy AI infrastructure, including Kubernetes selection, hardware choice (with options for public cloud or air-gapped environments), LLM deployment from various sources, and the creation of secure endpoints, all managed from a central location.

The presentation emphasized the comprehensive nature of Nutanix’s AI infrastructure approach, extending from LLMs down to the underlying hardware. Nutanix’s goal is to streamline the entire process, enabling seamless Day 2 operations. This allows IT professionals to centralize their AI infrastructure and provide a better experience for their developers and application owners.


Multi-Tenancy & Network Automation for AI Infrastructure Operators Demonstrated with Netris

Event: AI Infrastructure Field Day 2

Appearance: Netris Presents at AI Infrastructure Field Day 2

Company: Netris

Video Links:

Personnel: Alex Saroyan

Netris CEO Alex Soroyan demonstrated the multi-tenancy and network automation solution in AI infrastructure. The presentation began with a live demonstration of the Netris controller, showcasing how it facilitates the setup and management of AI infrastructure networking. Utilizing Terraform modules and a “CloudSim” simulation, Soroyan illustrated the process of initializing the controller, generating network configurations based on user-defined parameters, and creating a digital twin of the network for validation.

The core of the presentation focused on day-2 operations, specifically the creation and management of tenants and network isolation. Using templates, Soroyan showed how easy it is to establish isolated clusters (VPCs) for different tenants. These templates translate high-level server assignments into low-level switch port configurations, enabling a cloud-native approach to network management. The demo also highlighted the integration of Elastic IPs to expose the internal clusters to the outside world.

Finally, Soroyan discussed monitoring features, which automate the configuration of monitoring tools and provide network health checks, including link validation. The presentation also touched on InfiniBand networking, demonstrating Netris’s capability to manage InfiniBand fabrics and integrate them with Ethernet networks. The key takeaways were automating network tasks, simplifying complex configurations through templates, and comprehensive monitoring capabilities, all contributing to a more efficient and manageable AI infrastructure environment.


How it works. Multi-Tenancy & Network Automation for AI Infrastructure Operators with Netris

Event: AI Infrastructure Field Day 2

Appearance: Netris Presents at AI Infrastructure Field Day 2

Company: Netris

Video Links:

Personnel: Alex Saroyan

Netris, as presented by CEO Alex Soroyan, offers cloud-provider-grade network automation and multi-tenancy software tailored for AI Infrastructure operators. The core of their solution lies in the Netris Controller, which acts as the centralized source of truth for network engineers.  It allows for the modeling and simulating network infrastructure using tools like Terraform and CloudSim, while also providing APIs that can integrate into cloud provider platforms, facilitating the creation of VPCs and managing network functions. A key component of their offering is SoftGate, a gateway on Linux servers that provides functions such as elastic load balancing and NAT. It offers a streamlined, integrated solution compared to separate, third-party products.

The presentation details Netris’ approach to day-zero and day-one operations, highlighting the use of Terraform for infrastructure-as-code methodologies and how the controller facilitates the deployment and management of various switch vendors.  The system supports granular multi-tenancy through VXLANs and is designed to integrate with shared storage solutions.  Netris facilitates access and isolation by allowing access to the network from the storage and the tenants, intending to integrate directly with storage vendors via their API. This setup allows for a cloud-like experience for AI infrastructure operators, streamlining the onboarding of tenants and the allocation of resources.

Netris differentiates itself by being multi-vendor and providing cloud networking constructs not typically found in traditional network automation platforms. The presentation emphasized the efficiency and integration provided by SoftGate, which eliminates the complexity of connecting firewalls and load balancers while supporting InfiniBand through integration with NVIDIA UFM.  Alex expressed confidence in Netris’ position, particularly given the growing demand for cloud-provider-like capabilities in the AI infrastructure space.


