Home Blog Page 9

Storage news ticker – September 22

By
Chris Mellor
-

Acronis will incorporate Seagate’s S3-compatible Lyve Cloud Object Storage into its Acronos Archival Storage offering to help enable MSPs to deliver services with more cost-effective, and compliant storage. Learn more here.

…

China’s Biwin has introduced a prorietary MicroSSD card-like Mini SSD product, 15.0 mm × 17.0 mm × 1.4 mm in size, supporting PCIe gen 4, and delivering pp to 3700 MB/s sequewntrial read, 3400 MB/s write speeds. It uses TLC 3D-MAND, Land Grid Array (LGA) packaging and has 512 GB, 1 TB, and 2 TB capacities. It has a modular slot design, enabling plug-and-play functionality and wide compatibility with diverse devices, including laptops, tablets, smartphones, handheld gaming consoles, cameras, NAS systems, smart albums, and portable SSDs.

…

BlackFog, a global cybersecurity company providing on-device anti-data exfiltration (ADX) technology, announced a distribution partnership with Exertis Enterprise to expand BlackFog’s presence across the UK and European markets and equip resellers and service providers to deliver advanced prevention against unauthorised data exfiltration and evolving AI-enabled threats.

…

Data protector Commvault announced the findings of its latest report with ESG which assesses the real-world impact that its Cleanroom Recovery and Cloud Rewind offerings are having on customers worldwide. It found that customers:

  • Reduced recovery times by 99 percent
  • Rebuilt cloud infrastructure 94 percent faster 
  • Improved testing frequency by 30 times
  • Reduced testing time by 99 percent 

…

Cristie Software, which supplies system recovery, replication, and migration technologies, announced GA of Cristie System Recovery PBMR (PowerProtect Bare Machine Recovery) for Dell Technologies PowerProtect Data Manager. This new integration extends PowerProtect Data Manager environments with full system recovery capabilities, enabling bare machine recovery (BMR) to physical, virtual, or cloud-based targets. PBMR also features powerful automated recovery tools, including Continuous Recovery Assurance, which automatically performs a full system recovery within a clean room environment whenever a new backup is detected on a protected system. PBMR enhances Cristie’s system recovery portfolio, joining existing solutions for Dell Technologies Avamar and NetWorker platforms.

…

Databricks is partnering with London based AI startup Applied Computing to tackle emissions from heavy industry in India. Applied Computing is one of only three British AI startups to develop their own foundational model (others being Synthesia and Wayve). It’s developed Orbital, the first foundation model designed specifically for complex energy environments, including refining, upstream operations, LNG, and renewables. Orbital is built on physics-inspired, domain-specific AI that delivers accuracy, explainability and real-time optimisation at scale. It can cut energy operations emissions by 10 percent. By embedding Orbital into the Databricks Data Intelligence Platform, customers in energy and utilities can access industry-specific AI models directly within it.

…

Dell has updated its PowerProtect (deduping backup target) product set:

  •  PowerProtect Data Domain All-Flash appliance is now available, enhancing cyber resiliency with up to four times faster data restores and two times faster replication performance.
  •  PowerProtect Data Manager now features new ecosystem, security and virtualization enhancements.
  •  PowerProtect Cyber Recovery now supports Data Domain All-Flash and CyberSense Analytics support for Commvault client-direct backups of Oracle databases.
  •  PowerProtect Backup Services can now leverage Microsoft Azure Storage as a backup target.

…

The EU’s General Court has rejected a challenge to the European Commission’s approval of the EU-US Data Privacy Framework, the system that allows certified US companies to receive personal data from Europe under strict privacy rules. The challenge, brought by a French MEP, argued that US surveillance powers remain too broad and that the new US redress body lacks independence. The Court disagreed, finding that although the US has wide surveillance powers, they are subject to checks and remedies that are sufficient under EU law.

…

Hemanth Vedabargha

Postgres-compatible SQL and real-time data analytics and AI platform supplier Firebolt has promoted its president, industry veteran Hemanth Vedabargha, to CEO. He succeeds co-founder Eldad Farkash, who will now serve as Firebolt’s chairman of the board. Firebolt’s decoupled metadata, storage, and compute architecture gives organizations the flexibility to run in the cloud or self-host, while maintaining the same efficiency, Iceberg support, and Postgres compatibility.

…

UK-based Firevault has launched its three Firevault products:

  • Vault: A secured offline digital vault for directors, shareholders, HNWIs, family offices, and professionals. Protects board papers, contracts, financial records, identity documents, investments, IP, and personal archives. Identity-locked to one verified owner, available in 2 TB, 4 TB, and 8 TB models from £360/month, with Backup Vault and Vault Buddy for continuity.
  • Storage: A scalable secured offline data storage system for enterprises and industries. Keeps crown-jewel, commercial, critical, and customer data physically disconnected and unreachable. Acts as the “0” in the 3-2-1-0 model, delivering risk reduction and 30-plus percent OPEX savings versus connected systems.
  • Platform (FV-PaaS): An offline-first platform with nine offline modules, designed to help organisations take control of their data and protect critical assets. Supports secure, multi-site, sovereign-grade deployments for governments, enterprises, and critical infrastructure, ensuring sensitive information remains disconnected and beyond attacker reach. 

…

Gartner produced a new Enterprise Storage Platform Magic Quadrant with Pure Storage topping the charts, followed by NetApp, HPE, Huawei, IBM, and Dell in the top right Leaders’ box. Hitachi Vantara is the sole Visionary with DDN and IEIT Systems Niche Players. There are no Challengers. How long will it be before VAST Data and Infinidat appear in this chart? They are definitely enterprise storage platform suppliers but can’t pass Gartner’s stiff exclusion criteria barriers – which will be a source of satisfaction, if not relief, to the suppliers that do.

HPE said: “This marks the 16th consecutive year of recognition as a Leader for HPE’s Ability to Execute and Completeness of Vision.” Get the full report from HPE here, from NetApp here, and Pure Storage here.

Gartner says this is a new MQ, so no vendors have been added or dropped. It says: “Vendors must have a minimum total product and services revenue of $325 million (excluding support and maintenance), with at least 25 percent of total revenue coming from unstructured data product revenue. In addition, unstructured data product revenue must have a minimum year-over-year growth of 15 percent, as reported through 31 May 2025.” This and other restrictions exclude VAST Data and Infinidat.

A look at Gartner’s 2024 Primary Storage Platform MQ shows notable differences. As this 2025 Enterprise Storage Platform MQ is a new report, we shouldn’t make any comparisons. HPE’s comment certainly implies it’s comparing its status with previous such Gartner MQs though. As, obliquely, does a statement from Pure: “For the previous five years, Pure Storage was positioned highest and furthest in the Gartner Magic Quadrant for Primary Storage Platforms.” Here, therefore, we show the 2024 Primary Storage Platform MQ so that, like HPE and Pure, you won’t make any comparisons 🙂

A March 2025 Voice of the Customer research report from Gartner found that primary storage customers prefer Huawei, Infinidat, and NetApp over Dell, HPE, Hitachi Vantara, Pure Storage, and other suppliers.

…

Hardware RAID supplier Graid Technology, Giga Computing, and Miruware signed a Memorandum of Understanding focused on South Korea. The trio will ensure seamless integration of Graid’s GPU-accelerated storage solutions into Giga Computing’s server platforms, pursue technology integration initiatives, co-developing optimized solutions tailored to the needs of South Korean enterprises. They will engage in joint marketing and outreach, amplifying awareness of next-generation server and storage innovations, and collaborate on customer development, including lead generation, solution demos, and after-sales services.

…

A Forrester Total Economic Impact (TEI) study found that customers using Hitachi Vantara’s Virtual Storage Platform One (VSP One) storage product have achieved a 285 percent return on investment (ROI), $1.1 million in net present value (NPV) and payback in just seven months. The TEI study evaluated the VSP One experiences of six decision-makers across North America, Europe and Asia Pacific. Download the study here.

…

Earth Sciences New Zealand (formally NIWA) of Aotearoa selected HPE’s Cray XD2000, purpose-built for AI and simulation workloads, to accelerate the organization’s environmental science and precision of meteorological forecasting. The new system, named Cascade, replaces an aging Earth Sciences New Zealand’s HPC system, delivering 3x more computing power to run multiple AI-powered simulations simultaneously for more accurate weather predictions. It supports Earth Sciences New Zealand’s mission to strengthen the country’s resilience against weather and climate-related hazards, such as wildfires and flooding from heavy rainfall. Cascade is powered by AMD 4th Gen EPYC processors and tightly coupled with HPE GreenLake for File Storage to deliver a highly performant, simplified storage environment and enable 19 petabytes of data to accelerate local research, supporting faster decision-making and more informed crisis management.

