As Allyson Klein writes, VAST Data is revealing a major shift in AI strategy, joining forces with NVIDIA and Supermicro to bolster its role as a forward-thinking AI data platform. By embracing a novel architecture that eschews traditional x86 design for a powerful GPU-centric platform with NVIDIA DPUs, VAST Data is poised to redefine data storage, promising significant energy efficiency and enhanced performance for AI workloads. The company’s move shifts the AI training landscape towards GPU-native frameworks and sets VAST Data as a key innovator in an infrastructure industry ripe for disruption.
Does Storage Matter in AI Inferencing? What About the SSD?
Keith Townsend reacts to Solidigm’s presentation at AI Field Day, considering the role of storage systems in AI inferencing and the impact of SSD selection on AI system design. This video underscores the significance of considering storage performance and reliability when devising robust AI inferencing architectures. Solidigm’s discussion reflected a deeper industry focus on the intricate relationship between storage solutions and AI capabilities, suggesting that the choice of SSDs could be a pivotal factor in optimizing AI inferencing operations.
VMware Private AI at AI Field Day
VMware’s presentation with Intel at AI Field Day centered on optimizing on-premises AI workloads, highlighting the capability of Intel Sapphire Rapids CPUs with Advanced Matrix Extensions (AMX) to efficiently perform large language model (LLM) AI inference, traditionally a task for GPUs. Demonstrating that AI can be resource-effective on CPUs, the discussion covered the technical prerequisites for harnessing AMX in vSphere environments and the ongoing integration of these accelerators into popular AI frameworks. With CPUs increasingly capable of handling AI tasks through built-in matrix math acceleration, VMware showcases a sustainable, cost-effective approach, potentially reshaping the hardware strategies for mixed workload servers. Read more in this analyst note for The Futurum Group by Alastair Cooke.
Gemma and Building Your Own LLM AI
At AI Field Day 4, Intel invited the Google Cloud AI team to showcase their Gemma large language model (LLM), revealing insights into the advanced infrastructure used for building such models on Google Cloud. The presentation underlined Gemma’s efficiency with fewer parameters for inference, highlighting Google Cloud’s strength in analytics and AI, particularly in managing differing resource needs between model training and application inference phases. Google Cloud’s integration of AI in products was illustrated with Google Duet, an AI-based assistant that aids in software development, exemplifying the potential future where AI handles more coding tasks, freeing up developers for high-level problem-solving and design. Read more in this analyst note for The Futurum Group by Alastair Cooke.
Intel Xeon CPUs on VMware vSphere – A Powerful and Cost-Effective Twosome for AI/ML Workloads
With AI ingrained in our daily routines, Forward Networks delivered a strategic approach at Networking Field Day, demonstrating how even complex networking data can be made manageable through AI integration. Their platform uses a data-first principle, enabling AI to interact effectively with a digital twin of network infrastructure, simplifying tasks for network engineers. The innovative AI Assistant within Forward Networks’ ecosystem assists in constructing queries for the Network Query Engine, fostering trust through verifiable, human-readable outputs, and providing a gateway for more intuitive network management. Read more in this article by Sulagna Saha on Gestalt IT.
Forward Networks Now With More AI
Amidst the ubiquitous influence of artificial intelligence in our daily lives, Forward Networks stood out to Josh Warcop by integrating AI with a robust data-first approach, as showcased at Networking Field Day. Their creation of a digital twin for network infrastructure ensures precise data, allowing AI to effectively communicate with it, bypassing the challenges of manual data modeling. Forward Networks’ Network Query Engine, equipped with an AI Assistant, generates human-readable queries, bridging the gap between complex data systems and operator expertise and fostering trust in AI’s capabilities within network management.
Let Someone Else Install Your Campus Network
As Aaron Conaway posted, Nile showcased their comprehensive campus network solutions at Networking Field Day, emphasizing their end-to-end service that extends beyond hardware delivery to include detailed network planning and on-site evaluations. Their approach ensures not only installation but also up-to-date firmware and secure configurations managed via a cloud-based system, challenging the traditional “install now, secure later” methodology. Nile further distinguishes itself with ongoing management, monitoring, and proactive problem and replacement services, offering an innovative model for organizations lacking significant capital expenditure budgets.
Nile – Network as a Service – Considerations for Adoption
Attending Network Field Day gave Ryan Lambert a first-hand look at Nile’s innovative Network as a Service (NaaS) offering, a service model that manages and maintains a customer’s network infrastructure. Nile stands out by not only managing and monitoring networking equipment but also engaging with partners for on-site surveys to ensure top-notch performance, guaranteeing service levels for the hassle-free operation of customer networks. However, it’s important for customers to understand the shared responsibility in such managed solutions, particularly regarding the physical environment and specific requirements, ensuring Nile’s service guarantee aligns with their needs.
No Inventory, No Proper Management
At Networking Field Day 34, Forward Networks, Inc. showcased the substantial advantages of implementing a digital twin for network management. The digital twin provides a dynamic and thorough inventory by constantly analyzing the network to reflect its current state, automatically updating documentation, and enhancing compliance and automation efforts. Aaron Conaway emphasized the crucial need for up-to-date network visibility, asserting that reliable network management is impossible without a comprehensive and current understanding of one’s network infrastructure.
