Cloudian announced its integration with Nvidia Magnum IO GPUDirect storage technology, using RDMA parallel transfer, and putting file and object storage data on an equal footing in serving data to Nvidia GPUs.
This “simplifies the management of AI training and inference datasets – at petabyte and exabyte scales – while reducing costs by eliminating the need for complex data migrations and legacy file storage layers.”
Michael Tso
Cloudian CEO Michael Tso said of the move: “For too long, AI users have been saddled with the unnecessary complexity and performance bottlenecks of legacy storage solutions. With GPUDirect Storage integration, we are enabling AI workflows to directly leverage a simply scalable storage architecture so organizations can unleash the full potential of their data.”
Rob Davis, Nvidia VP of storage technology, said: “Fast, consistent, and scalable performance in object storage systems is crucial for AI workflows. It enables real-time processing and decision-making, which are essential for applications like fraud detection and personalized recommendations.”
GPUDirect, up until now, has been an Nvidia file-focused protocol that eliminates needless file data movement within a server. Normally a storage server CPU copies data from a storage drive into its memory and then authorizes its transmission over a network link into a GPU’s DRAM. The GPUDirect protocol enables a direct connection to be made between the storage drive and GPU memory, cutting out file movement into the storage server’s memory, lowering file data access latency and keeping GPUs busy rather than having them wait for file IO.
Cloudian diagram
However, GPUs still wait for object data IO, rendering the use of object data for GPU AI processing impractical. Cloudian and MinIO have broken this bottleneck, enabling stored object data to be used directly in AI processing rather than indirectly via migration into file data, which can then travel via GPUDirect to the GPUs.
The HyperStore GPUDirect object transfer relies on Nvidia ConnectX andBlueField DPU networking technologies, enabling direct communication between Nvidia GPUs and multiple Cloudian storage nodes.
Cloudian diagram. Object GPUDirect works like this: (1) Data requests are initiated via the S3 API, (2) Instead of routing through system memory and CPU, data moves directly to GPU memory. (3) RDMA enables parallel transfer from multiple Cloudian nodes
Cloudian says its GPUDirect-supporting HyperStore delivers over 200 GBps from a single system with performance sustained over a 30-minute period without the use of data caching. It “slashes CPU overhead by 45 percent during data transfers.”
There are no Linux kernel modifications, eliminating “the security exposure of vendor-specific kernel modifications.”
CMO Jon Toor wrote in a blog post: “Cloudian has spent months collaborating closely with Nvidia to create GPUDirect for Object Storage. With GPUDirect for Object, data can now be consolidated on one high-performance object storage layer. The significance is simplification: a single source of truth that connects directly to their GPU infrastructure. This reduces cost and accelerates workflows by eliminating data migrations.”
Cloudian HyperStore’s effectively limitless scalability enables AI data lakes to grow to exabyte levels, and its centralized management ensures “simple, unified control across multi-data center and multi-tenant environments.”
The company says it accelerates AI data searches with integrated metadata support, allowing “easy tagging, classification, and indexing of large datasets.” It says file-based systems depend on “rigid directory structures and separate databases for metadata management.” With Cloudian and other object stores, metadata is natively handled within the object storage platform, “simplifying workflows and speeding up AI training and inference processes.”
Michael McNerney, SVP of Marketing and Network Security at Supermicro, said: “Cloudian’s integration of Nvidia GPUDirect Storage with the HyperStore line of object storage appliances based on Supermicro systems – including the Hyper 2U and 1U servers, the high-density SuperStorage 90-bay storage servers, and the Simply Double 2U 24-bay storage servers – represents a significant innovation in the use of object storage for AI workloads. This will enable our mutual customers to deploy more powerful and cost-effective AI infrastructure at scale.”
David Small, Group Technology Officer of Softsource vBridge, said: “Cloudian’s GPUDirect for Object Storage will simplify the entire AI data lifecycle, which could be the key to democratizing AI across various business sectors, allowing companies of all sizes to harness the power of their data. We’re particularly excited about how this could accelerate AI projects for our mid-market clients who have previously found enterprise AI solutions out of reach.”
Nvidia is working with other object storage suppliers. Toor tells us: “GPUDirect is an Nvidia technology that we integrate with. We did foundational work on the code to support object, so we believe we are on the forefront. But like any industry standard, it will be driven by ecosystem growth. And that growth is a good thing. It will cement the position of object storage as an essential enabler for large-scale AI workflows.”
MinIO announced its AIStor with GPUDirect-like S3 over RDMA earlier this month. Scality introduced faster object data transfer to GPUs wIth its Ring XP announcement in October.
Cloudian HyperStore with Nvidia Magnum IO GPUDirect Storage technology is available now. To learn more, explore Cloudian’s views on GPUDirect in itslatest blog post.
Dell is extending its AI Factory range at the Supercomputing 2024 (SC24) event in Atlanta with Data Lakehouse and AI-focused server announcements.
It says the Data Lakehouse updates provide modern architectures for efficiently managing and analyzing data for AI tasks. It will expand to include Apache Spark for distributed data processing at scale. This Spark integration “will provide significant efficiency gains, with a unified approach for data analytics, management, processing and analysis for faster, more actionable insights.”
Arthur Lewis, president, Infrastructure Solutions Group, Dell Technologies, stated: “Getting AI up and running across a company can be a real challenge. We’re making it easier for our customers with new AI infrastructure, solutions and services.”
Dell’s Data Lakehouse was announced back in March and featured the Starburst Trino query engine, Kubernetes-orchestrated lakehouse system software, and scale-out S3-compatible object storage based on Dell’s ECS, ObjectScale, or PowerScale storage products. The Starburst code included so-called Warp Speed technology, with Apache Lucene indexing and caching technology.
Apache Spark is a distributed data processing engine designed for batch processing, real-time streaming, machine learning, and advanced analytics. Starburst Trino is a distributed SQL query engine intended for interactive and ad hoc queries across heterogeneous data sources.
The company boosted lakehouse query speed, added and upgraded connectors, and improved monitoring and security in June.
Dell announced two new servers in 4RU chassis for its 19-inch IR5000 rack:
The liquid-cooled PowerEdge XE9685L has dual 5th Gen AMD EPYC CPUs paired with the Nvidia HGX H200 or B200 platforms, and up to 12 PCIe gen 5.0 slots. It has customizable configurations designed for AI, machine learning, high performance computing (HPC) and other data-intensive workloads. There can be up to 96 Nvidia GPUs per rack.
The air-cooled PowerEdge XE7740 has dual Intel Xeon 6 CPUs with P-cores and up to 8 double-wide accelerators, including Intel Gaudi 3 AI accelerators or Nvidia H200 NVL, or up to 16 single-wide accelerators, like the Nvidia L4 Tensor Core GPU. It is intended for GenAI model fine-tuning or inferencing and analyzing large datasets.
Dell has a new XE server coming that will support Nvidia’s GB200 Grace Blackwell Superchip, with up to 144 GPUs per IR7000 rack (see bootnote).
It is also introducing:
Dell AI Factory with Nvidia gets HGX H200 and H100NVL support options, providing up to 1.9x higher performance compared to the HGX H100.