Introduction to Multi-Tenancy & Network Automation for AI Infrastructure Operators with Netris

Event: AI Infrastructure Field Day 2

Appearance: Netris Presents at AI Infrastructure Field Day 2

Company: Netris

Video Links:

Personnel: Alex Saroyan

Netris helps GPU-based AI infrastructure operators automate their networks, provide multi-tenancy and isolation, and offer essential cloud networking features like VPCs, internet gateways, and load balancers. Netris focuses on network software designed for AI and cloud infrastructure operators because the growing popularity of AI necessitates specialized networking solutions to handle demanding AI workloads. Netris’s technology is particularly well-aligned with NVIDIA’s networking offerings, which are based on the foundation of Mellanox and Cumulus networks.

The presentation highlights the importance of dynamic multi-tenancy for maximizing the utilization of expensive GPUs. Netris provides “cloud provider grade network automation software” that allows AI infrastructure operators to achieve security levels comparable to physical isolation while maintaining software-driven speed. This solves the problem of manual network configuration, which is time-consuming, error-prone, and doesn’t scale. Furthermore, Netris supports cloud networking functions like Internet gateways, NAT gateways, and load balancers, offering a complete solution that addresses the need for secure and flexible network management in AI environments.

Netris’s solution is built on three key pillars: VPCs for isolation, cloud networking functions for connectivity, and fabric management for network operations. They manage both Ethernet and InfiniBand fabrics, providing operators with a single pane of glass.  For InfiniBand fabrics, Netris integrates with NVIDIA’s UFM controllers. On the Ethernet side, Netris acts as the fabric manager for several vendors, including NVIDIA, Dell, and Arista, automating the management of network switches and streamlining operations. The goal is to offer a comprehensive, integrated network automation platform tailored for the demands of AI infrastructure.


The Pascari SSD portfolio by Phison

Event: AI Infrastructure Field Day 2

Appearance: Phison Technology Presents at AI Infrastructure Field Day 2

Company: Phison

Video Links:

Personnel: Chris Ramseyer

This Phison presentation at AI Infrastructure Field Day showcased their Pascari SSD portfolio, emphasizing innovation in the enterprise SSD market through performance and reliability. The presentation, led by Chris Ramseyer, focused on Phison’s high-performance Pascari Enterprise X series, designed for data-intensive applications like AI and machine learning. The X series boasts impressive speeds, including up to 15 GB/s sequential read performance and 3.2 million IOPS, positioning it as a leader in the market. The discussion highlighted the shift in data center workloads due to AI, particularly the need for higher queue depths and increased read bandwidth.

The presentation delved into the specifics of Phison’s SSDs, highlighting their various form factors and benefits, such as increased storage density and sustainability through improved power efficiency. The Pascari D-series, in particular, was showcased for its high capacity, with the D205V reaching 122.88 TB, and different form factors, including E1.S. Phison also demonstrated their customization capabilities, emphasizing collaborations, for example, with VDURA, which allows them to tailor their products to meet specific customer needs. In addition, Phison also offers boot drives and SATA drives to the legacy market.

The presentation concluded with three key takeaways: Phison’s broad portfolio and deep expertise, their status as a trusted innovation partner, and their unmatched flexibility, reliability, and performance across hyperscale, AI, and legacy systems. The presentation underscored Phison’s commitment to innovation, evidenced by their vertically integrated approach with in-house IP, controller knowledge, and testing capabilities. The team also highlighted the value of their products used in space and how their technology “trickles down” for use on Earth. Overall, Phison presented itself as a flexible and capable partner for enterprises seeking high-performance, reliable, and customizable SSD solutions to meet the evolving demands of modern data-intensive workloads.


Innovation in the enterprise SSD market driven by Phison

Event: AI Infrastructure Field Day 2

Appearance: Phison Technology Presents at AI Infrastructure Field Day 2

Company: Phison

Video Links:

Personnel: Michael Wu

Phison, a leading innovator in the enterprise SSD market, is driving the future of data-intensive applications with its Pascari SSDs. Michael Wu, GM & President, presented at AI Infrastructure Field Day, highlighting Phison’s commitment to innovation, focusing on performance and reliability in the enterprise space. Wu shared insights into the company’s journey, from its origins as an engineering-focused company to its current status as a $2 billion enterprise, and its evolution from a behind-the-scenes technology provider to a brand recognized for its cutting-edge SSD solutions.

