VMware by Broadcom Presents Private AI With Intel at AI Field Day 4

During AI Field Day 4, VMware stepped up to showcase how Intel AMX CPUs can be leveraged for Large Language Models (LLM) on vSphere, presenting a CPU-centric approach to AI tasks commonly handled by GPUs. As discussed by Gina Rosenthal, Earl Ruby of Broadcom (VCF) demonstrated the potential of Intel’s AMX technology in both older Ice Lake and newer Sapphire Rapids systems, achieving model fine-tuning and inference without the use of GPUs. This approach champions using CPUs for AI when feasible, reserving GPU use for scenarios demanding lower latency, and highlights the compatibility requirements for effectively implementing AI with Intel’s hardware in a vSphere environment.


Storage for AI: Data Professional Overview

The use of the term ‘data management’ is subject to varying interpretations between data professionals and storage providers, leading to some confusion when discussing the scope of services. Solidigm highlights the unique storage requirements for AI, as Karen Lopez wrote, emphasizing that AI servers need significantly more capacity and have specific data demands across different stages: Data ingest, Data prep, Training, Checkpointing, and Inference. As data professionals, our understanding of these distinct data workloads enables us to contribute valuably to discussions on data architecture, especially in collaboration with companies like Supermicro, who integrate these storage solutions into their AI server offerings.


Private AI Foundation With NVIDIA: Data Professional Overview

VMware by Broadcom, in partnership with NVIDIA, has introduced the Private AI Foundation, focusing on enhancing in-house data management and AI processing through privacy, choice, cost management, performance optimization, and compliance agility. As highlighted by Karen Lopez, data quality and protection are essential for accurate AI results, prompting data professionals to stay alert to components like vector databases for fast, complex data retrieval. Key takeaways emphasize the importance of not overlooking data fundamentals amidst AI advancements, reminding professionals that AI cannot replace the need for robust data management.


AI Field Day 4 Kicks Off With VMware’s Private AI

Gina Rosenthal is delivering real-time insights on each AI Field Day presentation, starting with her take on VMware’s advancements in AI. In their presentation, VMware highlighted the ease of deploying AI applications on existing infrastructure with a focus on robust security and privacy, leveraging their Private AI and partnerships with industry leaders like NVIDIA. VMware’s session delved into how their solutions, geared towards the generative AI market, are enabling customers to significantly improve operational tasks such as documentation search, boasting a 500% effectiveness increase.


Intel Day Kicks Off at AI Field Day 4

As AI Field Day 4 continues into day two, Gina Rosenthal turns her attention to the capabilities of Intel Xeon CPUs in AI, particularly in inference workflows, as presented by Ro Shah, AI Product Director at Intel. Shah delineated the growing trend towards generative AI with large models, while recognizing that enterprises are more inclined to adopt smaller language models, positioning Xeon as a suitable solution for these scenarios. Intel’s commitment to serving the AI market extends beyond hardware, showcasing their strategy of enabling AI across the board through developer support, tool extensions, and a strong partnership ecosystem.


Why Storage Matters for AI – Solidigm

During AI Field Day 4, Solidigm, alongside partner Supermicro, spotlighted the pivotal role of storage in AI, as discussed by Gina Rosenthal. Ace Stryker of Solidigm emphasized the need to shift from HDDs to solid-state drives, aligning with the trends of chip spending growth and the demand for higher storage in AI servers. Supermicro’s Wendell Wenjen and Paul McLeod further discussed the integration with WEKA and the importance of storage in AIOps, indicating that a substantial portion of Supermicro’s revenue is derived from AI-related ventures.


Cisco Live EMEA and Tech Field Day Extra Embodies “Better Together”

Attending Cisco Live EMEA and Tech Field Day Extra as a single mother on maternity leave, Zoe Rose highlighted the significance of community in shaping careers while juggling the demands of parenthood. Her experience underscores not only the flexibility and inclusion of these industry events but also the support structures that make participation feasible for parents. Rose’s story is an inspiring testament to the power of community and organizational support in fostering professional growth, even amidst the complexities of personal life transitions.