…

Global AI and data consultancy Indicium has new funding from Databricks Ventures, the strategic investment arm of Databricks. Indicium and Databricks have been partners since 2017, working to modernize data platforms and deliver AI systems to some of the world’s biggest companies, including Burger King, PepsiCo and Volvo. Going forward, Indicium will be more tightly integrated with Databricks’ product roadmap, including Agent Bricks and Lakebase, engineering resources, go-to-market efforts and training. Indicium will work more closely with Databricks on features such as advanced AI capabilities, Unity Catalog enhancements and Lakehouse upgrades. And Indicium and Databricks will work to co-develop AI and data solutions.

…

Jan Ursi.

Cloud-based SaaS data protector Keepit appointed Jan Ursi as VP of Global Channel. He has led partner and alliance initiatives at Rubrik, UiPath, Nutanix, and Infoblox, and earlier helped expand NetScreen and Juniper in Europe.

…

Japan’s Nikkei reports Japan will provide Micron Technology a ¥536 billion ($3.63 billion) subsidy for R&D and capital spending at its fab in Hiroshima, Nikkei Asia reports. Micron plans to invest ¥1.5 trillion to expand the fab to a maximum capacity of 40,000 wafers per month and make advanced memory chips there by end of fiscal 2029, so ¥500 billion of the subsidy will cover a third of that cost, while the other ¥36 billion is for R&D over five years.

….

During the week of Sep 18, RDIMM demand has increased unexpectedly. Wedbush analysts say eSSD demand has lifted significantly, in line with greater RDIMM demands, driving NAND ASPs higher. A large portion of this increase appears tied to greater CSP requirements for compute SSDs. However, a shortage of HDDs (and the need to find other storage) also appears to be creating incremental demand for QLC. It thinks Micron’s revenues will be karger as a result of this unexpected shift.

…

MSP-focused data protector N-able has launched Anomaly Detection as a Service (ADaaS), to strengthen Cove Data Protection’s defense against cyberthreats, with no additional management or cost impact. The first available ADaaS capability, Honeypots,is an always-on defense mechanism designed to detect brute-force attacks on infrastructure. Honeypots will be able to identify access attempts by unauthorized parties, ensuring that infrastructure and data remain fully protected. N-able plans to expand Cove’s Anomaly Detection functionality with capabilities including critical configuration changes, which will alert users to suspicious backup policy modifications.

…

A NAKIVO Backup & Replication release will enable customers (enterprises, SMBs and MSPs) to store backups anywhere: local, cloud, deduplication appliances, file shares and even tape with immutability, encryption, air gapping and other cybersecurity features supported. The new features/enhancements are designed to deliver disaster resilience, near-zero downtime and data loss across diverse IT environments – while also simplifying accessibility and MSP-tenant management. Should there be any incident or disruption to an organisation’s digital assets, data can be seamlessly restored.

…

The PCI-SIG announced the PCI Express 8.0 specification is well underway, with revision 0.3 now available to members. Check out a blog post below to learn more. PCIe 8.0 spec feature objectives:

  • Delivering 256.0 GT/s raw bit rate and up to 1.0 TB/s bi-directionally via x16 configuration
  • Reviewing new connector technology
  • Confirming latency and FEC targets will be achieved
  • Ensuring reliability targets are met
  • Maintaining backwards compatibility with previous generations of PCIe technology
  • Developing protocol enhancements to improve bandwidth
  • Continuing to emphasize techniques to reduce power

…

Pinecone has just become the first and only vector database integrated into Mistral AI’s Le Chat platform, as part of that company’s major enterprise expansion. The integration is part of Mistral’s broader push to become the enterprise AI assistant of choice, featuring 20+ connectors across data, productivity, and development tools from Notion, Asana, PayPal and others. This makes Pinecone the sole vector database available to Le Chat’s growing user base across free and enterprise tiers, enabling users to search and analyze vector data directly within their AI workflows.

…

Aquant, an Agentic AI platform for service organizations that maintain complex equipment, chose Pinecone’s vector database to replace its previous previous vector search infrastructure, built on PostgreSQL extensions, which was too slow. The results:

  • 98 percent retrieval accuracy
  • Cut full response time from ~24s to ~13.7s
  • Initiated responses 2x faster at 2.89s
  • Reduced no-response queries by 53 percent

…

UK-based mainframe to-cloud connectivity supplier Precisely is expanding the Precisely Data Integrity Suite with a new generation of AI agents and a context-aware Copilot – designed to bring autonomous, yet fully governed, data integrity to enterprises worldwide. Built to operate across complex hybrid and multi-cloud environments, these innovations will empower organisations to keep data accurate, consistent, and contextual by automating and optimizing data management processes without constant human intervention. 

…

Progress Software launched Progress Agentic RAG, a SaaS Retrieval-Augmented Generation (RAG) platform designed to make trustworthy and verifiable generative AI accessible across organizations and teams of all sizes. The new platform expands Progress’ portfolio of end-to-end data management, retrieval and contextualization solutions that empower businesses to leverage all their data to gain a competitive edge. Progress Agentic RAG platform is available now as a self-service offering on AWS Marketplace, as well as at Progress.com. Pricing starts at $700/month, providing small businesses, departmental teams and individuals the immediate power to transform unstructured data into actionable intelligence at an affordable price. For more information or to start a free trial, click here.

…

Tom Leyden

Object storage supplier Scality has appointed Tom Leyden as VP of Product Marketing. Leyden brings over two decades of experience in SaaS, storage, backup, and AI. His background includes leadership roles at Amplidata (acquired by HGST), DDN, Excelero (acquired by Nvidia), and Kasten by Veeam. 

…

Here are some recent Scality blogs:

…

Research house TrendForce says its latest findings indicate that over the next two years, AI infrastructure will mainly focus on high-performance inference services. As traditional high-capacity HDDs face significant shortages, CSPs are increasingly sourcing from NAND Flash suppliers, boosting demand for nearline SSDs designed specifically for inference AI and catering to urgent market requirements. NAND Flash vendors are quickly validating and adopting nearline QLC NAND Flash products to address the supply gap. QLC technology offers higher storage density at lower costs, making it essential for meeting large-capacity demand. Suppliers are also increasing QLC SSD production, with capacity utilization expected to steadily grow through 2026. The demand is likely to persist into 2027 as inference AI workloads expand, leading to tight supply conditions for enterprise SSDs by 2026.

To enhance the competitiveness of nearline SSDs in AI storage and better replace HDDs, future products will focus on increasing capacities and lowering prices. Manufacturers are working on new nearline SSDs that surpass mainstream HDDs in capacity, offer better cost efficiency, and significantly cut power consumption.

Beyond inference AI, NAND Flash suppliers are also focusing on AI training applications by introducing High Bandwidth Flash (HBF) products, dividing the industry into two main technology groups. The first is led by SanDisk, which is developing a hybrid design that integrates HBM with HBF. This approach seeks to balance large capacity and high performance to fulfill the dual needs of data throughput and storage in AI model training. The other group is led by Samsung and Kioxia, focusing on storage-class memory technologies like XL-Flash and Z-NAND. These offer a more affordable alternative to HBM, aiming to attract a wider range of customers.

…

VAST Data has been ranked 24th in the 2025 Forbes Cloud 100, the definitive ranking of the top 100 private cloud companies in the world, published by Forbes in partnership with Bessemer Venture Partners. VAST recently experienced 5x year-on-year sales growth in FY25, on top of a 2023 valuation of $9.1 billion. It has achieved a verified Net Promoter Score (NPS) of 84 – the highest in the industry. Earlier this year, Forbes also recognized VAST on the 2025 Forbes AI 50 List, which honors the top privately-held companies developing the most promising business use cases of artificial intelligence. Databricks is at no. 4 on the Forbes Cloud 100. Cohesity is at no. 21 with a $9 billion valuation. Cyera is no. 82.