Agentic RAG with Nvidia uses PowerEdge, PowerScale and Nvidia AI Enterprise software and GenAI tools including NeMo Retriever microservices and the Nvidia AI Blueprint for multimodal PDF data extraction.
Dell Validated Designs for AI PCs open source guides include NPU technology. Developers can customize the modular designs to integrate features like LLMs, vision, text and speech into applications.
It says the Agentic RAG with Nvidia design “helps organizations with large datasets use AI agents to improve RAG workflow performance, handle complex queries, and deliver higher quality outcomes.”
Dell’s Professional Services can, it says, ease AI adoption with Advisory and Implementation Services for Sustainable Data Centers, Data Management Services, Design Services for AI Networking, and Implementation Services for ServiceNow Now Assist.
Availability
The PowerEdge XE9685L will be globally available Q1 calendar 2025.
The PowerEdge XE7740 will be globally available Q2 calendar 2025.
Updates to the Dell Data Lakehouse are globally available now.
Dell Validated Designs for AI PCs are globally available now.
Dell Generative AI Solutions with Nvidia – GPU update will be available Q4 calendar 2024.
Dell Generative AI Solutions with Nvidia – Enterprise-scale RAG is globally available now.
Dell Data Management Services are available in select countries now.
Dell Services for Sustainable Data Centers are available in select countries now.
Dell Design Services for AI Networking are available in select countries now.
Dell Implementation Services for ServiceNow Now Assist are available in select countries now.
Read more in a server blog here and, on AI PC adoption, here.
Bootnote
The Dell Integrated Rack Scalable Systems (IRSS) program is a turnkey factory integration program that delivers fully loaded, plug-and-play rack-scale systems with Dell Smart Cooling. IRSS has one-call service and support options for the entire rack and Dell takes care of the packaging waste and recycling.
The IR5000 is a standard 19-inch rack engineered for general-purpose datacenter applications. The IR7000 is a 50U standard rack incorporating liquid cooling, with the ability for near 100 percent heat capture, and has 21-inch wide sleds. This rack is for large-scale HPC and AI workloads requiring high power.
MLCommons has revealed new results for MLPerf Training v4.1, which the org says highlights “the strong alignment between the benchmark suite and the current direction of the AI industry.”
MLCommons is an open engineering consortium and a leader in building benchmarks for AI data processing across different segments, including AI training. It says the MLPerf Training benchmark suite measures how fast systems can train models to a target quality metric.
MLPerf Training v4.1 includes preview category submissions using new and upcoming hardware accelerators Google Trillium TPUv6 and Nvidia Blackwell B200.
Hiwot Kassa
“As AI-targeted hardware rapidly advances, the value of the MLPerf Training benchmark becomes more important as an open, transparent forum for apples-to-apples comparisons,” said Hiwot Kassa, MLPerf Training working group co-chair. “Everyone benefits: vendors know where they stand versus their competitors, and their customers have better information as they procure AI training systems.”
The benchmark suite comprises full system tests that stress models, software, and hardware, for a range of machine learning (ML) applications. The open source and peer-reviewed benchmark suite promises a level playing field for competition that “drives innovation, performance, and energy efficiency for the entire industry,” says MLCommons.
The latest Training v4.1 results show a substantial shift in submissions for the three benchmarks that represent generative AI training workloads: GPT-3, Stable Diffusion, and Llama 2 70B LoRA, with a 46 percent increase in submissions in total across these three.
The two newest benchmarks in the MLPerf Training suite, Llama 2 70B LoRA and Graph Neural Network (GNN), both had notably higher submission rates: a 16 percent increase for Llama 2, and a 55 percent increase for GNN. They both also saw significant performance improvements in the v4.1 round compared to v4.0, when they were first introduced. There was a 1.26x speedup in the best training time for Llama 2, and a 1.23x speedup for GNN.
The Training v4.1 round includes 155 results from 17 submitting organizations, including ASUSTeK, Azure, Cisco, Clemson University Research Computing and Data, Dell, FlexAI, Fujitsu, GigaComputing, Google, Krai, Lambda, Lenovo, NVIDIA, Oracle, Quanta Cloud Technology, Supermicro, and Tiny Corp.
David Kanter
“We would especially like to welcome first-time MLPerf Training submitters FlexAI and Lambda,” said David Kanter, head of MLPerf at MLCommons. “We are also very excited to see Dell’s first MLPerf Training results that include power measurement. As AI adoption skyrockets, it is critical to measure both the performance and the energy efficiency of AI training.”
On the new accelerator competition between Google and Nvidia, Karl Freund, founder and principal analyst at Cambrian-AI Research, wrote: “The Google Trillium TPU delivered four results for clusters ranging from 512 to 3,072 Trillium accelerators. If I contrast their performance of 102 minutes for the 512-node cluster, it looks pretty good until you realize that the Nvidia Blackwell completes the task in just over 193 minutes using only 64 accelerators. When normalized, always a dangerous and inaccurate math exercise, that makes Blackwell over 4x faster on a per-accelerator comparison.”
This may well illustrate that Nvidia is set to continue leading the AI processing market going forward.
MLCommons was founded in 2018, and has over 125 industry members.
Data orchestrator Arcitecta is partnering with Wasabi Technologies to allow organizations to integrate Wasabi’s “hot” cloud storage into their workflows.
Arcitecta provides MediaFlux Universal Data System products covering file and object data storage and management software. This technology has a single namespace and tiering capability covering on-premises SSDs, disk and tape, and the public cloud, with a data mover and metadata database. While Arcitecta says integrating cloud storage into an organization’s workflow can offer many advantages, such as improved scalability, accessibility, and collaboration, it can be “challenging” with concerns around data security, compliance, migration, compatibility, performance, latency, and data governance.
Data management across multiple environments presents more complexity. Storing data on-premises and in private or public clouds can create silos, making it difficult to have a unified view of the data. In addition, ensuring that data remains consistent and synchronized across different environments can become tricky, especially when updates occur simultaneously in multiple locations.
Arcitecta diagram with Wasabi cloud in bottom right position
Arcitecta says it can address these issues, claiming this will make it simpler for organizations to integrate Wasabi’s cloud storage into workflows. The Wasabi cloud appears like any other storage managed by Mediaflux, which acts as a gateway, allowing users to access Wasabi cloud storage and all their data, regardless of where it resides, through one unified view.
Jason Lohrey
“Organizations need fast and easy access to their data and, for distributed workflows in particular, cloud storage is a great option,” said Jason Lohrey, CEO and founder of Arcitecta. “Our partnership with Wasabi enables customers to seamlessly store, manage, and access their data from Wasabi’s cloud into their day-to-day workflows, all within a single, unified view and namespace, to accelerate decision-making.”
Mediaflux allows customers to use any mix of storage technologies to best meet their requirements, whether on-premises, in a public or private cloud, or a hybrid of both. Users have a global view of all the data, no matter what storage it is on, with identity and policy-based access controls.
Laurie Coppola Mitchell, SVP of global alliances and partner marketing at Wasabi Technologies, said: “Wasabi offers high-performance, secure cloud storage at a fraction of the cost of other providers. By integrating cloud storage with Mediaflux, organizations can optimize costs, enhance security, and maintain control over their IT infrastructure.”
As a use case example, the two companies have illustrated how their partnership can work in higher education and research institutions. More details here.