Meraki Switching 17 – Encapsulating ‘Better Together’ by Providing Context

Meraki’s upcoming MS17 features, highlighted at Cisco Live EMEA by Brennan Martin and Alex Burger, are tailored to reduce burdens on IT teams through efficient data aggregation, contextualized insights, and trend visualization. As Zoe Rose discusses, the update is set to streamline operations by introducing dynamic profiles for switchports, enhanced packet capture analysis, and comprehensive device health dashboards, with an API to further customize thresholds and alerts. While a ‘single pane of glass’ for network management remains a complex goal, these changes reflect a product evolution responsive to customer feedback and aimed at alleviating workload, underscoring Cisco’s commitment to user-centric innovation.


Runaway K8 Clusters, Meet Your Match: Platform9’s Elastic Machine Pool (EMP)

Although Jim Czuprynski prefers monolithic architectures like Oracle’s Exadata, as an experienced Oracle DBA he recognizes the common issue of memory over-allocation in containerized environments, driven by the containerization trend. Platform9’s EMP toolset was presented as a solution which introduces an alternative virtualization layer, streamlining resource usage and potentially curbing the over-provisioning habits of Kubernetes engineers. With a focus on enhanced performance and cost efficiency, EMP’s potential to optimize Kubernetes configurations was highlighted at Cloud Field Day, a novel approach eagerly eyed by those in cloud resource management.


Winning the Private (Cloud) War: SoftIron’s Hyperscaled Solutions

Jim Czuprynski reacts to SoftIron’s presentation at Cloud Field Day. They presented a private cloud vision that readily adapts to the challenges of interconnectivity, scalability, and disaster resilience, necessities underscored by recent calamitous events. SoftIron’s tailored solutions, capable of delivering up to 16PB clusters, show their commitment to meeting the elastic computing needs of diverse, growth-oriented IT infrastructures. Read more in this LinkedIn Pulse article!


Cloud to Ground Deployment Magic

Dell Technologies introduced the concept of “Cloud to ground” at Cloud Field Day 19, reflecting the growing interest in cloud repatriation and the desire to bring cloud-like operations to on-premises datacenters. Their APEX Cloud Platform aims to simplify this transition, providing a management solution that integrates with existing tools like Azure console and OpenShift Advanced Cluster Management to automate and manage Cloud OS deployments on Dell hardware. While this innovation aids in the automation and management of on-prem hardware, Ned Bellavance notes that it’s not a full public cloud experience nor a private or multicloud management plane, but rather a significant step towards simplifying hybrid cloud environments for Dell users.


Hardware Vendors Are Providing Better Network Monitoring Than Most Software Vendors

As Thomas LaRock wrote about earlier, 70% of enterprises use private clouds, including those with public-facing websites, pointing out the implicit challenge these companies face in securing their networks against DDoS attacks. Broadcom is innovating in this space with their Trident5-X12 chip featuring the NetGNT engine, which offers real-time traffic analysis and anomaly detection to bolster network security measures. However, despite the advancement in hardware-based security solutions, the onus remains on customers to develop and train the neural network models needed to identify and categorize threats—a task that requires specialized skills.


Better Switch(es) to Fight

Cloud Field Day 19 concluded with Broadcom, as Jim Czuprynski writes in this LinkedIn Pulse article, turning the often-overlooked networking components into a topic of significant impact for cloud infrastructure. Broadcom’s Trident5-X12 chip features smart built-in capabilities to scrutinize every network packet, potentially catching and blocking suspicious traffic such as DDOS attacks in real-time. This technology underscores the importance of robust network security measures, requiring system administrators to be proactive in training and adapting to emergent threat patterns, reinforcing our defense against sophisticated cyber threats.