…

Parallel file system supplier VDURA has launched its first scalable AMD Instinct GPU reference architecture in collaboration with AMD. The validated blueprint defines how compute, storage and networking should be configured for efficient, repeatable large-scale GPU implementations. The design combines the VDURA V5000 storage platform with AMD Instinct MI300 Series Accelerators to eliminate performance bottlenecks and simplify deployment for the most demanding AI and high-performance computing (HPC) environments. It supports 256 AMD Instinct GPUs per scalable unit, achieves throughput of up to 1.4 TB/s and 45 million IOPS in an all-flash layout, and delivers around 5 PB of usable capacity in a 3 Director and 6 V5000 node configuration. Data durability is assured through multi-level erasure coding, while networking options include dual-plane 400 GbE and optional NDR/NDR200 InfiniBand. The full 20-page architecture reference guide is available for download here.

…

ZDNet Korea reported that China’s YMTC is interested in entering the HBM market, possibly via a collaboration with CXMT.

…

IBM, Supermicro, and software RAID supplier Xinnor have produced an “IBM Storage Scale with Supermicro and Xinnor XiRAID” reference architecture. Check it out here. 

IBM, Supermicro, and Xinnor architecture diagram

…

Xinnor has produced xiRAID Opus 1.2, an NVMe Composer in Lunux user space, which is GA. It features more NVMe over Fabrics capabilities and Virtualized access protocol (VirtIO-BLK) performance and scalability. It ‘unifies local and network-attached NVMe drives into a single, high-performance storage platform that maximizes speed and reliability while minimizing hardware overhead and power consumption. The solution’s breakthrough architecture delivers an impressive 1 million IOPS and 100 GB/s throughput per CPU core with ultra-low latency, providing linear scaling across cores and nodes. xiRAID Opus 1.2 native NVMe-oF initiator and target as well as VirtIO-BLK implementation allow it to seamlessly integrate network storage capabilities across drives, drive-namespaces, and RAID arrays and volumes. This integrated approach enables organizations to simplify NVMe deployments by removing the complexity of integrating various Linux components.”

Neo4j backs new graph query standard for AI era

By
Chris Mellor
-

Interview. AI searches that query any part of an organization’s data will need to look at structured as well as unstructured data, and the structured data won’t just be relational databases, where SQL rules. There are graph databases that store relationships between entities. The entries in a graph database can’t be vectorized, ruling out GenAI similarity search responses to natural language queries into such database content, nor can SQL be used.

Neo4j reckons it has a way for GenAI to access its graph database records, and we interviewed Andreas Kollegger, lead for GenAI Innovation, to find out more.

Blocks & Files: Where should graph databases be used instead of relational or other databases? What do graph databases do that other databases don’t?

Andreas Kollegger

Andreas Kollegger: Graph databases take a different approach to data modelling and querying compared to relational databases, focusing instead on use cases where the relationships between data points are just as important as the data itself. They don’t just store data, they encode the semantics of how things relate. That’s why graphs excel at fraud detection, recommendation engines, supply chain bottleneck analysis, and other scenarios where uncovering hidden patterns can simplify complexity and surface insights critical to decision-making.

On the other hand, relational databases are designed to store structured data in tables and perform aggregations, sums, or filters, which are more likely to be used in accounting systems, inventory management, customer records and other transactional applications. In relational databases, uncovering complex insights often requires complicated JOINs and multiple queries, which become cumbersome and inefficient as the connections multiply. Graphs, however, allow you to traverse relationships directly, revealing insights that would otherwise remain buried. 

Put simply, (despite what the name implies) relational databases tell you what exists; while graph databases tell you how those things are connected, leading users to creative solutions for pressing problems.

Blocks & Files: The widely used relational databases have their SQL query language, which is pretty standard and cross-supplier. Graph databases are nowhere near as widely used. Is there a cross-supplier query language? Is one likely to evolve? 

Andreas Kollegger: It’s fair to say that SQL is the standard language of relational databases, which is why relational systems have such broad adoption. For many years, graph databases didn’t have a single, universal query language. However, as of April 2024, that changed with the announcement of the ISO approved Graph Query Language (GQL) – a concrete standard for querying graphs across platforms with broad industry backing. It’s closely aligned with Cypher and familiar to SQL users, which makes adoption straightforward. With all major graph vendors moving toward GQL compliance, this marks a significant step towards wider adoption of graph technologies across enterprises.

Blocks & Files: What is Neo4j’s graph database query language? Could you provide a simple example of what it looks like and how it works? 

Andreas Kollegger: Neo4j uses Cypher as its query language, which has organically evolved to become a fully GQL-compliant implementation. That way, developers can keep using Cypher as they always have, while knowing it aligns with the new ISO standard for graph querying. Most notably, Cypher is designed to be readable and intuitive. By matching patterns of nodes with relationships, users can navigate connected data easily, without needing to write complicated JOINs. It also scales naturally to more complex queries, whether analyzing social networks, supply chains, or recommendation engines. 

Cypher’s similarity with SQL makes it easier for developers with SQL experience to interpret Cypher queries, therefore supporting a smoother transition from relational to graph querying.

For example, one can create a query to find all movies connected to Tom Hanks and the type of relationship, like this: 

MATCH (tom:Person {name:’Tom Hanks’})-[r]->(m:Movie) 

RETURN type(r) AS type, m.title AS movie 

This query finds the Person node for Tom Hanks, follows all outgoing relationships [r] to Movie nodes, and returns both the relationship type (e.g., ACTED_IN or DIRECTED) and the movie titles. This illustrates how Cypher makes exploring and querying connected data simple and intuitive. 

Blocks & Files: How skilled do ordinary users have to be to use it? As skilled as an SQL coder? Do they need access to a graph database query building person? 

Andreas Kollegger: Historically, while there are similarities between Cypher and SQL, users needed some understanding of graph structures – nodes and relationships – to write queries effectively. This often meant relying on developers to write queries and interpret results, making graph databases feel like a specialist domain. However, today, graph technology is being democratized. Modern tools provide drag-and-drop workflows, ready-made algorithms, and integrations with familiar formats like spreadsheets, meaning you don’t have to be a Cypher expert to start exploring graphs. 

With graph technology now appearing in forms that business users can explore and understand for themselves, you don’t need SQL-level expertise. Developers now serve more as facilitators than gatekeepers, enabling a broader ecosystem of users to leverage graph data effectively. 

Blocks & Files: Could GenAI act as a natural language interface to Neo4j’s graph database, constructing a query from a user’s input request? How would that work? 

Andreas Kollegger: Absolutely, GenAI language models can map natural language requests to Cypher queries by first interpreting the intent of your question and then translating it into the right query structure. For instance, if you were to ask, “Which customers bought both product X and product Y in the last month?” a GenAI system can automatically generate the appropriate MATCH and WHERE clauses in Cypher.

The process typically works in three steps. First, intent extraction, where the system interprets what the user actually wants, followed by query generation, which turns that intent into Cypher. Finally, execution and post-processing run the query in our graph databases and format the results for the user.

This removes the need for the user to know the language, letting them interact with graph data conversationally. For example, with Neo4j’s LLM Knowledge Graph Builder, users can drag in documents, web pages, videos and more to create a queryable graph, then use a natural language interface to ask questions, with the LLM automatically extracting nodes, relationships, and generating queries. This makes it easy for anyone, regardless of technical expertise, to explore complex connected data. 

Blocks & Files: Could you describe the output from such a natural language query? 

Andreas Kollegger: The output could be tabular, visual or both, depending on the tool. Sometimes, results are returned as a simple table of rows and columns, similar to SQL, showing the nodes and their properties. In other cases, it could be a graph visualization that highlights how nodes connect and what relationships exist between them. Neo4j’s tools, like Bloom or Browser, can also overlay aggregated metrics or insights, including node counts or centrality scores, derived from algorithms run on the graph databases.

Since graphs encode relationships natively, the visual representation is often the most intuitive – letting users see the connections directly rather than having to infer them from rows and columns. 

Blocks & Files: I’m guessing it does not make sense to suggest a graph database could be vectorized. Why not? 

Andreas Kollegger: Vectorising a graph by itself may seem appealing, but it can’t capture the structured, navigable nature that makes graph databases powerful. Vectors convert data into numerical forms, making them great for similarity search, machine learning or embedding documents and images. However, graphs encode explicit relationships between nodes, which can be traversed and analysed. Converting a graph entirely into vectors would lose that native structure and semantics. 

That said, Neo4j does integrate native vector search as part of its core database capabilities to help capture implicit patterns and relationships based on items with similar data characteristics rather than exact matches. 