Laurie Coppola Mitchell
Pricing for the joint Mediaflux and Wasabi offering is predictable, transparent, and straightforward, with Mediaflux licensing decoupled from the volume of data stored so organizations can affordably scale storage needs to hundreds of petabytes without financial strain. Wasabi customers pay one “low rate” for capacity, with no hidden fees or egress charges. The offering is available now and can be purchased through Arcitecta’s Mediaflux channels. Wasabi cloud storage can be purchased online at a per TB rate per month.
At last month’s IT Press Tour of Boston and Massachusetts, Wasabi said it had reached 100,000 customers after continuing sales growth.
Comment
We covered Arcitecta and Wasabi’s partnership in a brief note yesterday. It is part of a fundamental change in the unstructured data management market that deserves more attention.
Arcitecta and Hammerspace both provide file-based distributed, multi-vendor, multi-storage-media, and on-prem/public cloud location file and object management and orchestration, providing local-speed access to centrally managed files and objects in a global namespace. They partially overlap with Komprise, coming from a file lifecycle management background, and also Datadobi with its data migration background. In addition, there is a partial functional overlap with cloud file services suppliers CTERA, Nasuni, and Panzura, which each serve cloud-based file access to distributed locations.
The growth in GenAI training and inference has made access to the entirety of an organization’s unstructured data more important, both for better trained models and for retrieval-augmented generation (RAG), which needs access to an organization’s private data. Because of this, previous file lifecycle management and migration specialists are encountering a need to become GenAI unstructured data supply generalists. GenAI is encouraging these suppliers to come out from their niches into a more open market.
That is because these GenAI data sourcing needs drive a requirement for organizations to understand where their data is stored and to make it available to AI workloads. That is providing a rising data storage market demand for products that can provide maps of an organization’s data and move it or make it available to GenAI. It is for this reason that Arcitecta, CTERA, Datadobi, Data Dynamics, Egnyte, Hammerspace, Komprise, Nasuni, and Panzura will find themselves meeting each other in bids as customers cast about for products and services that help them map, organize, and orchestrate their distributed unstructured data for GenAI use.
Hitachi Vantara and Hammerspace have made a deal to combine multi-source data orchestration facilities from Hammerspace with Hitachi Vantara’s enterprise-level file and block storage products, targeting data provision for GenAI workloads.
Hitachi Vantara is a long-standing supplier of block, file, and object storage to large and medium enterprises. It is unifying its portfolio under a Virtual Storage Platform (VSP) One concept with recent VSP One block and object appliances as part of this initiative. The company has also introduced its IQ product with NVIDIA-specific GPU server targets, like BasePOD, integrated with Hitachi Vantara storage systems. This includes HCFS, rebranded WekaFS file system software (Weka Data Platform), a POSIX-compliant parallel file system integrated with Hitachi Vantara’s object storage.
Integration with Hammerspace’s Global Data Platform adds another dimension to this storage portfolio.
Kimberly King
Kimberly King, SVP of strategic partners and alliances at Hitachi Vantara, said: “This partnership represents a strategic move to enhance our portfolio with advanced and intelligent capabilities to orchestrate data, addressing the growing demand for robust and scalable AI solutions.”
Hammerspace’s Global Data Platform (GDP) presents a global namespace, in which data is placed, orchestrated, made available for use, and accessed as if it were local. It uses parallel NFS-based file system software to manage file and object data in globally distributed locations, across SSDs, disk drives, public cloud services, and tape media. Hitachi Vantara customers can now use GDP to access other suppliers’ storage in a single environment, and use it in GenAI training and inferencing applications.
By using Hammerspace’s advanced data orchestration capabilities, Hitachi Vantara customers will be able to unify data access, provide data for in-place AI, and consolidate data into a central data lake for AI workload access. The two companies say that Hammerspace’s “standards-based parallel file system will ensure that all your performance requirements are met for your AI workloads and will scale easily and linearly as your environment grows.”
David Flynn, CEO and founder, Hammerspace, said: “We are thrilled to announce our strategic partnership with Hitachi Vantara. This collaboration represents a significant step forward in our mission to revolutionize data orchestration and storage solutions.”
David Flynn
Hitachi Vantara is the first enterprise storage system supplier to ally directly with Hammerspace. It, along with NetApp and others, has existing storage array software capabilities to virtualize other suppliers’ arrays and treat their stored data as part of, in this case, the Hitachi SVOS environment. The Hammerspace deal extends this concept much further and works higher up the stack, so to speak. More information on Hitachi Vantara’s offerings in AI here.
Bootnote
Hammerspace is contributing its acquired RozoFS Mojette transform-based erasure coding to Linux for client-side erasure coding.
PEAK:AIO, the UK-headquartered software-defined storage platform built for AI and GPU computing, has named a new CEO amid a claimed 400 percent growth in US sales over the last year.
As a private company, is not required to provide any figures for American sales, so the initial base is unknown.
PEAK:AIO says Roger Cummings will now officially be the company’s president and CEO, replacing Mark Klarzynski, who will now hold the title of founder. Cummings is located in Chicago, Illinois.
Roger Cummings
According to Cummings’ LinkedIn profile, he has been serving as PEAK:AIO CEO since May, and the PEAK:AIO website leadership section still has Klarzynski pictured as the “CEO and founder.”
Under Cummings’ leadership, “PEAK:AIO has forged impactful partnerships with key industry leaders, and PEAK:AIO’s influence will be evident at SuperComputing 24, where numerous stands will showcase PEAK:AIO-based solutions, reflecting its rapid adoption across industries,” said the provider.
SuperComputing 24 takes place over November 17-22 in Atlanta, Georgia. PEAK:AIO partner stands showing off its technology include Western Digital, Hammerspace, Solidigm, and Boston Limited.
At the back end of last year, PEAK:AIO had its top-tier GPU data delivery performance validated through testing by HPE.
The company provides storage software for the AI market based on storage servers from vendors such as Dell, HPE, Supermicro, and Gigabyte. It aims to deliver the same or better performance than multi-node parallel file systems such as IBM’s Storage Scale and DDN’s Lustre, using basic NFS, NVMe SSDs, and rewritten RAID software.
In the high-performance computing market, PEAK:AIO focuses on energy efficiency and scalability. “Our transformation from a prominent EMEA leader in AI data infrastructure to a key force in North America underscores our dedication to meeting the sophisticated demands of today’s data-driven world,” the company said.
“Being at the forefront of PEAK:AIO’s transformative journey and witnessing our extraordinary growth in North America is exhilarating,” said Cummings. “Our partnerships underscore the trust and demand for PEAK:AIO’s technology in powering next-gen AI and GPU workflows.”
He added: “The market’s rapid embrace of our innovative, energy-efficient solutions underscores a significant shift toward sustainable data performance. I look forward to deepening our collaborations and propelling PEAK:AIO to new heights in this evolving landscape.”
Data orchestrator Arcitecta has partnered with Wasabi Technologies to enable organizations to seamlessly and transparently integrate Wasabi’s cloud storage into their workflows. Arcitecta’s data management platform enables organizations to leverage all their data across multiple environments. Its flagship product, Mediaflux, acts as a gateway, allowing users to access Wasabi cloud storage and all their data regardless of where it resides through one unified view.