Defeating Data Gravity? – Hammerspace
According to Keith Townsend, Hammerspace presented a compelling argument for a shift in overcoming data gravity by moving data closer to accelerated computing resources at AI Field Day. Their solution, a parallel file system, acts as a bridge between dispersed data sources, offering a unified metadata view that streamlines data preparation for AI tasks. While Hammerspace’s technology appears to enhance user experience, it also requires strategic GPU placement and considerations around data governance and movement across geopolitical boundaries.
Forward Networks – The Tale of an Aptly Named Company
In Ryan Lambert’s LinkedIn Pulse article he delves into the capabilities of Forward Networks, a company he says lives up to its name. The company creates a Digital Twin of network data, encompassing router configurations, route tables, and firewall policies. Lambert emphasizes the practicality of this approach, enabling users to gain insights into network behavior, potential vulnerabilities, and operational breakdowns.
Taking on AI Inferencing With 5th Gen Intel Xeon Scalable Processors
Intel’s 5th Generation Xeon Scalable Processor, known as Emerald Rapids, offers an advantageous solution for AI inferencing, providing a compelling alternative to GPUs in certain applications. Highlighted during the AI Field Day event, Intel showcased the processor’s suitability for general-purpose AI workloads, especially for private AI deployments requiring lower latency and mixed workloads. In his presentation, Ro Shah illustrated that Xeon CPUs are well-equipped to handle AI models with fewer than 20 billion parameters, making them a cost-effective and efficient choice for many enterprises. Read more in this article from Gestalt IT.
Replacing Human-Driven Automation With One That Is Event-Driven, With Cisco
In a move to enhance network management efficiency, Cisco is replacing the existing human-driven automation with an innovative event-driven approach. Network controllers centralize operations by maintaining an inventory of network states, and event-driven automation streamlines this further by automating responses to network changes, reducing manual intervention and speeding up the resolution process. Cisco’s event-driven solutions were explored in detail at Tech Field Day Extra at Cisco Live EMEA 2024, promising elevated operational capabilities with a focus on observability and swift auto-configuration. Read more in this article from Gestalt IT.
Policy Management Gets Easier With Cisco AI Assistant
Cisco’s AI Assistant, unveiled at Tech Field Day Extra at Cisco Live EMEA 2024, promises to revolutionize policy management within Secure Access, offering swift and reliable creation and enforcement of security policies. Emphasizing simplicity and risk reduction, the AI tool handles English language prompts to generate comprehensive policies, demonstrating its capability to securely augment and automate IT administrator tasks. Showcasing AI’s monitoring role, Cisco Secure Access also ensures protected use of AI tools within organizations, preventing sensitive data breaches and incorporating intelligent user access control.
Insights From the AI Field Day: A Futurum Group Overview
In this LinkedIn Pulse article, Paul Nashawaty of The Futurum Group summarizes all of the AI Field Day presentations, highlighting VMware’s deep dive into Private AI in collaboration with industry giants like NVIDIA and IBM, and Intel’s focus on deploying AI inference models with Xeon CPUs across diverse environments. Next-generation AI-infused storage solutions from Solidigm and SuperMicro underscored the critical role of optimized storage in AI, while Vast Data focused on addressing the growing data demands of AI and HPC workloads. Google Cloud’s session on AI platforms and infrastructures showcased innovative approaches with Kubernetes at the core, paving the way for accessible and powerful AI development and deployment.
“AI Assist” Is Better Than “AI Do It for Me”
John Herbert feels cautious towards AI, particularly when entrusting it with network control, advocating for human oversight to validate AI-generated solutions. But he appreciates Forward Networks’ AI Assist feature for creating natural language queries that help users engage with the company’s Network Query Engine (NQE), while retaining the option for manual evaluation and tweaking. Furthermore, Herbert finds value in the Summary Assist tool, which provides natural language explanations of NQE queries, enhancing user understanding and trust in the AI’s output.
From Server Farm to Table: How Nature Fresh Uses AI and CPUs to Improve Crop Yields
At AI Field Day 4, Keith Bradley from Nature Fresh Farms highlighted the practical application of AI in agriculture, revealing how the company utilizes AI models and data from IoT devices to boost crop yields significantly. As Jim Davis writes, Nature Fresh Farms’ unique approach relies heavily on edge computing with Intel Xeon CPUs for AI inferencing, debunking common misconceptions that AI always requires cloud connectivity or GPU resources. Bradley’s presentation emphasized the critical importance of IT system reliability in the agricultural sector, where even brief periods of downtime can result in substantial losses.
Qlik Presents About AI on AIFD4
At AI Field Day, Qlik unveiled their comprehensive AI strategy, focusing on enabling customers to tackle the complexity and scale of AI integration into their operations with solutions like predictive analytics, AutoML, and their 170K-strong model portfolio. With initiatives like Qlik Staige, the company facilitates the use of AI for those new to the field, using established algorithms and bringing structured and unstructured data together for added value. Read more about the Qlik presentation in this LinkedIn Pulse article by Gina Rosenthal!
VAST Data Upends Storage in the AI Era
Allyson Klein offers an insightful look at VAST Data’s innovative approach to the AI data pipeline, a process crucial for preparing data for AI training, through their advanced NAS solution spotlighted at AI Field Day. VAST Data’s platform addresses the challenges organizations face with data prep and movement in AI, proven by their success in the HPC sphere and their natural progression into supporting AI training clusters.