This allows users to perform similarity searches or embed ML features while still preserving the graph structure. This approach combines the strengths of vector-based methods for AI/ML tasks with the rich, traversable relationships that make graph databases inherently powerful.

Blocks & Files: An AI agent could receive a query about an organization’s data, some of which is in relational databases, some in a graph database, and some held in unstructured files and object storage systems. Am I right in thinking that the agent could deconstruct the query into three sub-queries? One would use trad SQL to search the relational database, one would use an LLM sub-agent to look at the vector embeddings of the unstructured data, and a third would take its part of the query and build a graph database query and execute it. Then the overall agent would combine the results from the three types of search and generate a response for the user. Does this make sense? 

Andreas Kollegger: Yes, that is one approach that makes sense when an organisation’s data is spread across multiple types of storage. Modern AI agents can act like orchestrators, breaking a complex query into specialised sub-queries tailored for each data store. For your example, the three sub-queries would be: 

  • Relational databases – an agent can generate SQL queries to filter, join and aggregate structured tables.
  • Graph databases – an agent can translate the relevant part of the query into Cypher to uncover patterns or connections that wouldn’t be obvious from a flat table. 
  • Unstructured data (documents, emails, etc.) – an LLM-powered agent can use vector embeddings to find semantically relevant information, even when the exact words don’t match. 

The agent then aggregates the results, using retrieval mechanisms such as GraphRAG, which leverages the structure of knowledge graphs to pull relevant nodes and relationships before synthesizing a coherent response. This hybrid approach enables users to leverage the strengths of each data paradigm without forcing one system to do all the work, delivering faster, richer insights than any single database could provide. By using GraphRAG, AI agents can generate answers that are not only correct but also more accurate, contextual, and explainable.

Bootnote

GraphRAG combines knowledge graphs with Retrieval-Augmented Generation (RAG), enabling you to build GenAI applications that deliver better results. Here is a brief history of Neo4j:

  • 2000: The concept of Neo4j began when founders Emil Eifrem, Johan Svensson, and Peter Neubauer, while working on a content management system, identified the need for a database to handle complex relationships more effectively than relational databases.
  • 2007: Neo4j, Inc. (originally Neo Technology) was founded in Malmö, Sweden, and the first version of the Neo4j graph database was released as an open-source project.
  • 2010: Neo4j gained traction with the release of version 1.0, introducing features like the Cypher query language, a declarative language for querying graph data, which simplified development.
  • 2011-2014: Neo4j grew in popularity, with enterprise adoption increasing. The company raised funding to expand, and Neo4j transitioned to a dual-licensing model (open-source Community Edition and commercial Enterprise Edition).
  • 2018: Neo4j 3.4 introduced native graph processing improvements and cloud integrations.
  • 2020: Neo4j launched Aura, a fully-managed cloud service, making it easier for organizations to deploy and scale graph databases.
  • Present (2025): Neo4j provides advancements in AI integration, scalability, and cloud-native features. It has more than 80 Fortune 100 customers, 170-plus partners and 300,000 developers in its ecosystem.

Contact Neo4j to find out more. 

Dell reclaims top spot in all-flash array market

By
Chris Mellor
-

Dell is claiming the number one position in the all-flash array (AFA) market, citing IDC numbers. Yet, just last month, NetApp claimed it led the AFA market.

NetApp said in August, when reporting its first quarter fiscal 2026 results (ended July 25, 2025), that it “achieved the #1 market share position in all-flash storage for calendar Q1 2025, as reported by IDC.” CEO George Kurian said: “By helping customers modernize with cutting-edge and cyber resilient storage solutions, we have taken the lead position in the all-flash market.”

Jeff Clarke

The IDC report was the Worldwide Quarterly Enterprise Storage Systems Tracker, June 2025.

Dell vice chairman and COO Jeff Clarke then put out an X post on 19 September, saying: “#1 in all-flash storage vendor revenue. Proof that innovation and execution win!” He linked it to a blog by Travis Vigil, Dell’s SVP for Product Management in its Infrastructure Solutions Group (ISG).

Vigil’s blog was headlined by this statement: ”Undisputed: Dell Technologies Ranked #1 in All-Flash Storage Vendor Revenue,” and followed by this one: “Dell Technologies is the leader in all-flash storage vendor revenue per IDC.” A chart illustrated the claims:

It refers to calendar Q2, 2025, and is backed up by IDC’s WW Quarterly Enterprise Storage Systems Tracker, 2025 Q2 historical release, September 11, 2025. That would be roughly equivalent to NetApp’s first fiscal 2026 quarter, ended July 25, 2025, when its AFA revenues were $893 million. If that represents 16.9 percent of total vendors’ AFA revenues in the quarter, then Dell’s 23.7 percent equals $1.25 billion and Huawei’s 13.5 percent equals $713.3 million.

This means that Dell regained the number 1 AFA market revenue share position in the second calendar quarter after losing it to NetApp in the first quarter. NetApp’s calendar first quarter is roughly equivalent to its fourth fiscal 2025 quarter, ended on April 25, when its AFA revenues were $1.03 billion.

Druva rolls out AI agents to mine backup metadata in real time

By
Chris Mellor
-

SaaS data protector Druva is providing new AI agents that can query aggregated backup data in real time to get summaries of risks, anomalies, and trends, and speed and simplify backup data management.

It says that historically, “backup intelligence has been limited to static dashboards and siloed reports.” In many environments today, it says, teams must extract, transform, and load backup metadata into separate systems before it can be analyzed, which takes time and adds cost. These limitations can be swept away by using graph theory to map backup metadata and letting natural language-driven AI agents search it.

Jaspreet Singh

Jaspreet Singh, CEO and co-founder of Druva, states: “Backup has always been an untapped goldmine of intelligence, and Druva has steadily unlocked that potential with our growing family of AI agents. With Dru MetaGraph and the new DruAI agents, we’re taking the next step and giving every team the ability to query and act on their metadata. This democratizes access to backup intelligence, extending its value beyond IT and security to functions like compliance and legal.”

Druva already has a set of data, help, and action AI agents to enable customers to restore entire applications with one command, orchestrate full workload recovery, and eliminate the need to manually piece together infrastructure components. Now it has gone further by graph-mapping its metadata and having a couple of new agents access this graph structure to provide real-time backup management intelligence and actions.

The Dru MetaGraph metadata foundation aggregates all backup metadata – such as file attributes, permissions, and identity information – into a secure metadata layer combined with graph intelligence. It is not vectoring its metadata, realizing that vectors can’t capture the relationship between events and items in the backup metadata. In graph theory, entities (nodes) and their relationships (edges) are represented in a graph data structure. This can be analyzed with algorithms investigating the relationships and interconnections to uncover patterns, predict outcomes, and optimize processes. 

Druva DruAI infrastructure diagram

The two new DruAI agents use such algorithms to look into MetaGraph and analyze metadata relationships and context in real time. This can “help organizations pinpoint risks, streamline compliance reporting, and improve operational efficiency – all without moving or exposing data.”

Dru Insights Agent, with its real-time intelligence, distills complex metadata into clear, prioritized insights – summaries of the most important risks, anomalies, and trends, plus recommended steps to deal with them.

The Dru Lifecycle Agent enables admins to ask natural language queries to identify stale or non-compliant data, surface orphaned accounts before they become risk vectors, and enforce retention policies at scale. Customers can ask questions like “Show me data that is not compliant to PCI [Payment Card Industry] retention adherence.” The Lifecycle Agent translates these questions into queries against the MetaGraph and returns contextual answers, helping teams find things that need action and enable the action to take place.

DruAI agent user session example

DruAI can present the option for users to view results as a chart that can be pinned to their dashboard for future use.

Druva says admin teams can pull full audit trails, logs, and anomalies on demand to pinpoint threats, confirm impact, and trigger the right response instantly. They can detect unusual patterns, policy violations, or suspicious activity in real time then take immediate action to contain them. Teams can also view backup trends, check license usage, verify policy coverage in seconds, and make changes on the spot.

Dru MetaGraph and all DruAI agents run entirely within the Druva platform, analyzing metadata only, not the actual customer data. Each customer has their own isolated MetaGraph in their tenant with end-to-end encryption and compliance with global standards like FedRAMP, SOC 2, and GDPR. Customer data never leaves its boundary. Druva uses private retrieval-augmented generation (RAG) and isolated large language models (LLMs) to deliver accurate, context-aware answers without exposing information or sharing data across tenants.