…
Catalogic’sCloudCasa Kubernetes backup and management business announced expanded support for open-source virtualization platform KubeVirt, that enables the management of virtual machines (VMs) as Kubernetes-native resources. This integration allows users to manage the backup and restoration of both VMs and containerized workloads in hybrid environments.
…
Cloudera has launched Cloudera Copilot for Cloudera AI at EVOLVE24 Paris. It enables enterprises to get trusted data, analytics, and AI applications into production faster. Cloudera Copilot helps users write high-quality, consistent code with:
Automated code generation, data transformation, and troubleshooting.
Reliable coding assistance across languages, libraries, and workflows.
Insight and on-demand guidance to maintain high coding standards and minimize errors.
…
Cloudera announced a definitive agreement to acquire Octopai, a data lineage and catalog platform. Octopai’s automated solutions for data lineage, discovery, cataloging, mapping, and impact analysis complement Cloudera’s data architecture strategy. With Octopai’s metadata management and multi-dimensional data lineage, Cloudera customers gain visibility across diverse data environments, enabling trusted data to power AI, predictive analytics, and other decision-making tools. Customers can also expect enhanced data discoverability, data quality, data governance, and migration assistance.
…
New York-based enterprise data management startup Conduktor has raised $30 million in Series B funding to grow its real-time data streaming business. The funding round, led by RTP Global, will accelerate the startup’s expansion into the US and product development to support data flows and new use cases – including AI. Its scalable framework integrates with Apache Kafka and platforms such as Confluent, Amazon MSK, and Microsoft Azure. Conduktor’s platform is used by global organizations like BMW Group, Capital Group, Flix, SIX Group, DraftKings, and Lufthansa for data streaming across their distributed organizations. The global stream processing market is expected to grow from $22.34 billion in 2023 to $185 billion by 2032.
…
Cloud file services supplier CTERA updated the visualization and analytics CTERA Insight Service for its CTERA Global File System, providing next-generation data observability. It delivers end-to-end observability across edge locations, enabling organizations to make data-driven decisions that enhance security, optimize storage, and reduce costs. The new generation of CTERA Insight empowers enterprises to:
Monitor activity – Real-time dashboard visualizations of file activity across filers and cloud, with granular detail and performance metrics.
Discover usage – Customers gain a comprehensive view into storage consumption, file types, access distribution, and ownership, enabling end users to pinpoint opportunities for storage optimization, better predict capacity requirements, and maintain compliance.
Investigate security incidents – CTERA Insight provides long-term audit logging, enabling detailed filtering and activity forensics and in-depth investigations of suspicious behaviors, helping enforce security measures across the organization.
Curate datasets for AI – Enterprises can identify and refine high-quality datasets from vast volumes of unstructured data to advance large language model (LLM) and machine learning (ML) initiatives.
…
DataCore Software’s annual 2024 State of Storage Survey showed:
Data storage capability gaps and management hurdles – More than half (54 percent) of respondents stated that they prefer to keep the data of their organization central in local datacenters and/or decentralized across distributed facilities.
Lack of essential storage capabilities – Ninety percent of respondents indicated that their current storage infrastructure is missing critical features, elaborating that the most pressing gaps are high availability (26 percent), sufficient storage performance (25 percent), and tamper-proof data protection (23 percent).
Impacts of AI usage within organizations – AI is currently being used internally by 57 percent of the interviewed organizations (69 percent in the US). Key internal departments using AI are IT (60 percent), Marketing (37 percent), and Customer Service (30 percent). Yet only 27 percent of total responses stated they were “Extremely Confident” that their present data management and storage could handle AI workloads, with the remaining 73 percent citing their confidence levels on a sliding scale between “Moderately Confident and Not at All Confident.”
AI usage grows amid uncertainty – More than half of respondents reported using AI within their organizations, yet nearly 73 percent are unsure if their infrastructure can handle AI’s current impact, let alone future impacts, as increased AI adoption is expected by 58 percent of respondents in the future.
AI capabilities within data storage management – The audience was questioned on their intent (current and future) to deploy AI directly within their storage infrastructure, and 69 percent were interested in doing so, with AI looking to take an increasingly important role in shaping storage efficiency and performance. Multiple drivers for implementing AI capabilities in data storage were cited. Most mentioned the automation of repetitive storage tasks (43 percent), intelligent storage operations (43 percent), more efficient space management (39 percent), and identifying cost saving potential (38 percent). AI in storage is expected to streamline complex operations, optimize resource use, and enable organizations to tackle performance demands with greater agility and precision.
Download the full 2024 State of Storage Survey report here.
…
SaaS data protector Druva announced new support for both Microsoft Dynamics 365 – enabling enterprises to secure data across the Sales and Customer Service CRM modules – and Microsoft 365 Backup Storage – offering customers expanded protection options for Microsoft 365. With these additions, Druva offers a centralized platform to simplify backup, security, and compliance across infrastructure, end user data, and business-critical applications across the Microsoft ecosystem.
…
Elastic announced its AI ecosystem aimed at helping enterprise developers accelerate building and deploying their RAG applications. The Elastic AI Ecosystem provides developers with a curated, comprehensive set of AI tools integrated with the Elasticsearch vector database. With pre-built Elasticsearch vector database integrations from a network of AI companies, developers are able to:
Deliver more relevant experiences through RAG.
Prepare and ingest data from multiple sources.
Experiment with and evaluate AI models.
Leverage GenAI development frameworks.
Observe and securely deploy AI applications.
…
Data mover Fivetran announced a collaboration with Microsoft to offer scalable, secure data solutions for enterprises through Fivetran’s Managed Data Lake Service. This collaboration empowers organizations to centralize, manage, and scale data in OneLake – part of Microsoft Fabric. George Fraser, CEO of Fivetran, explained: “Our Managed Data Lake Service eliminates manual data engineering and ensures models are built on clean, compliant, and AI-ready data. This collaboration empowers companies to innovate faster while adhering to the highest security and governance standards, including GDPR and HIPAA.”
…
HighPoint’s RocketAIC 6542AWW turnkey storage system (4.84 inches tall, 8.27 inches deep, and 9.25 inches in length) is equipped with 8x Solidigm 61.44TB D5-P5336 NVMe SSDs, with 491.52TB of enterprise storage, at up to 28GB/sec of transfer speed. It’s aimed at edge applications that need fast, high-density storage with a small hardware footprint. The RocketAIC 6542AAW 491.52TB External PCIe 4 x 16 NVMe Drive has an MSRP of USD $78,999.00.
…
Hitachi Vantara’s VSP One Object appliance has a 4RU chassis with six NVMe SSD drives per node. It supports 7.68, 15, and 30TB drives. The base chassis (not expansion, one with management nodes + worker) can have up to 750TB of raw capacity with 30TB drives. An expansion chassis can hold 30 drives, so up to 1.6PB of capacity. 25GbE and 100GbE are supported. It supports all-flash on the nodes plus any media type (HDD, SSD, NVMe, etc.) using a disaggregated architecture.
VSP One Object appliance
Compared to HCP, VSP One Object is built using Kubernetes microservices, which helps improve performance. It also has an improved metadata engine more suited to AI/ML, analytics, and data lakes. It provides stronger protection for metadata, better durability, faster processing, and the ability to scale more easily.