Dell – Streamlining Cloud – On and Off Premises

Camberley Bates’ LinkedIn Pulse article examines Dell’s latest solutions—APEX Storage for Public Cloud and APEX Cloud Platform—introduced at DTW 2023 and further elaborated during Tech Field Day. These offerings aim to enhance cloud management both on and off-premises, with the SDS-based APEX Storage designed for scalability in AWS, Azure, and eventually GCP, complemented by the APEX Navigator for comprehensive management. The article also highlights APEX Cloud Platforms’ on-premises integrations for OpenShift, AzureStack HCI, and VMware Tanzu that simplify deployment and lifecycle management.


Multi-Cloud: Because Life Isn’t Painful Enough

Thomas LaRock’s LinkedIn Pulse article wittily addresses the complexities of multi-cloud management and introduces Dell’s APEX Cloud Platform with APEX Block Storage, focusing on how it can ease the burdens of handling high-performance workloads across different cloud providers, despite the potentially higher costs associated with enterprise-level storage solutions and multi-cloud management tools like APEX Navigator.


When CPUs/GPUs Are Not Enough: NeuroBlade’s SPU

Drawing from experience with high-performance Exadata DBMs, Jim Czuprynski considers the challenges in running complex analytic or ML workloads in containerized databases like Apache Spark. NeuroBlade has stepped in with a potential solution for these scalability issues in massive, containerized environments by introducing the SQL Processing Unit (SPU), designed to offload and accelerate these demanding tasks. Touted to boost analytical performance by up to 10x, the SPU exemplifies hardware innovation tailored for the needs of private cloud computing, heralding a promising future for data-intensive workloads.


Kubernetes With Platform9’s Elastic Machine Pool

Platform9’s Elastic Machine Pool (EMP) for AWS addresses the critical issue of EKS cluster underutilization, offering Kubernetes operations enhanced efficiency and cost-effectiveness. By dynamically rebalancing resources and leveraging advanced virtualization, EMP can potentially cut AWS costs in half for EKS users, presenting a considerable advantage for both DevOps and #FinOps teams. Beyond the cost benefits, EMP’s intelligent resource management promotes operational efficiency, allowing teams to devote more time to innovation instead of the manual oversight of cloud resources, according to Jon Myer.


The Dell Alternative to Cloud

In this LinkedIn Pulse article, Joep Piscaer analyzes Dell Technologies’ multifaceted approach to multi-cloud, which includes APEX Block Storage—a software-defined, mature storage solution that extends on-prem software to cloud providers, boosting performance and resilience—and the APEX Cloud Platform, a standardized on-prem hardware and software package that integrates cloud features and operations into Dell’s ecosystem. Piscaer commends Dell’s strategy of not directly competing with cloud service giants but rather expanding its ecosystem within the cloud and enriching the on-premises infrastructure with cloud-like functionalities, providing significant value to its existing customer base.


SoftIron – A HyperCloud That Is Not Your Typical HCI

Camberley Bates describes SoftIron’s distinctive approach within the Hybrid Cloud Infrastructure market, highlighting its appeal to sectors with zero-trust and high-security requirements. SoftIron stands out by manufacturing its own hardware for compute, network, and storage in the US and Australia, with a software foundation geared towards high-security standards. Their systems can scale significantly beyond the usual limitations of HCI, supporting large clusters with high performance, even in HPC or image processing contexts. The article notes SoftIron’s unique architecture, including a stateless device setup and control plane integration in top of rack networking nodes, and mentions the intention to further investigate their data management and system interface.


Saving Money in EKS Is a Good Start

Cloud elasticity, allowing for scaling out and in, can lead to cost inefficiency when resources remain underutilized, a situation exacerbated in Kubernetes environments with studies citing overprovisioning as a significant issue. Ned Bellavance discusses Platform9’s Elastic Machine Pool (EMP), which seeks to address excessive resource allocation in EKS by optimizing node placement and memory on AWS EC2 instances, potentially leading to substantial cost savings. However, while EMP is a clever intermediary fix, the underlying challenge of workload right-sizing and application team behavior remains, signifying the need for a more profound, long-term solution.