The Insights Agent is generally available today. The Lifecycle Agent will be available soon. Look further into Dru MetaGraph here.

Qumulo and Cisco partner to store stranded enterprise data for AI

By
Chris Mellor
-

Scale-out filer Qumulo has partnered with Cisco to run its Unified Data Platform on UCS servers and integrate Splunk observability for streaming and cleaning machine data.

The two say that this deal “enables the consolidation of billions of files and petabytes to exabytes of unstructured data into a single, globally consistent network-attached namespace,” which “eliminates decades of stranded data, giving AI systems complete and instant access to the information needed for more accurate models and better outcomes.” The Data Platform software provides unified file and object storage in a global namespace that can run on-premises – at edge and core data center sites – or in the cloud, as Cloud Native Qumulo on AWS, Azure, Google, and Oracle.

Brandon Whitelaw

Brandon Whitelaw, Qumulo’s SVP of Product, stated: “This partnership makes it radically easier to gain real-time insight across complex environments by enabling AI reasoning across decades of unstructured data. Together with Cisco, we’re delivering a modern data platform that overcomes the ‘data gap’ by simplifying operations and managing data at a petabyte-to-exabyte scale across the edge, data center, and cloud while enhancing situational awareness with deep integrations into Splunk Observability Cloud.” 

Cisco-owned Splunk provides software that collects, analyzes, and visualizes machine-generated data from various sources, such as servers, networks, cloud environments, applications, and devices in real time. Qumulo says that when enterprise data is unified onto Qumulo and Cisco UCS, the Qumulo Data Platform natively streams OpenTelemetry data directly into Splunk Observability Cloud. Customers can then get “visibility into real-time troubleshooting to drive faster anomaly detection, helping strengthen resilience against threats to sovereign data.”

Key workloads and data include medical imaging, signals intelligence, autonomous driving telemetry, life sciences research, geospatial mapping and imagery, video surveillance, and enterprise document management. The AI focus is on inferencing and not training.

This is an on-premises deal, for now, and different from the converged infrastructure (CI) deals Cisco has with NetApp (FlexPod), Pure Storage (FlashStack) and VAST Data – each of which include Cisco Nexus networking. These are rack-scale hardware and software deals.

Qumulo’s Unified Data Platform is now available on Cisco’s global price list, and through Cisco’s global channel. It includes pre-validated, jointly supported on-premises architectures with future support for Cloud Native Qumulo deployments in all major public clouds. Download a Unified Data Platform solution brief here.

Nvidia bids $900M+ for server fabric startup Enfabrica

By
Chris Mellor
-

Nvidia is buying GPU cluster interconnector and server memory offload appliance developer Enfabrica in a deal valued at more than $900 million. Multiple sources, such as CNBC and Reuters, are reporting this deal, with Nvidia acquiring Enfabrica, its team, and its intellectual property in a cash and stock deal.

From left, Rochan Sankar and Shrijeet Mukherjee

Enfabrica was founded in 2019 – officially incorporating in 2020 – by CEO Rochan Sankar and chief development officer Shrijeet Mukherjee to develop a massively distributed, Ethernet-based, server fabric interconnect ASIC and associated software. 

Its Accelerated Compute Fabric Switch (ACF-S) chip provides a multi-terabit and elastic compute fabric to interconnect hyperscalers’ scaled-out and hyper-distributed servers running AI and machine learning applications, and combine PCIe and CXL technology with RDMA-NIC functions. PCIe and CXL use memory semantics whereas Ethernet and RDMA links use a networking protocol, and Enfabrica bridges the two domains.

Enfabrica has raised a total of $290 million in VC funding – a $50 million A-round in 2022 with a $50 million valuation, $125 million B-round in 2023 with a $250 million valuation, and a $115 million C-round in 2024, with its valuation now around $600 million, at least according to Pitchbook. Nvidia invested in the B-round.

The EMFASYS (Elastic Memory Fabric System) chassis, built around the ACF-S chip, was announced by Enfabrica in July, providing elastic memory bandwidth and capacity to interconnect more than 100,000 GPUs and their limited high-bandwidth memory (HBM) capacity.

Enfabrica’s EMFASYS chassis

NVLink is Nvidia’s high-bandwidth, GPU-to-GPU interconnect networking technology, used within GPU servers and racks. Enfabrica’s IP can add a rack-scale disaggregated memory fabric. We could envisage a future ACF-S development adding NVLink Fusion IO chipsets to further speed system interconnectivity.

Nvidia has just invested $5 billion to buy a stake in Intel and $700 million in Nscale, a datacenter startup building a UK Stargate center with OpenAI and Nvidia. It bought Israel-based Run:ai and its GPU orchestration software for $700 million in 2024. Nvidia is also involved with Kioxia in developing a 100 million IOPS AI SSD with CXL features.

Pure: ignoring data sovereignty concerns can cause unwanted financial loss

By
Chris Mellor
-

The rise in data sovereignty requirements and risks mean that country, regional and global organizations need to be aware that they they could face revenue loss, financial penalties, and reputationally damaging loss of trust if they break data sovereignty regulations.

Pure Storage has run a study with academics from the University of Technology Sydney (UTS) which found that geopolitical uncertainty and regulatory evolution mean data sovereignty has moved from being a compliance issue to one that can affect revenues, competitiveness, and customer trust.

Alex McMullan.

Alex McMullan, International CTO for Pure Storage, stated: “The potential consequences of not having a modern and realistic data sovereignty strategy are acute. Loss of trust, financial damage and competitive disadvantage are possible outcomes that cannot be ignored. We recommend a hybrid approach to data sovereignty: start with a risk assessment across workloads, keep critical workloads sovereign, and use the public cloud for the rest.”

Data sovereignty is about who has the legal authority to access and govern data, regardless of where it is actually stored. Data residency is about where the data is physically stored. For example, the EU’s General Data Protection Regulation (GDPR) enforces both data sovereignty and residency, requiring that EU Residents’ personal data is stored and processed within specific geographic locations or under adequate safeguards, regardless of whether it’s handled outside the EU. On top of that, there are 60 articles in the EU’s Digital Operational Resilience Act (DORA) regulations applying to financial sector operators in the EU territory. Article 12 requires secure, physically and logically separated backup storage. 

US-based public clouds, like AWS, Azure and the Google Cloud, were built with a global reach and without any country or regional data sovereignty requirements in mind. Nowadays, such data sovereignty requirements can mandate that certain kinds of data must be physically stored within a geographic territory and not be moved outside.

There are US data sovereignty regulations. For example, US goverment Executive Order 14117 (2024) and DOJ Final Rule (Effective April 8, 2025) prohibits or restricts bulk transfers of”sensitive personal data” (e.g., genomic, biometric, health, financial, or precise geolocation data) and “government-related data” to “countries of concern” (China, Cuba, Iran, North Korea, Russia, Venezuela) or associated persons. Another example: the Health Insurance Portability and Accountability Act (HIPAA) requires protected health information (PHI) to be stored, accessed, and transmitted in ways that ensure US jurisdictional control, often implying domestic handling for compliance.

Data sovereignty concerns are rising in the current AI era because countries and regions want their own, local AI competency and not to be dependent on US-based providers subject to the somewhat capricious effects of Donald Trump’s presidency with, for example, tariff rises and data access requirements. Dan Middleton, VP UK and Ireland, Keepit, tells us that: “The US CLOUD Act allows authorities in that country to compel US organizations to hand over data, regardless of where that data is stored. This creates an unavoidable conflict between US law and European data protection laws, forcing US companies to choose between criminal liability in their home country or violating European data protection regulations.”

The Pure/University of Sydney study relied upon qualitative interviews with experts and practitioners from across industry and the research sector in nine countries between July and August 2025; Australia, France, Germany, India, Japan, New Zealand, Singapore, South Korea and the United Kingdom.

The study found that;

  • 100 percent of respondees confirmed sovereignty risks, including potential service disruption, have forced organisations to reconsider where data is located
  • 92 percent said geopolitical shifts are increasing sovereignty risks
  • 92 percent warned inadequate sovereignty planning could lead to reputational damage
  • 85 percent identified loss of customer trust as the ultimate consequence of inaction
  • 78 percent are already embracing different data strategies, such as implementing multi service provider strategies; adopting sovereign data centers; and embedding enhanced governance requirements in commercial agreements

Gordon Noble, Research Director at UTS’s Institute of Sustainable Futures, said: “These are wake-up call numbers. Every single leader we interviewed is rethinking data location. The message is clear: sovereignty is no longer optional, it is existential.”