…
Hitachi Vantara’s VSP One Block with QLC has a 2RU controller chassis with 24 drive slots supporting up to 720TB with 30TB NVMe drives. It has a PCIe 4 bus and there can be up to 2x 24 drive expansion chassis. For networking it supports up to 32x 64Gbit FC (SCSI & NVMe) or 16x 25Gbit iSCSI or 8x 100 Gbit NVMe/TCP or various combinations.
VSP One Block appliances
…
Pro creative user-focused supplier iodyne, developer of Thunderbolt storage for media professionals, announced the official release of Windows support for iodyne Pro Data. Editors, colorists, VFX pros, animators, and digital imaging technicians can now use the fastest, most reliable, and secure path from camera to edit on both Mac and Windows.
…
Mirantis introduced its fourth-generation Kubernetes distribution, Mirantis Kubernetes Engine (MKE) 4, based on k0s – a scalable CNCF-certified Kubernetes. More than 300,000 nodes of MKE have been deployed in production. MKE 4 uses declarative lifecycle-management that is highly automated – with platform configurations that are continuously monitored and can be corrected by Kubernetes operators to prevent configuration drift. Installation of MKE Virtualization (KubeVirt) enables virtual machine (VM) workloads to run in tandem with those on containers.
…
A survey of companies over 500 employees across all industries conducted by Nexsan found that six out of ten organizations have experienced some form of cyber attack targeting their data storage in the past 12 months. More than 90 percent of organizations indicated they regularly perform offline backups to prevent ransomware attacks, but only 40 percent utilize immutable backups.
…
Veeam backup target supplier Object First reported a 347 percent year-over-year increase in bookings for Q3 2024. Appliance bookings rose by 235 percent in North America and 409 percent in LATAM year-over-year. Object First saw a 191 percent year-over-year increase in transacting partners in Q3 2024, and over 400 percent growth in partners closing multiple deals since the start of 2024.
…
OpenDrives announced Atlas software version 2.9 and the introduction of Atlas Professional, a new mid-tier feature bundle delivering enterprise-class performance and data management under an unlimited capacity licensing model. Atlas Professional sits in between the starter Atlas Essentials bundle and the flagship Atlas Comprehensive bundle, both of which were introduced in July 2024 during the Atlas 2.8 software launch. This new offering rounds out OpenDrives’ efforts to provide customized, high-performance storage that is scalable to meet the needs of media organizations of all sizes.
…
OWC announced the GA launch of the OWC Thunderbolt 5 Hub, which turns a single cable connection into three Thunderbolt 5 ports and one USB-A port. With up to 80Gbit/sec of bi-directional data speed – up to 2x faster than Thunderbolt 4 and USB 4 – and up to 120Gbit/sec for higher display bandwidth needs, users can connect portable SSDs exceeding 6,000MB/sec, up to three 8K displays @60Hz with DSC, today’s USB-C connector compatible devices, and yesterday’s USB-A desktop accessories, while keeping notebooks and tablets powered and charged up for mobile use with 140 watts.
…
Pliops will show its XDP LightningAI offering, which enables sustainable, high-efficiency AI operations when paired with GPU servers, at SC24. The data-focused startup’s newest Extreme Data Processor (XDP), XDP-PRO ASIC, plus a rich AI software stack and distributed XDP LightningAI nodes, address GenAI challenges by utilizing a GPU-initiated key-value I/O interface as a foundation, creating a memory tier for GPUs, below HBM.
…
Portworx by Pure Storage unveiled as-a-service and enterprise enhancements:
As-a-service capabilities for AI/ML workloads – Portworx is introducing as-a-service delivery for databases and curated foundation AI/ML models. This includes support for vector databases like Milvus, PostgreSQL, and Elasticsearch, as well as graph databases like Neo4j, simplifying the deployment and management of AI/ML workloads.
Unified data management for VMs and containers on Kubernetes – Portworx is supporting enterprises through an assessment-based approach which enables organizations to streamline operations, reduce expenses, and accelerate application modernization on a single platform.
Enterprise enhancements across portworx platform – Portworx has security features, role-based access control (RBAC), and granular controls to improve resource utilization and ROI. Additionally, data resiliency for SQL Server is enhanced with support for Availability Groups, while automated database deployments are streamlined through a Terraform provider, integrating seamlessly into existing GitOps and infrastructure-as-code workflows.
…
Red Hat OpenShift 4.17 brings enhancements to Red Hat OpenShift Virtualization, improving the management of virtualized workloads. Key features include improved safe memory oversubscription, which increases workload density by allowing more virtual machines to run than the available physical memory, and improved dynamic workload rebalancing, ensuring resource optimization and stability during cluster upgrades or changes to workload demand. Storage live migration is introduced as a technology preview, allowing non-disruptive movement of data between storage devices and storage classes while a VM is running. The release also introduces, in technology preview, a dedicated virtualization admin console experience, providing a focused view for managing the OpenShift cluster which is limited to the features, add-ons, and plugins relevant to virtualization, thereby driving greater administrator efficiency.
Red Hat OpenShift Lightspeed provides a GenAI virtual assistant integrated into Red Hat OpenShift that lets teams ask technical questions in straightforward sentences and receive detailed answers, improving how teams learn and work with OpenShift.
…
RelationalAI, the industry’s first knowledge graph co-processor (software, not hardware) for the data cloud, announced GA of its Snowflake Native App on the Snowflake Marketplace, with production support to its customers. RelationalAI enables users to build and modernize intelligent applications with ten times less code and complexity, using a data-centric architecture based on relational knowledge graphs. As an extension of Snowflake’s AI capabilities, RelationalAI allows customers to combine GenAI models with compound AI techniques for rule-based reasoning, graph analytics, prescriptive analytics, and predictive analytics in Snowflake.
…
Rubrik announced a collaboration with Red Hat to support Red Hat OpenShift Virtualization (RHOSV) on Rubrik Security Cloud. OpenShift Virtualization is a feature of Red Hat OpenShift that allows users to run and manage VMs alongside container workloads in a unified platform. Rubrik Security Cloud, using RHOSV, is designed to help organizations more easily migrate and protect virtual machines (VMs) and applications. General availability is expected in early 2025.
…
Cloud data warehouser Snowflake announced Snowflake Intelligence (in private preview soon) – a platform that will enable enterprises to ask business questions across their enterprise data and, in just a few steps, create data agents that take action on those insights. Snowflake Intelligence will give businesses enterprise-grade data agents built on Snowflake’s data foundation that efficiently get organizational work done, while protecting customer IP and delivering answers backed by reliable, trusted enterprise data. It enables everyone to easily access their data, and seamlessly connect with third-party tools – including sales transactions in a database, documents in knowledge bases such as SharePoint, productivity tools such as Slack, Salesforce, and Google Workspace, alongside business intelligence data in Snowflake – so they can talk to their data using natural language. Snowflake Intelligence also supports API calling to enable actions and data modifications to advance business users’ work.