Phison is uniquely positioned to meet the evolving needs of the enterprise market. Their strategy revolves around a vertically integrated approach, encompassing controller, firmware, and hardware design, allowing them to offer customized solutions through their Imagine Plus design service. This focus on customization, coupled with their early adoption of dual-port technology for enterprise applications and a commitment to providing legacy support, sets them apart.  They also emphasize their commitment to innovation with their world’s first achievements and unique NAND emulator system.

Looking ahead, Phison is strategically expanding into the enterprise market. They are investing heavily in R&D, with 75% of their workforce dedicated to it. Furthermore, Phison is actively establishing regional presences through joint ventures and partnerships, particularly in India and Malaysia, to provide local customers with tailored solutions. Their approach to the market, supported by their proprietary NAND emulator, allows Phison to be first to market with new technologies. As a result, they anticipate substantial growth in enterprise revenue, fueled by the rising demand for high-density storage solutions and their commitment to being the leader.


GPU Memory Offload for LLM fine-tuning and inference with Phison aiDAPTIV+

Event: AI Infrastructure Field Day 2

Appearance: Phison Technology Presents at AI Infrastructure Field Day 2

Company: Phison

Video Links:

Personnel: Sebastien Jean

With aiDAPTIV+, Phison makes on-premises AI processing more accessible and affordable, especially for small and medium-sized businesses, government entities, and educational institutions. CTO Sebastien Jean explained that the primary goal of Phison’s solution is to facilitate fine-tuning of large language models (LLMs) on-site. Fine-tuning often demands significantly more memory than inference, making it expensive and difficult for those without massive budgets or a lot of resources.  The presentation highlighted the massive memory requirements for fine-tuning, which can be up to 20 times the memory needed to run an LLM, driving up costs and making it impossible for some organizations to begin with this approach.

Phison’s solution addresses this challenge by decoupling compute and memory. Sebastien Jean, Phison’s CTO, focused on how Phison’s technology, with its AI-optimized SSDs and middleware, enables on-site LLM training and inference. The product uses a combination of their proprietary middleware, Adaptive Link, and custom-built ProSuite software to manage and extend the memory available to PyTorch, effectively turning an SSD into an extended memory pool. This architecture allows for training large models using fewer GPUs.  The system uses a software layer within PyTorch that intercepts calls and then offloads slices of the model to the SSD, which helps in memory management.

By leveraging SSDs and their proprietary controller technology, Phison offers a cost-effective alternative to expensive GPU-intensive setups and targets the SMB, government, and education markets with this solution.  The presentation concluded with a focus on the financial benefits and the sustainability of the solution. By allowing for more efficient hardware utilization, Phison provides not just a financially smart solution but one with power and cooling benefits as well.  Also, by using repurposed NAND, the solution can increase the lifespan of hardware, reduce electronic waste, and extend the useful life of data center infrastructure.


Affordable on premises LLM training and inference with Phison aiDAPTIV+

Event: AI Infrastructure Field Day 2

Appearance: Phison Technology Presents at AI Infrastructure Field Day 2

Company: Phison

Video Links:

Personnel: Brian Cox

Phison’s show their aiDAPTIV+ technology, designed to make on-premises AI processing more affordable and accessible, particularly for small to medium-sized businesses, governments, and universities. The core of their innovation lies in leveraging flash storage to offload the memory demands of large language models (LLMs) from the GPU. This approach addresses the growing challenge of limited GPU memory capacity, which often necessitates buying more GPUs than needed, primarily for the memory capacity.

Phison’s solution enables the loading of LLMs onto high-capacity, cost-effective flash memory, allowing the GPU to access the necessary data in slices for processing. This significantly reduces the cost compared to traditional deployments that rely solely on GPU memory. The company is partnering with OEMs to integrate their technology into various platforms, including desktops, laptops, and even IoT devices, with a focus on providing pre-tested solutions for a seamless user experience. They are also expanding their partnerships to include storage systems.