The AI angle has increased interest in data sovereignty. On Asia-based AI governance, a respondent said: “AI is fast becoming a geopolitical force, and data centres are the chess pieces on the board. Decisions about where to build, how to power them, and their environmental impact are increasingly central to global negotiations.”

Pure suggests that organizations adopt a data sovereignty process that analyses the risk landscape to identify which services and data sets are most critical and sensitive, place these workloads in sovereign environments, while leveraging public cloud for less crucial functions. It says that this “enables organizations to maintain compliance and control without sacrificing the innovation and agility that organizations need in order to remain relevant in today’s fast-paced business environment.”

Archana Venkatraman, Senior Research Director, Cloud Data Management, IDC Europe, said: “We expect to see data sovereignty treated as a strategic priority in 2025 and beyond.”

Here at B&F, we expect an increase in territorial data residency requirements for sensitive data sets.

Read more in a McMullan blog. Download the research report information here.

Elastio detects real-time, zero-day, malware backup infection

By
Chris Mellor
-

Elastio’s Model Context Protocol (MCP) Server embeds agentless ransomware detection and backup validation directly into AWS workflows, developer tooling, and AI assistants, with claimed zero-day detection capability.

Najaf Husain

The company was founded in 2020 by CEO Najaf Husain and CRO Todd Frederick to combat zero-day ransomware threats that evade traditional defenses by cloud-native SW running agentless detection, malware scanning of backups, automated recovery testing, and immutable storage to enable rapid, clean restores after malware detection. Husain previously founded data center composability business Cloudistics whose assets were bought by Fungible in 2020. Husain started up AppAssure before that and it was acquired by Dell in 2012. Elastio started out with a seed round in 2020 and then raised $18 million in a mid-2023 A-round.

Greg Aligiannis, CISO at Elastio, stated: “With the MCP Server, we bring detection, validation, and compliance-ready reporting straight into the environments teams already use.”

Greg Aligiannis

The MCP Server has been developed with the idea that modern enterprise operations will increasingly depend on agentic AI workflows, autonomous systems where AI agents reason, act, and adapt with minimal human oversight. These dynamic workflows orchestrate, correct, and recover in real time but need real-time cyber-resilience. Elastio claims its MCP Server makes incident response, resilience, and recovery an integrated part of agentic workflows and enables self-healing.

It provides:

  • Managed cyber resilience: Continuously monitor backups, restores, deployments, and files, directly within IDEs, AWS workflows, and chat-based AI assistants,.
  • Agentic and extensible: Integrates across ecosystems as MCP delivers resilience data into agentic tools and platforms, exposing compromised data caused by ransomware, misconfigurations, and optimization opportunities in real time.
  • Incident response with real-time detection and live context: Gain continuous visibility at the asset, volume, and file level, identifying threats as they emerge and delivering live context through AI assistants to accelerate response and “guarantee uncompromised recovery.”

Which backups does it constantly monitor and how does it identify them? The two main targets are AWS and Azure. On AWS, it says it monitors recovery points for EC2 instances, EBS volumes, EFS file systems, S3 object storage, and VMware workloads created via AWS Backup. It also inspects snapshots and data in logically air-gapped vaults. There are integrations with Cohesity, Commvault, NetBackup, Rubrik, and Veeam backup, allowing validation of their backup data for ransomware and malware presence.

Elastio scans AWS accounts and integrated environments to detect new workloads, resources, and backup instances dynamically via IAM roles, SQS messages, and Lambda invocations, with no direct VPC connectivity. This triggers inspection of eligible backups without agent installation. Backups are identified through API integrations with AWS Backup and third-party tools, correlating them to specific assets (e.g., EC2/EBS recovery points). It uses metadata like resource IDs, timestamps, and storage locations to catalog and track backups.

Elastio dashboard example.

We understand that backups are scanned using RansomwareIQ, Elastio’s data integrity engine, which employs machine learning for behavior-based detection (beyond signatures or anomalies). This includes heuristics, detonation in isolated environments, and forensic mounting (read-only) to verify integrity, quarantine infected points, and pinpoint the last clean recovery state. RansomwareIQ mounts backups in a secure, read-only environment and uses ML-driven analysis to detect ransomware signatures, encryption anomalies, and behavioral red flags. Infected recovery points are flagged and isolated, preventing restoration of compromised data. Then it provides detailed reports on threat exposure, clean recovery points, and compliance status, accessible through Elastio’s unified dashboard.

The MCP Server will also soon integrate with Azure Backup, Azure Site Recovery (ASR), Azure Blob Storage, Azure VMs, and Azure Block Storage, and has full support for Veeam backups stored in Azure. It says it will inspect Veeam backups stored in Azure Blob Storage, “detecting encryption patterns and insider threats that often slip past antivirus or EDR tools.” Request an preview of Elastio on Azure here.

The MCP Server can provide monitoring and clean recovery from third-party on-premises backups, where integrated. We understand that these include Cohesity DataProtect and (Veritas) NetBackup, Commvault Complete Data Protection, Rubrik CDM, and Veeam Backup & Replication.

Existing Elastio customers include the Boston Celtics, healthcare IT provider Cloudwave, legal tech SaaS company CARET, and a global payment business handling >40 billion transactions annually.

The Elastio MCP Server for AWS is available today, with installation guides and documentation. Continuous feature updates and integrations will be released via AI-assisted channels. Read more about Elastio’s software in its blog and solution brief docs.

Cohesity is protecting identities and Active Directory

By
Chris Mellor
-

Cohesity announced five new cyber-resilience capabilities covering the top 3 public clouds, identity, AI-native data security, data sovereignty and its Gaia AI-based knowledge discovery assistant.

It declares there are five steps needed to boost cyber-resilience; protect all data, ensure data is always recoverable, detect and investigate threats, practice application resilience, and optimize data risk posture.

Sanjay Poonen.

CEO Sanjay Poonen stated: “Identity is the new battleground, and enterprises need solutions that combine prevention, resilience, and recovery without adding complexity to their existing cyber resilience strategy. Cohesity Identity Resilience delivers a modern, proven approach that helps organizations protect the foundation of their digital operations, so they can recover quickly from even the most critical attacks and remain cyber resilient.”

Cohesity has a library of connectors that it says supports >1,000 datasources. It plans to add 40 additional connectors by the end of 2025, including the most popular compute, container, storage, and database services across AWS, Azure, and Google Cloud.

It is partnering with cyber-security company Semperis, with its automated forest recovery and identity protection tools, to provide Cohesity Identity Resilience protecting Active Directory. This helps users protect and secure on-premises Active Directory and extends these safeguards to Microsoft Entra ID in the cloud. Overall, customers get;

  • Proactive Active Directory hardening – The ability to scan on-prem or hybrid Active Directory environments for hundreds of indicators of exposure (IOEs), see risk scores, apply remediation guidance, and identify and prioritize attack paths that bad actors could use to reach critical privileged assets (Tier 0).
  • Secure, immutable backups with capability for cyber vaulting – Enables consistent backup and recovery workflows while providing isolation of AD backups.
  • Rapid and secure recovery – Restores Active Directory forests after a cyberattack or catastrophic failure, ensuring minimal downtime.
  • Specialized identity forensics and incident response capabilities – Cleanses the system of attacker artifacts, ensuring Active Directory is restored to a trusted environment.
  • Comprehensive post-breach analysis – Offers comprehensive post-attack forensic support, including rapid assessments within defined incident windows, and validation of Active Directory integrity before restoring production systems.

A new NetBackup DirectIO feature enables NetBackup to write backups to the Cohesity Data Cloud target. All NetBackup data sources can have their backups stored in Cohesity’s environment to get faster restoration and, Cohesity says, up to 53 percent in direct cost and storage efficiency savings. 

Cohesity is now able to provide an on-premises installation of its FortKnox cloud-based vault called the FortKnox Self-managed option. It enables customers  with data sovereignty requirements to implement an isolated, secure virtual air-gapped vault within their own data centers.