…
On Solidigm’s 122.8TB D5-P5336 SSD, Wedbush analyst Matt Bryson writes that the questions in his view are: (1) Can Solidigm take its work around QLC, and translate the same advantage to hynix (or another vendor’s) charge trap solutions elongating its solution roadmap beyond current output from Dalian, given hynix/Solidigm have ended progression of Intel’s floating gate technology; and (2) Will competitors improve their execution on high capacity eSSDs or will Solidigm’s advantage only continue to elongate giving hynix’s subsidiary more time to establish a future direction (currently we believe the latter outcome appears more likely)?
…
Data protector Veeam announced updates to Veeam Data Cloud Vault – a fully managed, secure, and Azure cloud-based storage service for storing backups of data and applications offsite. The release, developed in collaboration with Microsoft, includes benefits such as:
Security and durability – Veeam Data Cloud Vault storage-as-a-service (STaaS) now offers up to 12 nines of durability that protect against entire datacenter failure.
Predictability – Veeam Data Cloud Vault offers predictable, flat, per-TB pricing on two editions of cloud storage, inclusive of read/write requests, and egress fees.
Simplicity – Veeam Data Cloud Vault minimizes cloud and security skills gaps with on-demand, pre-configured and fully managed cloud storage built on Azure, directly integrated with the Veeam Data Platform interface. This includes the ability to procure, provision, and monitor Veeam Vaults directly from Veeam Data Platform and directly restore to on-premises and Azure VMs.
…
XConn Technologies has joined the Ultra Accelerator Link (UALink) Consortium as a Contributor member. The newly incorporated UALink Consortium aims to establish an industry-standard, high-speed interconnect to enable scale-up communications between AI accelerators and switches, delivering low latency and high bandwidth for AI workloads in datacenter environments. XConn’s participation in the UALink Consortium follows its introduction of the “Apollo” switch – a pioneering 256-lane CXL and PCIe switch fabric architecture.
…
Yellowbrick Data announced that its cloud-native SQL Data Platform is now optimized for on-premises Dell Infrastructure in addition to AWS, Azure, and GCP, and was shown at the Dell Technologies Forum in Arlington, Texas, at Global Life Field on November 14.
…
Zettar, MiTAC Computing, and Nvidia claim they have revolutionized high-speed data movement, both bulk and streaming, with Nvidia BlueField-3 DPUs embedded with Zettar zx. Zetta says the DPUs can do both bulk and streaming data transfers at high speed and nearly the same data rates for 1 MiB to 1 TiB, locally (in the same host and/or cluster) and over any distance – latency is irrelevant. It doesn’t need any site-specific tuning and it also addresses a common challenge – casual exchange (i.e. up to a few TB) of data with collaborators – efficiently and easily.
COMMISSIONED: As the AI era unfolds, I have been reflecting on my journey in the tech industry.
What’s happening with AI at the edge right now reminds me of the early days of the internet, when the potential seemed limitless, but the path was unclear. Similarly, edge computing and AI are at a pivotal moment — with vast potential waiting to be unlocked on the other side.
While GenAI is capturing all the headlines, right now it’s AI inferencing that is driving the majority of the growth at the edge. In this light, it’s hardly surprising that IDC is predicting a surge of hardware spending at the edge that will be double the pace of data center and cloud investments, with a quarter of new infrastructure deployed to edge locations. Other reports concur, predicting that 62 percent of data compute will reside in edge environments within the next few years.
Right now, AI at the edge is revolutionizing business operations across all industries. While compute-intensive AI training typically takes place on large, powerful infrastructure stacks in a cloud, colocation or data center environment, AI inferencing needs to happen close to where data is generated and decisions get made — at the edge. AI inferencing is the stage where the trained model applies what it has learned to real-world scenarios to make predictions or decisions. Generative AI (GenAI) relies on AI inferencing to generate new content, making it a crucial component for GenAI to function and produce meaningful outputs.
Running AI inferencing at the edge provides the low latency required for real-time decisions. For example, if you’re riding in an autonomous vehicle, your life may depend on a split-second decision. You can’t wait for data to go to the cloud and back. AI inferencing at the edge also optimizes costs by avoiding the need to transfer gigabytes of data over the web to the cloud or a core data center and back. And it can enhance operational efficiency for geographically distributed locations with different needs than the home office.
Many organizations with extensive edge environments — such as manufacturers, retailers, healthcare providers and utilities — are in the beginning phases of thinking about and planning for AI. It’s important to identify the challenges that you’d like to solve at the edge and then start to work toward an efficient way to deploy, secure and manage your edge AI over time.
You will also need to consider the unique demands of edge computing. Unlike the controlled environments of data centers, colocation facilities or clouds, the edge presents a diverse set of challenges that need to be addressed for edge AI deployments to be successful. For example, edge IT tends to have a distributed footprint, different environmental conditions, different data sources, much less IT support and a wide attack surface that can be difficult to protect.
This is where Dell NativeEdge comes into the picture. It’s an innovative edge operations software platform designed to streamline and secure edge operations. The platform enables you to bring AI to the edge, faster and simpler, so you can rapidly deploy and manage edge solutions without compromising security or efficiency.
In an edge environment with diverse infrastructure, NativeEdge helps you securely onboard and provision devices with zero touch and Zero Trust. You can then orchestrate AI applications and workloads with other infrastructure as well as other virtual and container environments. In addition to NativeEdge, Dell offers an extensive portfolio of technologies and solutions designed with the edge in mind so that you can easily deploy your AI, at scale, at the edge.
Running AI inferencing at the point of data creation enables real-time, intelligent insights at the point of decision. This is a winning combination that is driving an explosion of AI inferencing at the edge.
To learn more, check out Dell.com/NativeEdge and watch the Capitalize on Your Edge with AI webinar where Alison Biers and I walk you through the business cases, uses cases and technical requirements for edge AI along with step-by-step demos of Dell NativeEdge in action. Whether you’re a tech enthusiast or a business leader, this is a conversation you won’t want to miss.
AI cloud platform operator CoreWeave has closed a $650 million secondary share sale to investors in a deal insiders claim now values the startup at $23 billion.
While the investors were led by Jane Street, Magnetar, Fidelity Management, and Macquarie Capital, both Cisco and Pure Storage took part in the sale, along with others.
Combined with its software, CoreWeave provides access to high-performance GPUs in the cloud to allow organizations to process their AI workloads. Its existing backers include GPU provider Nvidia. The young biz competes against the likes of AWS and Microsoft Azure in this market, and is seen as a rising star by market watchers.
In the secondary stock sale, existing investors sold shares to new investors. According to sources close to the transaction, CoreWeave is now valued at around $23 billion – a massive jump from its valuation of $7 billion around a year ago. CoreWeave was valued at $19 billion only this May, following a $1.1 billion Series C funding round led by private equity firm Coatue.
To cash in on the AI data processing craze, CoreWeave is believed to be moving toward an IPO some time next year.
For its part, alongside its undisclosed financial investment, Pure Storage said it was enabling CoreWeave customers to leverage the Pure Storage flash platform within CoreWeave Cloud, as part of a strategic alliance.
Rob Lee
“Integrating the Pure Storage platform into CoreWeave’s specialized cloud service environments enables customers that require massive scale and flexibility in their infrastructure, the ability to tailor their infrastructure and maximize performance on their own terms,” said Rob Lee, chief technology officer at Pure Storage.