Beyond hardware, Phison also addresses the knowledge gap in LLM training by offering educational programs and working with universities to provide students with access to affordable AI infrastructure. Their teaching PC, offered in partnership with Newegg, aims to democratize LLM training by making it accessible in classrooms. The company’s efforts focus on fine-tuning pre-existing foundational models with domain-specific data, allowing businesses and institutions to tailor AI to their unique needs and keep their data private.


Industry First SSDs Tackle AI Challenges with Solidigm

Event: AI Infrastructure Field Day 2

Appearance: Solidigm Presents at AI Infrastructure Field Day 2

Company: Solidigm

Video Links:

Personnel: Scott Shadley

Solidigm’s Director of Leadership Narrative and Evangelist, Scott Shadley, introduces their new SSDs designed to tackle the challenges of AI infrastructure. Shadley begins by highlighting the industry-first status of their 122TB SSDs, emphasizing their potential to revolutionize storage solutions. He shares a case study involving Dell, who is excited to potentially introduce this product, illustrating a significant reduction in footprint and power consumption compared to traditional hybrid storage solutions using hard drives and SSDs for cache. A 9-to-1 reduction in rack space and a 90% reduction in network footprint lead to the possibility of utilizing 50% more servers within the same infrastructure.

The presentation then delves into the specifics of the 122TB SSD, emphasizing its high reliability and efficiency and its design to ensure longevity. It also reveals how the product is not just a drive but an entire ecosystem. The slide show then goes on to detail the product’s inner workings, from the stacking of the drives to the packaging and how the firmware has been adapted to work in the current state of the technology.

Finally, Shadley showcases a liquid-cooled direct-attach storage solution, designed in partnership with NVIDIA for their GB300 platform. The design uses a single-sided cold plate for cooling, allowing hot-swappable drives within a fully liquid-cooled system. This design innovation eliminates the need for fans while maintaining the flexibility and reliability expected from traditional storage. The presentation concludes with a reiteration of Solidigm’s commitment to continuous innovation and its focus on providing solutions that address the evolving needs of AI infrastructure.


Why Storage Matters to AI in 2025 with Solidigm

Event: AI Infrastructure Field Day 2

Appearance: Solidigm Presents at AI Infrastructure Field Day 2

Company: Solidigm

Video Links:

Personnel: Ace Stryker, Steen Graham

Solidigm focused on the evolving role of storage in AI, specifically highlighting its significance in the AI data pipeline through 2025. The presentation emphasizes that the AI workflow involves a series of distinct tasks, each with unique demands on hardware, and that it’s often distributed among different organizations. The presentation identifies six primary stages in this pipeline: ingestion, data preparation, model training, fine-tuning, inference, and archiving. Solidigm’s perspective underscores the increasing data intensity of these steps, especially during data ingestion, inference, and archiving.

Solidigm explores how storage choices impact these stages, distinguishing between direct-attached storage within GPU servers and network-attached storage. Direct-attached storage is optimized for performance-intensive tasks, while network-attached storage is used for larger datasets and offers capacity and cost-effectiveness. Ace Stryker highlights the rising importance of network storage, driven by advances in network bandwidth and the growing size and complexity of AI models and larger datasets for RAG. The key takeaway is that great storage facilitates larger models, longer interactions, and improved outputs on the inference side.

Finally, the presentation showcases a collaboration with Metrum AI, presenting a real-world demo that addresses the challenges of data-intensive inference by offloading model weights and RAG data to SSDs. This allows running 70 billion parameter models on less powerful hardware, which would have been impossible without storage-based offloading, saving on costs and hardware requirements. The demo emphasizes the potential of SSDs to enhance performance in AI applications by reducing GPU memory usage. The collaboration offers insights into the benefits of leveraging storage in AI, as the partnership also showed that offloading to SSD could provide the same or better performance as DRAM in this application of retrieval augmented generation.


« Previous Page
Next Page »