It is adding NetBackup’s “ultra-fast, hash-based threat scanning” to its DataProtect product to deliver “near-instant search results for indicators of compromise” across customers’ data. Cohesity will also incorporate Google Threat Intelligence into Cohesity Data Cloud’s threat scanning capability, at no extra cost, as part of its Cohesity Enterprise edition.

As a way of practising app resilience, Cohesity is introducing a cyber recovery orchestration tool, RecoveryAgent. This automates recovery testing, rehearsals, and recovery execution. It has embedded malware scanning and agentic AI capabilities. By using RecoveryAgent, customers can organize recovery workflows and forecast recovery timelines with some precision.

An integration with Cyera embeds additional data classification and governance capabilities directly into the Cohesity Data Cloud platform. It will enable customers to identify sensitive and regulated data within backups, eliminate redundant or obsolete data, and enforce compliance requirements in near real-time to avoid risks like sensitive data restoration to unauthorized locations.

The AI-driven Gaia assistant, available for both cloud and on-prem environments, gets a new search interface, sensitive data redaction, multi-language support, and integrations with Slack and Google Agentspace.

Lastly Cohesity is expanding its CERT (Cyber Event Response Team) offerings with new cyber resilience consulting services.

Download its 5-step cyber-resilience eBook here. Visit the Catalyst1 website for session replays and on-demand content.

Bootnotes

Protecting data everywhere includes mainframe data by definition. Cohesity protects mainframe data through integrations with partners like Model9 and Luminex, enabling backup, archiving, disaster recovery, and ransomware protection for mainframe environments (such as IBM z/OS).

An Active Directory forest is is the top-level structure that represents a collection of one or more domains – groups of objects (users, computers, etc.) – that share a common schema, configuration, and global catalog, forming a single security and administrative boundary. Domains in a forest can be organized hierarchically into trees.

SingleStore sidesteps into private equity ownership

By
Chris Mellor
-

Unified database startup SingleStore has been bought by private equity house Vector Capital Management instead of raising more cash or running an IPO.

SingleStore provides a combined transactional, analytical, and vector database that can run in memory, supports external storage, operates with low latency on-premises or in the public cloud, and has a growing AI focus. It has just announced strong second quarter fiscal 2026 results with ARR of over $123 million, growing 23 percent year-on-year. Total net new ARR grew 34 percent Y/Y. It has free cash flow for the last 12 months of negative ~$2 million, within touching distance of break even. At the end of the period, the company had more than $150 million in cash and zero debt, providing a robust balance sheet for continued growth both organic and inorganic. 

Raj Verma.

Raj Verma, CEO of SingleStore, stated: “We are seeing clear evidence that our strategy and execution are working. Customers are scaling with us, retention is world-class and we’re operating with outstanding financial discipline. Our pipeline is deeper and healthier than ever, and I have tremendous confidence and excitement for what the future holds for us.”

Against this background why accept a private equity takeover when an IPO was being discussed after a 2022 $116 million funding raise with a $1.3 billion valuation?

Amish Mehta, Chief Investment Officer and Managing Director of Vector Capital, said, “Our investment in SingleStore underscores Vector’s commitment to investing in category-defining technology companies … As the world’s best-fit database for applications which demand ultra-low latency with transactional and analytical requirements, we believe SingleStore has the potential to become one of the most important data platforms for the AI era and are proud to support its next phase of growth alongside an exceptional group of investors and shareholders.”

Verma, staying on as CEO, said: “Building a database natively engineered for the AI era, well before the moment arrived, has been the work of more than a decade. … Now, as we look to accelerate our growth trajectory, I am delighted to partner with Amish, Stephen, and the Vector team who share our passion for innovation and our vision for what is possible to achieve when you start with an AI-ready database of unparalleled speed, scale and simplicity. I look forward to working with Vector on the next wave of innovation and further accelerate SingleStore’s growth to best serve our customers.”

We learn that Vector has bought a majority stake, speculated to be around $500 million according to Bloomberg, and that long-term SingleStore shareholders including Google Ventures, Dell Technologies Capital, IBM, and REV Venture Partners will remain as investors. Several Vector Capital limited partners will invest alongside Vector itself. A Vector MD, Stephen Goodman, said: “SingleStore represents the largest new platform investment Vector has made in over 15 years. … We are excited to partner with Raj and the SingleStore team to build on its strong momentum, expand the Company’s global reach and continue delivering the real-time, AI-ready data solutions that today’s leading enterprises require.”

Vector’s typical playbook is to buy into middle- or late-stage growth companies and bulk them up with innovative developments, expanded go-to-market, financial discipline and possible bolt-on acquisitions to build a stronger business it can sell on or make public a few years later.

How did SingleStore get to this point?

History

SingleStore was started up as MemSQL in 2011, to provide an in-memory SQL database, by Adam Prout (CTO), Eric Frenkiel (left 2017), and Nikita Shamgunov (left 2022 to be Khosla Ventures partner) in San Francisco. It created a high-performance, distributed SQL database that could handle both transactional (OLTP) and analytical (OLAP) workloads in real time, addressing the limitations of traditional databases.

MemSQL developed Helios, a public cloud version of its database in 2019, and Raj Veerma became CEO the next year. It added support for external drive-based columnar storage in 2016, increasing the amount of data it could store, and worked with Apache Kafka and Spark to support real-time data pipelines.

The public funding history was solid and stable up to 2020 when there was an acceleration from the previous $30 – $36 million rounds to an $80 million E-round plus $50 million in debt finance followed by a second $80 million round in 2021, with it having a $1.3 billion+ valuation, and a $116 million round in 2022. The total raised was said to be $464 million; not all raises were public. Then fund raising stopped, until now with the buyout transaction. 

AI became an increasing focus from 2023 onwards with vector support added in 2024.

More and more VC focus has been placed on AI companies and AI data storage and analysis companies such as Snowflake, Dremio and Databricks. They can all be described as data platforms for the AI era.

Competitors

SingleStore’s main competitors are Snowflake, PostgreSQL, MySQL and MongoDB. It also competes with public cloud databases; Amazon Redshift, Google Cloud BigQuery, Azure SQL Database, and the Oracle Database.

Cloud data warehouser Snowflake was started up in 2012, a year after MemSQL, and grew very quickly, with its funding rate rapidly overtaking that of Snowflake to reach a total of $1.4 billion in 2020 when its G-round raised $479 million, more than SingleStore’s total funding, and then it ran an IPO in September of that year.

As well as being out-competed by Snowflake, SingleStore had to compete with the rise of overlapping data lakes and lakehouse, notably with Databricks, which grew its funding at an enormous rate, more and faster than Snowflake, with $20 billion total funding, $10 billion of that alone raised in 2024, and a $100 billion valuation. That makes SingleStore fundraising look anaemic in comparison.

Snowflake revenues are now at the $1 billion/quarter level and growing at 30 percent Y/Y. SingleStore revenues are a fraction of that; its $123 million ARR translates to around $30 million/quarter, and it wouldn’t look that impressive as an IPO prospect. We think the VC investors’ telescope is focused more on the big AI analytics companies rather than the AI database suppliers like SingleStore. 

Somehow the flurry of growth and investment in the 2020-2022 period has slowed down and SingleStore found its growth prospects limited, a near-term IPO unlikely to succeed, and a private equity buyout become more and more attractive.

It is an excellent company that needs more development as this AI era progresses. Vector and its partners will provide the cash and market expertise for that and we’ll see where this takes SingleStore in the next few years.

Identity protection from Rubrik and CrowdStrike

By
Chris Mellor
-

Rubrik has done a deal with CrowdStrike and its Next-Gen Identity Security so that customers can reverse malicious identity changes and restore identity providers back to a safe state.

Rubrik has already done a deal with CrowdStrike to send data to its malware-detecting Falcon XDR (Extended Detection and Response) and look for threats in Rubrik backup data. Commvault has done a similar deal with CrowdStrike. Now cyber-security concerns are focussing on identity, as malware attackers may well log into a victim organization’s websites after having gained access credentials, by phishing attacks for example. Rubrik’s Identity Resilience offering can work bi-directionally with CrowdStrike to roll back malicious changes and return identity systems to a secure, immutable state.

Rubrik’s Chief Product Officer, Anneka Gupta, stated: “By expanding Falcon Next-Gen Identity Security with rollback and recovery, we’re giving customers a complete solution – detect, adapt, and reverse – that minimizes disruption and keeps operations running in the face of identity-based threats.”