The Pure Storage platform is available as an option within CoreWeave’s dedicated environments, which customers access through the CoreWeave Platform.
“Partnering with Pure Storage gives our customers the flexibility to select storage solutions tailored to their specific AI needs. This collaboration strengthens our commitment to providing high speed performance, reliability, and flexibility for our customers who trust our cloud platform to accelerate the development and deployment of AI,” said Brian Venturo, chief strategy officer at CoreWeave.
On the win for Pure Storage, global investment bank William Blair wrote in a report, just published: “CoreWeave represents an exciting new customer opportunity for Pure, especially as media reports indicate that CoreWeave has plans to go public next year, and is experiencing rapid revenue growth on the back of the GPU farms that it has built.
“We believe CoreWeave is the eight-figure Evergreen//One deal that Pure announced in the fourth quarter, which management called out as being with ‘one of the largest specialized GPU cloud providers for AI.'”
The bank added: “Importantly, we do not believe this is the much-anticipated top ten hyperscaler design win that Pure’s management has been telegraphing all year, partly because CoreWeave is not a top ten hyperscaler.”
Pure Storage CEO Charlie Giancarlo has previously said a top ten hyperscaler win is expected by the end of this year.
Both Backblaze and VAST Data have previously been announced as storage providers to CoreWeave.
Data orchestrator Hammerspace has discovered how to add a GPU server’s local NVMe flash drives as a front end to external GPUDirect-accessed datasets, providing microsecond-level storage read and checkpoint write access to accelerate AI training workloads.
As an example, a Supermicro SYS-521GE-TNRT GPU server has up to 16 NVMe drive bays, which could be filled with 16 x 30 TB SSDs totaling 480 TB or even 16 x 61.44 TB drives amounting to 983 TB of capacity. Hammerspace says access to these is faster than to networked external storage, even if it is accessed over RDMA with GPUDirect. By incorporating these drives into its Global Data Environment as a Tier 0 in front of Tier 1 external storage, they can be used to send data to GPUs faster than from the external storage and also to write checkpoint data in less time than it takes to send that data to external storage.
We understand that checkpoints can run hourly and take five to ten minutes, during which time the GPUs are idling. Hammerspace Tier 0 drops the time from, say, 200 seconds to a couple of seconds.
David Flynn
David Flynn, founder and CEO of Hammerspace, stated: “Tier 0 represents a monumental leap in GPU computing, empowering organizations to harness the full potential of their existing infrastructure. By unlocking stranded NVMe storage, we are not just enhancing performance – we’re redefining the possibilities of data orchestration in high-performance computing.”
Hammerspace points out that, although NVIDA-supplied GPU servers typically include local NVMe storage, this capacity is largely unused for GPU workloads because it is siloed and doesn’t have built-in reliability and availability features. With Tier 0, Hammerspace claims it unlocks this “extremely high-performance local NVMe capacity in GPU servers.”
It is providing this Tier 0 functionality in v5.1 of its Global Data Platform software. It points out that using GPU server’s local storage in this way reduces the capacity needed in external storage, thereby reducing cost, external rack space take-up, cooling, and electricity draw. This can save “millions in storage costs.” A figure of $40 million savings was cited for a 1,000 GPU server installation.
The company has also developed a software addition to speed local storage access and contributed it to the latest Linux kernel 6.12 release. It says this Local-IO patch to standard Linux enables I/O to bypass the NFS server and network stack within the kernel, reducing latency for I/O that is local to the server.
This development “allows use of the full performance of the direct-attached NVMe which multiple devices in aggregate can scale to 100+GB/s of bandwidth and tens of millions of IOPS while maintaining mere microseconds of latency, making Tier 0 the fastest, most efficient storage solution on the market to transform GPU computing infrastructure.”
Hammerspace told us: “LocalIO is way more powerful than GPUDirect. It allows the Linux OS to auto recognize it’s connecting to itself and handle the IO request zero copy – by pointer to the memory buffer.”
Altogether, Hammerspace claims this makes it possible to “unlock the full potential of local NVMe storage by making it part of a global file system that spans all storage from any vendor. Files and objects stored on local GPU server storage can now be shared with other clients and orchestrated within the Hammerspace Global Data Platform. Data can be automatically and intelligently tiered between Tier 0, Tier 1, Tier 2, and even archival storage, while remaining visible and accessible to users.”
Hammerspace says it can work its local GPU server storage and Local-IO magic in the cloud as well as on-premises. Also, this is not just for GPU computing. It can run it in an x86 virtual machine (VM) farm to feed VMs with data faster.
v5.1 of Hammerspace’s software also includes:
A more modern and dynamic user interface.
A highly performant and scalable S3 object interface allowing users to consolidate, access, and orchestrate file and object data on a single data platform.
Performance improvements for metadata, data mobility, data-in-place assimilation, and cloud-bursting.
New Hammerspace data policies (called Objectives) and refinements to existing Objectives make it easier to automate data movement and data lifecycle management.
We understand that a second hyperscaler customer after Meta has adopted Hammerspace and prospects in the US government lab market are looking good.
Quantum reported subdued results for its second FY 2025 quarter but said operational improvements, a product portfolio refresh, and go-to-market enhancements would return the company to growth.
Revenues in the quarter ended September 30 were $70.5 million vs $75.7 million a year ago and down 6.9 percent, with a GAAP loss of $13.5 million compared to a $3.3 million loss a year ago. The revenue fall was largely due to lower primary storage issues, meaning all-flash systems predominantly, and non-recurring project spend. That includes restructuring, getting back on SEC file, and new product introductions.
Jamie Lerner
Quantum chairman and CEO Jamie Lerner said: “Sales bookings and customer win rates for the quarter were consistent with our overall business expectations as we continued to transform the company. However, operational headwinds with the supply chain continued this quarter, resulted in exiting the quarter with higher than anticipated backlog.”
This was approximately $14 million, above the typical $8 million to $10 million run rate.
Lerner said in the earnings call: “We’ve been rotating our portfolio more to high-speed all-flash offers. And as you’ve been watching, high-speed all-flash systems, particularly those from Supermicro, have long lead times. So we’ve been finding that the … SSDs and high-speed servers that use SSDs just have longer lead times. And so we used to have about two to three week lead time on that type of server. Now it can be up to ten weeks.”
Lerner said the Quantum ship was turning around: “Evidence of our transformation can be seen in the progress of gross margin improving 490 basis points sequentially to above 41 percent, as well as non-GAAP operating expenses being reduced by more than 8 percent year-over-year. These actions contributed to our achievement of breakeven adjusted EBITDA for the quarter.”
Financial summary for Q2 FY 2024:
Gross margin 41.5 percent, up 490 basis points
ARR: $146 million vs $145 million last quarter
Subscription ARR: $19.6 million, up 28 percent year-over-year and 5 percent sequentially
Cash, cash equivalents, and restricted cash at quarter end: $17 million vs $25.8 million at Sep 30, 2023.
Cash, cash equivalents, and restricted cash were $25.9 million last quarter vs $26.2 million the year before that. Interest expense increased to $6.1 million from $3.9 million a year ago, and total debt rose to $133 million from the year-ago $109.4 million.