Daniel Bernard, Chief Business Officer at CrowdStrike, matched this, saying: “Enterprises need more than detection – they need identity security that can adapt, defend, and outpace today’s adversaries. Together with Rubrik, we’re delivering unified identity security that combines CrowdStrike’s AI-driven protection with Rubrik’s rollback innovation. The result is simple: customers stop identity attacks faster, minimize business disruption, and strengthen resilience across hybrid environments.”

Rubrik Identity Resilience (RIR) integrates identity security with risk detection and threat response. Customers can monitor identity changes in real time for forensic analysis, improve collaboration between security and IAM teams, and continuously assess identity risks across human and non-human identities. Rubrik says RIR  “can quickly identify and address overprivileged or misconfigured accounts before they are exploited.”

CrowdStrike’s Falcon Next-Gen Identity Security protects an accessing entity’s identity, whether it be human, non-human machine or app, and AI agent , across an identity lifecycle, blocking initial access, preventing privilege escalation, and stopping lateral movement.

Rubrik Identity Resilience screenshot.

The combined RIR-CrowdStrike integration provides:

  • Real-time identity threat detection and malicious change correlation. Falcon Next-Gen Identity Security provides AI-driven correlation of suspicious changes across identity providers (IdPs) such as Active Directory, Entra ID, and Okta. Rubrik ingests those alerts and identifies the malicious changes made by the compromised identity. 
  • Rollback of malicious changes. Rubrik Identity Resilience recovers and rolls back malicious actions, restoring them to a known, safe state, leveraging immutability to prevent re-exploitation. In a worst case scenario, Rubrik delivers a full, clean IdP recovery. 
  • Accelerated investigation and workflow in Falcon console. With Rubrik Security Cloud’s integrations with Falcon Fusion SOAR, Next-Gen SIEM, Falcon Threat Intelligence, and Charlotte AI, security teams can streamline the investigation and response process by initiating rollback actions, tracking completion, and orchestrating recovery workflows.

RIR is generally available. The Rubrik-CrowdStrike identity resilience integration is available on the CrowdStrike marketplace. Find out more about RIR and its CrowdStrike partnership here. Download a Rubrik Identity Resilience datasheet here.

HyperBUNKER touts diode-based offline vault for ransomware recovery

By
Chris Mellor
-

HyperBUNKER provides a backup of backups, and air-gapped, offline, multiple immutable data copies as a last resort for solidly dependable critical data recovery.

Data is written using data diode technology to a target device operated by PLCs (programmable logic controllers). It has its own backup software for file storage and is thus an addition to existing backup software. The data diode technology enforces a one-way, data-in communications path to the HyperBUNKER vault device. This technology is commonly found in highly-secure power grids, water utilities, nuclear facilities, and military installations, but has not been seen in traditional enterprise data protection and cyber-resilience products. The company was founded as an offshoot of data recovery business InfoLAB in Zagreb, Croatia, by CEO Bostjan Kirm and CTO Imran Nino Eškić. Nino Eškić is InfoLAB CEO.

Investor and advisor Matt Peterman told us why he became appreciative of the HyperBUNKER concept: ”You see servers and drives shipped in [to InfoLAB] from all over Europe – companies locked out of their own data. And why does this happen if they have perfect cyber protection tools? Sometimes, it is due to hardware failure, and often due to ransomware. And in those ransomware cases, Nino could do very little except suggest negotiating through brokers. That frustration is what pushed him to design an offline protection that actually preserves the most critical data.”

HyperBUNKER chassis

The key tech is the data diode idea, applied while a backup is being written to the HyperBUNKER rack-shelf chassis’ SSDs or disk drives. The company says this is disconnected from the external networks, such as the internet. Its software performs a comprehensive backup of entire hierarchical folder structures, such as a NAS system, preserving all subfolders and their contents. During recovery, which uses a separate access path, the original structure is fully restored. 

The device uses optocoupler-isolated monitoring, whose signals prevent unauthorized access. It says its tech remains invisible within network infrastructures, making it undetectable and unreachable to online hackers.

From left, HyperBUNKER COO Denis Eškić, CTO Imran Nino Eškić, and HW/SW engineer Hrvoje Marošević. They are holding the HyperBUNKER device

Peterman tells us: “This is not a new technology. But in practice, traditional diodes are one-way network devices. They still rely on protocols, handshakes, and ultimately leave the protected side online. One bad update or exploited protocol – and the vault is exposed. Three years ago, Nino tested diode systems extensively, then developed a different approach: optical isolation and butlering logic. That’s what HyperBUNKER patented in October 2024. It means data is kept physically offline at all times, no dependency on network protocols, no exposure to handshake exploits.

He says there are public examples of existing cyber-protection failing: “Public cases back this up. Veeam client breaches, even NHS repeat failures. Tools passed ‘compliance,’ but when tested by real incidents, backups collapsed or were bypassed. See also recent M&S and Jaguar case.”

Peterman said HyperBUNKER tracks “65-plus ‘cyber protection’ tools in the market, but ransomware is still growing.”

HyperBUNKER screenshot

“From my background in insurance fraud detection (my previous investment), the contrast is striking. In fraud, we always knew why a case failed, whose fault it was. In a cyber attack, postmortems are often not revealed or shared. The more complex the stack, the more likely a human or config error opens the door. HyperBUNKER is built for simplicity: if my most critical data is in my desk drawer, there’s very little chance a North Korean actor can lock it.”

HyperBUNKER admits “the only vulnerability is the physical theft of the device. However, by distributing HyperBUNKER units across multiple locations and encrypting all stored data, the risk of data compromise is minimized.” We asked HyperBUNKER some questions.

B&F: What is the data diode technology about?

HyperBUNKER: Inbound writes traverse optocouplers that physically enforce one-way flow. Data can enter the vault but nothing electronic can come back out. There is no logical bypass.

B&F:What are the dedicated PLC electronics used for storage management?

HyperBUNKER: Independent micro-PLC controllers govern drive power states, media rotation, and write windows. They are air-gapped and cannot be reached or reprogrammed over a network, so ransomware cannot flip modes or trigger erasures.

B&F: When existing backup software can write to tape libraries, with offline tape cartridges, why do I need HyperBUNKER?

HyperBUNKER: Tapes are only “offline” once ejected. During write or restore the library and control plane are online with credentials exposed. HyperBUNKER’s vault media are never network-addressable. Inbound is one-way optical so there is no session to hijack and no remote delete. And restoring data from tapes can last for a while.

B&F: If I use HPE Zerto why do I need Hyperbunker?

HyperBUNKER: Zerto is strong software but it lives on the network and depends on credentials, updates, and policy. HyperBUNKER sits outside that plane as a last-resort, physically isolated vault when software defenses are breached.

B&F: What prevents the HyperBUNKER backup software backing up already corrupted data? How do you know the backup is clean?

HyperBUNKER: HyperBUNKER does not try to detect malware or decide what is clean. It is content-agnostic by design. Protection comes from physically enforced immutability and spaced generations. New writes cannot alter or delete earlier sets. If a poisoned write reaches the vault, it lands as a new generation, and recovery means selecting a pre-event generation upstream. But you can do numerous “restores” of data in a virtual environment and remove malware. That’s what Nino is quite often doing in his data recovery labs in Verona and Zagreb.

B&F: How is HyperBUNKER priced? By capacity? What is its capacity?

HyperBUNKER: Appliances are priced by usable capacity, not per-seat or license. Current ranges are 1 TB and 4 TB configurations. Typical entry deployments start around €50,000.

Bootnote

HyperBUNKER incorporates encryption, access controls, and regular audits to meet GDPR and other privacy standards. It employs a proprietary version of File Security Encryption (FSE), meets HIPAA standards for protecting electronic protected health information (ePHI) and complies with PCI DSS requirements, ensuring secure storage for payment card information with encryption and restricted access. HyperBUNKER aligns with the NIS2 Directive. A rugged, military-grade version with EMP/EMI protection is planned.

1...8910...470Page 9 of 470

ABOUT US

Blocks & Files is a storage news, information and analysis site covering storage media, devices from drives through arrays to server-based storage, cloud storage, networking and protocols, data management, suppliers and standards. We publish news, views, opinions and analysis of storage industry events and developments, aiming to keep storage users, developers, consultants, marketeers, salespeople, support staff, analysts and everyone else involved in IT storage up to date with what’s going on.

FOLLOW US

© Situation Publishing, 2018-2024