The chart shows that there was a rising revenue trend starting in FY 2021’s first quarter, which ended ten quarters later in Q3 FY 2023. Then revenues plunged for three quarters in a row and have now stabilized in the $70 million to $71 million area for three quarters. The missing Q3 FY 2024 revenue number will probably be in the $75 million to $72 million area.
Quantum delayed filing its fiscal 2024 SEC report due to an accounting problem with its standalone pricing methods for sales in the period. Last quarter it provided Q1 FY 2024 results and this quarter the Q2 results have been revealed. We still await Q3 FY 2024 results and expect these in three months time when the firm reports its Q3 FY 2025 results.
The chart above shows that Quantum’s new normal for quarterly revenues is around $71 million and it has been cutting costs to try to regain profitability, saving almost $40 million since FY 2023.
CFO Ken Gianella said in the earnings call: “Over the last several years, the company has had significant cash spend on onetime consulting, a new ERP, updated infrastructure, new product introductions, and restructuring expenses. We are pleased to announce that we are substantially complete with these efforts.”
Lerner said restructuring and operational improvements are “improving our free cash flow, which is expected to be positive in the back half of fiscal year 2025 and driving fiscal 2026 to be cash flow positive for the first time in five years.”
A chart below showing Quantum’s revenues sequentially since FY 2016 shows a five-year downward trend, which Lerner then reversed for nine quarters, until that turnaround was itself reversed when the public cloud hyperscalers stopped buying Quantum’s tape libraries in FY 2023.
However, the tape library business is showing signs of growth, with Lerner saying there was “a multimillion-dollar purchase order in-house from one of the world’s leading cloud platforms” for Quantum’s Scalar i7 RAPTOR library. Gianella added: “We’re super excited about the i7 coming out and being a category killer to get those win rates back up.”
The company thinks growth prospects are picking up, with Lerner saying: “Our business strategy remains focused on high-priority growth initiatives, particularly around Myriad and ActiveScale as we are seeing demonstrated proof points of our ability to significantly expand within our target verticals. In Q2 2025, we achieved significant pipeline growth for Myriad and ActiveScale.”
The DXi T-series 1RU all-flash target backup appliances did well in the quarter, with Lerner saying: “We’ve had multiple strategic wins against the competition based on the DXi T-Series fast recovery times in the face of a cyberattack due to its leading data reduction and recovery rates.
“While our efforts are still short of the intended results, we are seeing positive proof points through our new product introductions, including Myriad traction, combined with driving a more operationally efficient business.”
Outlook
Gianella said: “While we are exceeding our expectations on product mix, gross margin, and cost improvements, we need to continue to focus on improving our overall revenue execution. We see improvements in second-half of FY 2025 continuing into FY 2026.”
Next quarter’s revenue outlook is $72.0 million +/- $2.0 million. This will be a sequential rise at the midpoint but, as we don’t yet know the year-ago Q3 revenue number, we don’t know if this will be a revenue rise or fall compared to a year ago. Gianella said that the outlook “reflects management’s view of ongoing operational headwinds including transition to a new manufacturing partner during the quarter … We’re consolidating our manufacturing operations into one new location.”
Lerner said Quantum is “evolving our sales model to focus dedicated sales resources on select product lines” as a way of growing revenue. He added: “We have completed the heavy lift on the operational model. We have fully refreshed our product portfolio and we are now actively engaged in re-energizing our go-to-market approach. All of these combined create positive momentum in the coming quarters and beyond.”
Its full-year outlook is $280 million +/- $5 million, a 10 percent fall on fiscal 2024’s $311.6 million. This implies Quantum’s fourth FY 2025 quarter will bring in around $66 million, a 7.7 percent fall annually.
Delisting
Having avoided a previous Nasdaq delisting threat, because its shares traded below a minimum $1 dollar value, by implementing a reverse stock split, Nasdaq told Quantum on October 4 that it again faced delisting. This time it was because its minimum market value, based on its publicly traded shares, was below the required $15 million for 30 consecutive days. It has 180 days to regain compliance by its traded share market value being above $15 million for ten consecutive days.
Storage array supplier Infinidat has devised a RAG workflow deployment architecture so its customers can run generative AI inferencing workloads on its InfiniBox on-premises and InfuzeOS public cloud environments.
The HDD-based InfiniBox and all-flash InfiniBox SSA are high-end, on-premises, enterprise storage arrays with in-memory caching. They use the InfuzeOS control software. InfuzeOS Cloud Edition runs in the AWS and Azure clouds to provide an InfiniBox environment there. Generative AI (GenAI) uses large language models (LLMs) trained on general-purpose datasets, typically in massive GPU cluster farms, with the trained LLMs and smaller models (SLMs) used to infer responses (inferencing) to user requests without needing massive GPU clusters for processing.
However, their generalized training is not good enough to produce accurate responses for specific data environments, such as a business’s production, sales, or marketing situation, without their response retrievals using augmented generation (RAG) from an organization’s proprietary data. This data needs transforming into computer-generated vectors, dense mathematical representations of pieces of data, that need storing in vector databases and made available to LLMs and SLMs using RAG inside a customer’s own environment.
Infinidat has added a RAG workflow capability to its offering, both on-premises and in the cloud, so that Infinidat-stored datasets can be used in GenAI inferencing applications.
Infinidat CMO Eric Herzog stated: “Infinidat will play a critical role in RAG deployments, leveraging data on InfiniBox enterprise storage solutions, which are perfectly suited for retrieval-based AI workloads.
Eric Herzog
“Vector databases that are central to obtaining the information to increase the accuracy of GenAI models run extremely well in Infinidat’s storage environment. Our customers can deploy RAG on their existing storage infrastructure, taking advantage of the InfiniBox system’s high performance, industry-leading low latency, and unique Neural Cache technology, enabling delivery of rapid and highly accurate responses for GenAI workloads.”
Vector databases are offered by a number of vendors such as Oracle, PostgreSQL, MongoDB, and DataStax Enterprise, whose databases can run on Infinidat’s arrays.
Infinidat’s RAG workflow architecture runs on a Kubernetes cluster. This is the foundation for running the RAG pipeline, enabling high availability, scalability, and resource efficiency. Infinidat says that with AWS Terraform, it significantly simplifies setting up a RAG system to just one command to run the entire automation. Meanwhile, the same core code running between InfiniBox on-premises and InfuzeOS Cloud Edition “makes replication a breeze. Within ten minutes, a fully functioning RAG system is ready to work with your data on InfuzeOS Cloud Edition.”
Infindat says: “When a user poses a question (e.g. ChatGPT), their query is converted into an embedding that lives within the same space as the pre-existing embeddings in the vector database. With similarity search, vector databases quickly identify the nearest vectors to the query to respond. Again, the ultra-low latency of InfiniBox enables rapid responses for GenAI workloads.”
Marc Staimer, president of Dragon Slayer Consulting, commented: “With RAG inferencing being part of almost every enterprise AI project, the opportunity for Infinidat to expand its impact in the enterprise market with its highly targeted RAG reference architecture is significant.”
NetApp has also added RAG facilities to its storage arrays, including integrating Nvidia’s NIM and NeMo Retriever microservices, which is not being done by Infinidat. It and Dell are also supporting AI training workloads.
If AI inferencing becomes a widely used application, all storage system software – for block, file, and object storage – will have to adapt to support such workloads.
